The Unwritten Book is Finally Written!
An in-depth analysis of: The sacrifice bunt, batter/pitcher matchups, the intentional base on balls, optimizing a batting lineup, hot and cold streaks, clutch performance, platooning strategies, and much more. Read Excerpts & Customer Reviews
Interesting article in Science News: "Odds Are It's Wrong: Science Fails to Face Its Statistical Shortcomings": http://www.sciencenews.org/view/feature/id/57091/title/Odds_are,_its_wrong
Well I imagine that we'll only know for sure if he publishes his comment. I'd surmise that he was calculating this from a game theoretic perspective (he is a card player, after all). My own experience is more in statistical testing, in the sense that I described.
But the point stands, I think, that taking the "safe" option will not get a coach in trouble. Deviating from it -- taking the risk of doing so -- requires that he have something more than just intuition to go on, but instead some other factor(s) that research or experience shows can shift the odds in a predictable way.
Mike: take a look at the detailed summary of Nate's argument on the blog.
If the "accepted wisdom" says you should choose A over B because A will work 2/3 of the time, are there any conditions when you should choose B?
The accepted wisdom is safe, defensible, and won't get a coach in trouble if it fails 1/3 of the time.
Nate calculates that choosing B will require information that increases its odds of success from the base 1/3 probability.
So the question is how much better odds than the base 1/3 success rate do you need to justify choosing B? And more importantly (but not stated in the summary that is linked), what kind of information is that?
To me this boils down to a standard issue in multivariate analysis, where you have to decide whether adding another variable to a regression model will improve the prediction enough (perhaps allowing for interaction among predictor variables) to justify the additional complexity -- and let the coach off the hook for defying the conventional wisdom.
It can't just be a coach's intuition, but most likely an assessment of contextual information of some kind (e.g., injured key player on the other team; offense/defense mismatch of some kind) that is not just randomly related to the outcome. WHEN do you "go for it" on 4th down rather than punt? When do you pull your goalie? (me: when you know there's a 95% probability you will lose and only such a desperate move gives you a significantly better chance of winning. I don't know the real odds here, but Tango probably does.)
So while I think Nate's logic is good, it doesn't answer the question when you should go against the conventional wisdom. Of course, that's what research is all about -- finding holes in the conventional wisdom by either improving measurement (of performance indicators) or looking for other predictive factors over outcomes that are also practical for managers/coaches to follow.
So those authors knew of other estimates but didn't even cite them. For example, in his "Ruth vs. Bonds" intro chapter in Baseball Between the Numbers, Nate Silver estimated that if Bonds had begun his career in 1916 (when Ruth did) he'd have had 444 career home runs. And if Ruth had begun his career in 1984 (when Bonds did), he'd have hit 913 homeruns in his career.
This is a vastly different set of numbers from what the authors of this article came up with -- yet they cited the book without citing Silver's estimates. Not saying Nate's right and they're wrong, but the authors' had an obligation to read the lit.
Agree that the bibliography in the paper is weak, and misses some preceding studies specifically asking "What do statistics tell us about steroids"? Nate Silver has an article with just this title in Baseball Between the Numbers, a book that the authors of this study cite but strangely attribute to James Click. He also wrote a chapter comparing Bonds to Ruth in that same book.
Doing it with # of fans is the most reasonable, it seems to me. When you use population size (in the metro area) as a predictor, you can also make an adjustment for whether there are competing ML teams in the region (LA, NY, Chicago, specifically). Don't split the population, just put in a dummy for "two teams".
"Making the playoffs the year before raises a .500 team's attendance by about 3,000 fans per game - a major boost. Obviously making the playoffs raises hype around the team, and this appears to manifest itself in the form of increased attendance."
Making the playoffs the previous year, winning the league, winning the world series all matter (in a study I did some years ago). But it's probably not just a matter of "raising hype." It's a matter of raising season ticket sales. Those sales are probably driven much more by previous season's record (specifically making playoffs, etc.) than by current.
So a refined analysis -- if data is available -- would look for lagged effects of wins (making playoffs, etc.) on season ticket sales and for in-season effects of current wins on walk-up or game-day sales.
An interesting but somewhat curious study. I can't follow the math but I can follow the logic.
Despite the title, it says nothing about steroids per se, as opposed to era in general.
(I'll leave aside the fact that they chose RBI as one indicator.)
I find it curious that they decided to scale the HR so that "The Babe" ended up with over 1200 HR. Maybe they could have anchored the baseline curve to the HR levels in the 20's and 30's and shrunk the more recent levels more than they did?
For the type of work that Clay has done with NOAA, there is no particular storm season for his job -- since he works on the entire globe not just the northern and western hemispheres.
Still, this is looking better in the sense that higher variation (less consistency) in either scoring or being scored upon is associated with a lower win probability.
Now you can try to add back one of the "enviromental variables," let's say RS only. That would capture the overall "level" of scoring. Or you could put in total runs in game or total runs per inning to capture the same effect.
I'm not even sure that you would end up needing the RG and RAG variables in a final equation of the type I've suggested, but they would in principle be interpretable.
The coefficient on runs per game may fall because it's correlated with the standard deviation of runs per game.
Another way to represent the variance is to use a coefficient of variation (CV), which is the standard deviation divided by the mean. In effect this "norms" the standard deviation -- its size is relative to the run environment.
I think you could try substituting CVRG and CVRAG for the two std. deviation variables that you now have. And then I wouldn't be surprised if the coefficients for RG and RAG became insignificant in such a setup, but that would be perfectly fine, because the CVRG and CVRAG would capture the run-environment-normed effects of variation/consistency in runs scored and allowed.
@salb#17: I get your point about the interpretation of the Pythag calculation. But as you know many prognosticators rely on the Pythag projected RS and RA to predict the number of team wins. Further, they take the difference between the Pythag projected wins and the actual wins at any point in the season (including the end) as indicating how lucky or unlucky a team has been.
However, this assumes that the difference between actual and Pythag wins is only due to luck. Suppose the ability to exceed the Pythag win projection were at least partly a "skill" or were predictable in some other systematic way. Perhaps a team's style (small ball vs. long ball) or its (unmeasured) baserunning skill (e.g., advancing on a ground out) were PERSISTENT. Suppose a particular team or manager showed an ability over several seasons to win more games than the simple Pythag projection.
In all of those cases, the difference between the Pythag and actual number of wins is not just due to chance. Rather, it's a "skill" (perhaps managerial skill or team style). And if it is persistent and means that, say, an otherwise Pythag 85 win team can be expected to win 89 games, then that's very meaningful for making projections of team wins.
So the question becomes, is the ability to exceed (or fall short of) a team's Pythag wins persistent? If so, then just like "consistency" (IF it is a consistent trait or habit of a team or manager), this ability (or inability) is not just a chance outcome.
(Additional question: I wonder also whether the return to "consistency" differs depending on the overall run environment.)
If there is no optimal strategy from a managerial standpoint, would there nonetheless be an advantage to taking a team's consistency in run production and run prevention into account as an adjustment factor in Pythaganpat estimates of W/L?
Just a gut intuition here but I suspect there's an analogy to stock investing. Those who try to "time the market" don't do as well as those who dollar cost average. In the case of stocks another factor is the transaction cost, but I think the logic still applies. You can't always predict what your opponent (or the market) will do next inning, and what they did the previous inning may not tell you that much (except on defense: change pitchers!). So do the little things right: keep investing steadily through thick and thin; don't rely on perfect timing or the home run.
This, BTW, is why I always feel when the Tigers are playing the Angels somehow the Angels are generating the extra run here or there (often via alert and aggressive baserunning). They accrete runs.
@Mike/#86.
(1) Re Bradbury: Colin criticized. Pretty much everyone else who commented criticized, including MGL and others from the outside. It was a thorough thrashing. I say "thank you BP" for helping to show its audience that the emperor has no clothes.
(2) Re did BP know there were problems with last year's PECOTA's? They read the criticisms; they know that PECOTA didn't perform nearly as well against the other systematic forecasts last year as it had in prior years. They don't have to have a big public ritual flagellation to expiate their sins. They just have to get to work to make sure things work better. There's plenty of evidence that they’ve tried to do that -- read Clay Davenport’s and Dave Pease’s comments, for example.
(3) What's wrong with PECOTA now? I think PECOTA is probably a significantly more complicated system than most of the other major forecasts. For one thing it is extremely sensitive to how the raw performance data on the referent population -- the population of potential comparables -- is evaluated and transformed (park effects, league and era adjustments, MLE's). I suspect that's where most of 2009's problems may have came from -- including the Wieters projection.
What I find least explicable is why the problems we are seeing now weren't addressed 6-8 months ago. If I were undertaking the translation of the Excel-based system into the new code then the first thing I would have done is to replicate past annual PECOTA projections using the new code -- while making no other changes in the methodology or he database. After late March 2009, Nate was no longer involved with PECOTA. They could have begun the code conversion any time after that. But it doesn't appear that they started this until after the 2009 season was over.
Only if I could use the new code first to reproduce Nate's results from 2008 or 2009 would I introduce any changes in the procedures. Did Clay first attempt such a direct replication? We don’t know.
At some point, Clay appears to have made some important changes -- he refers specifically to broadening the database of players from past seasons. Not only for the weighting of each players' baseline performance stats for the previous three seasons, but more importantly by adding a lot more player-seasons to the potential set of comparables. It seems likely that those added player-seasons -- ones that were not in Nate's database -- include mainly players with little or no ML experience. So these are players whose stats are highly subject to any assumptions and imputations made using the DT's and MLE's.
In addition they've had certain problems that seem specific to the Player Forecast Manager (PFM): taking the PECOTA estimates and plugging them in to the PFM. Now this is more complicated than it might at first appear to be -- chiefly because the PFM introduces depth charts and playing time estimates as constraints on the PECOTA estimates for each player. And lineups and batting order and the projected performance of all players on a team 'interact' to affect the projected performance of each every player as well as the team as a whole. In that way the PECOTA's aren't simply input to the PFM; they are simultaneously dependent on the projected performance of other members of the team.
Nonetheless, the PFM worked reasonably well in previous years (maybe not so much in 2009?). Why has it apparently become such a problem this year? Is it due to efforts to change the PFM this year? Or was this, too, transformed from an Excel spreadsheet macro to a new system of instructions for the first time this year?
And then there's the issue of how similar players are identified. It requires another complicated piece of analysis to search the database of player seasons who are most comparable to each current player – on the set of physical characteristics and performance indicators that PECOTA employs. This matching probably didn’t use an Excel procedure but instead some kind of analysis of covariance. Maybe that's where some of the problem lies this year. We haven't heard anything about this.
Finally there's the step of using the performance of the set of comparables to make projections for each current player's 2010 performance. More complicated code but probably fairly straightforward, if I understand the methodology.
In sum, there are a lot of places where things could have gone awry. As Tango wrote, we have to wait for BP to explain what's going on. As a long-time BP subscriber, I mainly want them to solve the problem. But my confidence in the PECOTA system has been shaken, and it will have to withstand two tests to rebuild it. First, I want to know why it took so long to get the new projections ready. Second, and more importantly, I want to know that PECOTA is once again among the very best at what it tries to do.
I see a lot of dumping going on, people grinding again on issues they've raised before. But very little analysis of the comparative value of EqA (whatever it's called) and alternatives. Did EqA predate or postdate WOBA? If they are comparable (as Colin Wyers pointed out on THT), why was the "second" one, whichever one that is, created?
Ragging on BP for publishing the Bradbury article is truly strange. By publishing it they didn't endorse it. And both BP writers and the vast majority of commenters on the article pretty much rubbished Bradbury's analysis. That's a mark in favor of BP, not against it.
As for the comments on PECOTA, the strongest and, in my view, most relevant question is why this year's problems are occurring in getting the individual player forecasts stabilized and the forecast manager (hence also the team forecasts) working consistently. While it does seem there were some problems with the 2009 projections, it's less relevant that BP explains what went wrong last year than that they learned that there were some problems and worked to solve them for this year.
I don't buy the argument that PECOTA -- the internal logic, the basic method of using comparables -- suddenly broke, after it had performed fairly well since 2003. But what I would like to know is how much of last year's problem was due to issues are are mainly external to PECOTA's logic, such as issues in the DT's and MLE's, i.e., in the data that is the input to PECOTA's forecasts. It appears from Clay's latest BP entries that this year there are definitely issues brought on by changes that Clay has made in defining the pool of players from which the comparables are drawn. And further there seem to be problems with how the Player Forecast Manager is cutting up playing time and filling out the rosters for each team. Both of these problem are issues that are largely external to the logic of PECOTA though they are critical to the player and team forecasts.
And this brings up the question that was raised by at least one commentator above: why wasn't this addressed much earlier, both by replicating and improving the 2009 forecasts and by modelling the 2010 forecasts -- for example, by anticipating that expanding the player pool, which had the effect of making the comparables contain a larger number of players with little or no ML experience, might throw a lot of things out of whack. Why wasn't this worked on last summer and fall? Why was it only faced in December, January, and February?
The streets and hiways of Canada will be empty from 7:30 til 10:00 -- well maybe a quick run to get some more beer around 8:45. (For the Russians, Moscow is 8 hours ahead of EST, so they're pretty much going to be vodka'd up the yingyang before the game starts at 3:30 AM.)
Russia didn't start out so hot either. Can't hardly criticize the Canadians for this quarterfinals showdown. Guess: Canadians wear out Russian defense and win.
USA, with 2 days rest, dominates Swiss.
The Czechs came within an eyelash of getting beaten by the Latvians yesterday (3-2 OT win). Not clear whether Jagr is fit to play (thanks to Ovechkin?). I think they lose to the Finns tonight.
Swedes should beat Slovaks.
Tango, another question to ask is if people could buy tickets to just one Summer Olympics competition, which one would it be?
Same for Winter Olympics.
For me, Summer would be track and field. (Next up: gymnastics.) Winter would be (men's) hockey. (Next up: downhill skiiing.)
Just one? Depends on the purpose.
For everyday interest (especially in season): baseball.
For betting interest: football.
For occasional intense interest, in a sport that I was once pretty good in and have a feel for: track and field!
It wasn't clear whether Clay was saying this 75th percent fix applies only to the PFM settings to use when using the latest versions of the forecasts, or whether instead he was saying that the PECOTA's in the book also have this problem. The latter would be a more fundamental error. Time will tell.
I don't think BP is saying the projections in the book are wrong (they are, however, always preliminary, i.e., not based on final depth charts and playing times).
Steve Goldman just posted that some of the problems seen in the weighted means spreadsheets and player forecast manager are not in the book.
Right now it looks like they're having problems generating updates to the PECOTA estimates as well as getting the PFM to work as it should. And Clay is working on this. I suspect these are more data management problems than problems with the basic PECOTA algorithms. But since Clay had to take Nate's humongous spreadsheet macros and turn them into a different code, there's certainly room for slippage. And in the past, Clay's job basically involved getting the input data (translated stats, etc.) ready so that the PECOTA's could be run. Now he has the entire process in his hands. So now he could conceivably be checking the whole computational process from data input and translation to estimates to reporting.
I agree that it's embarrassing to BP to have these problems. They should have beta tested and worked out the kinks before they released the projections online.
I like creaky old stadiums and small upright ones (stadiums where you get relatively close to the diamond even if you're in the grandstand), but there's nothing like Dodger Stadium on a sunny day.
1. Dodger Stadium
2. Wrigley Field
3. Fenway Park
4. Progressive (Jacobs Field)
5. Old Tiger Stadium
6. Toronto (in sunshine) (don't like it with dome closed)
7. Anaheim
8. Oakland
9. Shea
Old and dead one: Candlestick (sucked)
Long gone: Wrigley Field in Los Angeles (where L.A. Angles PCL team played before Dodgers arrived). Loved this one as a kid.
Should point out that this article isn't behind the BPro pay wall -- anybody can read it. I think that's a good precedent for any new metric that they develop and incorporate into their reports.
Just a comment on the title of this thread. It was allegedly Senator Daniel Patrick Moynihan who coined the phrase, "Everyone is entitled to their own opinion, but not to their own facts."
I proofread my manuscripts very carefully, using various methods -- not just reading word by word but also reading aloud, searching for common errors, etc.
But once something has been published I don't dare read it, because I just KNOW I'm going to see an error somewhere.
Depite the hype, Macs need AV software too. Once Snow Leopard disabled my Norton AV, I went looking and tried Avast! but now have installed ClamXAV, which found a couple of things that both Avast and NAV had overlooked. But I'm really not sure this is the last word. If anybody has other suggestions, I'd appreciate them.
In agreement with Peter Jensen, more from observation than playing: an outfielder uses all his senses, including hearing (cf. Adair's article), and getting a jump means paying attention to the pitcher's and hitter's tendencies, the base-out situation, the count, the elements (wind, sun location, the cut and dryness of the grass), the catcher's setup, etc. Further an outfielder chooses an initial location before the pitch is thrown and may also "shade" his stance or lean in a particular way.
He can't just be reactive to the flight of the ball.
Getting a good jump on the ball apparently depends on accoustics -- the "crack of the bat." Here's an article by Adair about this.
http://www.acoustics.org/press/141st/adair.html
You have to figure that a lot of subs are at much less than the $40 that Tango assumed. If they really have 20,000 subscribers, then many are re-subscribers who pay $35 per year and many are fantasy subscribers who pay $20 per year.
BPro's revenue stream has multiple sources.
1. Premium subscriptions.
2. Advertising on the website.
3. Book sales (mainly the BPro annual).
4. ESPN sponsorship (most of which probably goes to the authors of insider articles, and some to PBro as an organization).
I would guess that BPro's annual gross revenue is 1.5 to 2 times what Tango calculated from the subscriptions alone.
Wikipedia has had a "Sabermetrics" article since 2001. It gives a definition and the origin of the term. But according to Hahvahd Law School,
"4. How do you acquire rights in a trademark?
Assuming that a trademark qualifies for protection, rights to a trademark can be acquired in one of two ways: (1) by being the first to use the mark in commerce; or (2) by being the first to register the mark with the U.S. Patent and Trademark Office ("PTO"). 15 U.S.C. � 1127(a)."
http://cyber.law.harvard.edu/metaschool/fisher/domain/tm.htm
"In your eyes therefore, it would be IMPOSSIBLE for me to post that picture without my beating a dead horse."
You're probably right because you've been beating that horse for years.
I don't think the hype is smart either, BTW. But who takes it seriously?
Tango: don't pretend you didn't post the cover to make the same point you've made many times before.
As for the issue of using Nate's name, I wonder whether they cleared that with him. If I were him I wouldn't want my name on the forecasts themselves unless I did them. However, they could have worded that promo differently on the cover while still drawing on Nate's name recognition: "Featuring projections for 1600 players based on Nate Silver's deadly accurate PECOTA."
re #2: "At what age do economists start to decline?"
You mean economists who never make it to a major league university? Kennesaw State University is Carolina League level.
Tango, you're beating a dead horse. It's hype for a book cover. B.F.D. Get over it already.
You could make a far better criticism of this promotional cover if you wanted to: Nate Silver invented PECOTA but has had nothing to do with the 2010 projections, as Kevin Goldstein made clear the other day. This is basically in the hands of Clay Davenport now, and no doubt he will be tweaking the procedure some more, as he did last mid-season.
While I was rooting for Chicago, Rio should be a great venue for the games. Let the samba begin!
The point about whether a country's delegates can vote for their own country (or at all if their country is a contender) is not necessarily how many votes they have in a given "selection" but their bargaining power on procedural and planning issues, including decisions such as whether baseball is excluded from the Olympic sports.
Also, there is logrolling on votes. I recall being in China when their first bid was being prepared. They were sooo disappointed that they didn't get the millenial games. But they were a lock for getting one down the road in part because they went along on that decision calmly, and methodolically prepared for the future anyway.
Two comments.
The Cash for Clunkers program is a clunker of a design. Agree completely that the size of the incentive should have been fit to the MPG differential. I got rid of a 20 mpg 13 y.o. Taurus for pennies. Bought a new car getting 38 mpg. Did get hybrid tax credit, which was far smaller than the $4,500 for the clunkers. Would have made sense to offer $600 for a 6 mpg improvement, and $100 for each additional 1 mpg improvement up to some limit (say up to 40 mpg).
On MGL's OP: the core issue of one of "empathy," something that the right tends to denigrate. To me the concept has to do with the a sense of understanding how other people feel and considering not just their feelings but their actual well-being. Many people tend to think that whatever they have is theirs, even if earned largely through the labor of others or through inheritance or luck. They feel no sense of social responsibility. A practical aspect of this is that we've totally lost the sense of the purpose of progressive income taxes, or, for that matter, the dangers of inherited wealth.
Why not go with B-R's spelling (Jurrjens)? That's pretty much where people are going to be looking up info on him. That's also the spelling by MLB.com.
Jurrjens has been underestimated his "whole" (very short) career. The Tigers gave him away and had nothing much left on their pitching staff last year. He's not impressive to watch. He's not a big K guy. But he's incredibly heady for a young pitcher.
With your own laptop and a mifi in your pocket or a USB modem you would be able to log in any time anywhere. About $60 per month.
http://www.usatoday.com/tech/columnist/edwardbaig/2009-05-13-mifi-wireless-mobile-hotspot_N.htm
As I have commented on before, some systems come up with updated versions as the depth charts change during the preseason.
The DM projection for BP (PECOTA) is obviously based on an early version of the numbers and perhaps also not even using a BP-based depth chart but instead only the early individual player projections as published in the annual or the first weighted means spreadsheet.
@Dackle: I've heard of something called the "rank-size principle" in urban settlements; maybe that's what you have in mind.
On this whole issue I think the revenue has to be thought of as all-sources: subscription (if any), advertising (if any), publication sales.
If you broaden your concept of "baseball sites" beyond those that have a major focus on sabermetrics, clearly some others such as Baseball America are drawing revenue from all three streams. And I suspect that BP's revenues are coming to them partly for the expanded coverage they've given in recent years to prospects (read: Goldstein's columns), and partly from adding more "common" coverage (read: Perotto and others). IOW, the audience and potential subscriber base for "sabermetric" analysis is probably smaller than the subscriber base of BP.
Perhaps most importantly, you have to figure that the audience of fantasy baseball information is much much greater than the audience of sabermetric analysis. And the one product that BP has that feeds that audience is PECOTA. Without PECOTA I bet BP would lose half of its subscribers. I don't now how much money the other fantasy sites make, or the fantasy services offered by some of the big news sites.
But right now you have most of the best sabermetically based "forecasts" being given away on their non-subscription websites! PECOTA and Shandler are prominent exceptions, and their sites and books appear to make enough money to sustain a fairly large stable of writers and analysts. (BP also gets what appears to be decent advertising revenue as well as revenue from subs and book sales.)
At the very least, I think you all need to give some thought to what products you have that can appeal to a bigger paying audience -- and this has to be fantasy players, and the product has to be preseason forecasts and tout sheets, coupled with in-season updates, player injuries reports, prospects, trades, etc. Fantasy players feed on that stuff. And then you have to question why you're giving it away -- other than your devotion to baseball stats -- and whether it's this part of your product that you most need to try to sell for cash.
I agree the the idea that at base the grading has to be observational (but some combination of this and "objective" metrics is also reasonable).
But what I would propose as a way to reduce the costs would be that a sample of trained observers. 100 seems reasonable but that gives a +/- of about 10% for any given observation/play; however, but if they observe all plays in a game the event log grows a lot so your number of observers can probably be much smaller. Also I'd have those observers grade the fielder's actions even when they aren't "in" on a play (nowhere near the ball, for example, but are expected to move to cover a base, e.g., the pitcher covering the plate when the catcher has to go after a nubber or a PB/WP).
The "grading" can be done on the basis of video, which could be rerun, and the video could even be enhanced by putting electronic markings on the ground to indicate location on a grid.
The "observers," then, could grade one player for a game, or one position for an entire game -- that player's movements would be captured in video, with a "pic in pic" perhaps for larger simultaneous views of what's happening on the given pitch or play (not just what that player is doing). (An occasional cutaway to a particularly luscious fan in the stands could be a welcome tension breaker! Sorry guys, no beaver shots!)
Grade each player for ca. 16 random games per year, i.e., about 10% of the regular season games -- but grade them on every pitch/play for those games. The observer panel can do this at their leisure and submit their grades online.
It would be a heckuva interesting experiment. You'd definitely need to develop rubrics for grading and to train the observers. I can't calculate the statistical "power" you'd need -- you might not need 100 observers of each play(er) to get what you want if the different plays over the entire game are pooled together as observations. Maybe 10 trained observers would be needed for each position. A competent sampling design expert would be able to figure this out.
To add to a possibly defunct thread, one thing that occurred to me is a term from football teams. After a game, the coaches "grade" each player. I don't think they do this by reviewing a lot (or even any) pbp film. But it's SOP in football.
Imagine in baseball if a team could do this for each defensive player. Maybe not every game. But imagine that they literally watched a player play an entire game -- where he positions himself, what he does on every pitch and batted ball. Do the infielders move to cover the base, do they move to the right cutoff location, how quickly do they get off the throw, do they throw to the right base, how accurate is the throw, how strong?
Imagine that by truly tracking a player's every move on every play you could learn whether he's doing the things that are going to put him in a position to make a play, to back up the play, etc.
Would a baseball player accept the idea of coaches "grading" him based on this -- giving him a score for a game even on the kinds of things that Tom has asked us fans to do (but only) for a player's entire season?
The player wouldn't be terribly receptive, I imagine if he hadn't made any errors, hadn't been out of position on any play he was expected to make, and so on. But the smartest team players are going to be in position to make a play on every pitch, and many of the things he does do not show up in any of the standard stats. And a smart team might evaluate and train thair players by doing such careful grading.
Why couldn't this be part of the assessment of how well players play defense?
Just a comment on why people join professional associations. Lots of academics belong to more than one association. Most of those associations offer (a) a journal, (b) an annual convention. Analogous to SABR.
For some years now, if you're affiliated with a university that has a good library, you've been able to get instant online access to thousands of professional journals without being a member of the sponsoring association.
So why join the association?
Answer: for the meetings. You often still have to pay a registration fee (maybe less than nonmembers, or maybe nonmembers can't attend at all). When some professional associations at first hesitated to make their (recent) journal issues available online -- thinking that this would discourage people from paying dues -- they learned that in fact this isn't the case.
Perhaps the opposite was true: if you can "sell your journal" (by getting libraries to buy access to online versions of it, perhaps through some kind of publishing consortium), then you may sell the value of your association -- people will come to present papers, get feedback, meet other experts and up-and-comers, have fun.
Following this logic, only if you're interested in the meet-ups are you likely to join SABR.
I used to be a real fan of OJ the football player. After that -- as actor, promoter of Hertz -- not so much. I loved one thing he once said about why he always seemed just a bit slow -- taking an extra couple of seconds to get up from the ground -- after a play. "I don't want them to know that they hurt me. I get up slow whether they hurt me or not." That was a signature of his.
I think he was guilty as hell of that double murder. A jury of his peers found him not guilty. And the state can't appeal a not guilty verdict. That's the system. It's not perfect. It's supposed to favor the accused against an overreaching government.
I agree with you that the crime for which he has just been sentenced was over-charged or shouldn't have been charged at all. The sentence is way out of line with any harm he did in this instance. The only explanation is that this is the law enforcement "system" getting even for its earlier failure to convict OJ, plus the fact that OJ has no friends in high places.
Scooter Libby will probably be pardoned by Bush, for committing several felonies including perjury and obstruction of justice. And OJ will rot in jail for at least 9 years for stealing back his own stolen property from some of the most unsavory characters around.
Every version of a simple or so-called flat tax that I've seen turns out to be deceptive for one basic reason: either it's a tax on consumption or it's a tax on earnings. "Unearned" income (read: income from investments, savings, inheritance, real estate gains, etc.) is totally excluded.
Thus such a tax is ultimately regressive by excluding from taxation a very substantial part of the income of the rich. You have to read the fine print to ask what kind of income is in the base that is subject to tax. Does the print say "earned income?" It's not truly a flat tax.
Tango,
Looking that Nate's list, I'm reminded about the Elo application that he made to professional baseball (and that Clay Davenport also incorporated into team season projecttions). Looking up Elo on Wikipedia I see that such ratings exist for some other major sports including basketball (Sagarin uses them in part).
But I don't see one for hockey. I'd love to see one for both NHL and D1 college hockey (which maybe USCHO could add to their power rankings?). Any chance you might take this on -- just for fun and for us hockey fans?
Dull? The last two games were electric, IMO. I only had NBC to watch. We used to get CBC on our basic cable but no longer, so didn't have the option to compare the two coverages.
I'm going to miss CBC when the Beijing Olympics come on because they always do a better job, and broadcast for more hours, as well as more events in "real time," than do the American networks. No insult intended here but I think one reason CBC has better coverage is that they pay more attention to the preliminary rounds and the heats while the Canadians are still competing, whereas the U.S. networks seem only to be interested in the finals and often don't even stay tuned to the entire event (say if it's a long track race).
Thanks, guys. I guess I was just thinking that WE or other "expectancies" are based on what are the general or overall or "large number" average probabilities. I'm thinking that while certainly it's an important baseline for evaluating managerial decisions, a more explicit modeling of some of the complexity might be worthwhile.
For example, if you were to systematically study managerial decisions in various areas (let's say you were trying to come up with a standardized metric for evaluating managerial competence), then you might test for the effects of various factors in accounting for the "mistakes" (cases where their decisions appeared to be counter to what was seemingly rational).
Then you could perhaps understand that that there is a method (perhaps an accepted "rule of thumb") or perhaps a "psychology" to such decisions, just like that study of football decision-making in which coaches were found to take the field goal rather than go for a first down far more often than was "rational."
In training managers to make "correct" decisions, then, you wouldn't just drill a WE matrix into their heads, but you would also address which other factors might be worth taking into account in certain game situations.
Thanks, Ryan. I guess what I'm thinking is that WE is a very static kind of factor to consider; that is, it presents an "average" probability (of winning) that, while situationally accurate doesn't allow for the multiple factors that are involved in decisions at any point in the game.
Just those considerations that you mentioned -- remaining outs available in the game, available pitching resources and pinch hitters -- could be explicitly modeled, instead of being subsumed by the "average" WE associated with the given situation.
To take this a step further, there are also "next game" considerations or "series" considerations. Do I burn a relief pitcher today who I may need tomorrow (or who I used yesterday)?
Thanks. Has anyone tabulated "runs lost" due to allowing pitchers to bat with RISP? Following the original post, this could be used as a systematic measure of the quality of managerial decision-making. (Along with other quantifiable questionable managerial decisions, e.g., bunting, steal attempts.)
It's probably a lot easier just to look at NL data and calculate PA's by lineup order and then, since pitchers are almost always in #9 slot (except if there's been a double switch during the game) figuring out how many PA's are saved for "better hitters" by comparing total PA's by pitchers with total PA's for #9. But getting such a calculation for cases with RISP would be pretty hard, I suppose. And wouldn't you have to look at the RS in the inning after the PH occurred, and compare it with the expected RS without the PH to net out the gain from removing the pitcher? (Not a data maven here.)
This isn't a comment on the particular decision by Cox to let Reyes hit.
But how much, if it all, should one weigh in both the run/win expectancy of pinch hitting for the pitcher in a given situation at a given point in the ball game and the assessment of how well the pitcher is performing on the mound, and the availability of better substitutes at that time? If the manager thinks the pitcher will hold the opponents for a couple more innings, and the chances are decent that his own hitters will score another run or two, it's probably still wrong to let the pitcher hit but the calculus is more complicated than has been allowed for.
Another way of putting this is if the situation were similar in terms of runners on base and outs, should the manager make the same decision whether the batter is a pitcher or a position player? Does the expectation of runs gained by pinch hitting always trump in that situation? And wouldn't that be just as true in the 2nd inning as in the 6th? (Should a .117 BA pitcher ever be allowed to hit with RISP? When and when not?)
Is "Leverage Index" a copyrighted name? If it's in the public domain, why do they need to offer to pay you for it? Answer (I think): you created it, and they want to keep other game manufacturers or any other commercial venture from using the term. Presumably that would also keep MLB from adopting the term as an official statistic as well, unless they paid a royalty to whoever owned the name.
(I assume they're not going to call it "Tango's Leverage Index" and therefor want to use your name as well.)
Re #14: The results in IN and NC were very much as expected if you had the right model, which (a) almost nobody did, and (b) didn't depend too much on the polls. To use a perhaps faulty analogy, those who relied on polls and the image of a horse race with late deciders, etc., are like baseball analysts who rely on short-term random factors to determine the underlying quality and skill of the players.
There was one forecasster who relied more on fundamentals (demographic analysis) who has nailed the last three major primaries within a few points.
In yesterday's primaries, he had projected +2 for Clinton in Indiana and +17 for Obama in North Carolina.
Here's the site: http://www.fivethirtyeight.com
"538" is also known as "poblano" on DailyKos.com. He lays out his methodology in great detail on his 538 site.
Re #7 and other comments on PECOTA projections. The full projections by team posted above were the preliminary ones before defense was considered AND before depth charts were set. There were many changes in the predicted wins in the revised projections based on establishing depth charts.
If last year is a precedent, those team projections will be revised a couple more times before the season starts but further revisions will be quite minor. If anybody wants to see the "latest" (and they have asubscription) then they should go to the "Depth Charts" here: http://baseballprospectus.com/fantasy/dc/
MGL, I have no inside knowledge of PECOTA methods. I've tried to figure out how it works, and I agree that it's an amalgam of "Marcel" with its reliance on last 3 years for a baseline (plus aging and regression adjustments), and "similarity scores" with their reliance on comparable players for the forecast. The biggest mystery about it, it seems to me, is the method of matching individual players to comparables in the database. The most thorough discussion of this that I'm aware of is in the BP 2003 annual, but there's not enough information there for anyone to mimic it or reverse engineer it, it seems to me.
Tom, a question for you. Is the regular season vs. playoff scoring environment as different in most ML sports as it is in baseball (as percent of base)? Maybe not in soccer. What's the situation in hockey, American football, or the NBA and NCAA basketball?
In any case, don't you have to correct for the quality of the teams or competition in the playoffs (i.e., selectivity or survival, especially the further along you go in the playoffs, say from first to second to third rounds), and not only look at aggregate differences between regular season and playoffs? Maybe the playoffs in baseball are more selective for quality because 8 of 30 teams qualify, while in other sports a larger proportion of teams qualify?
On some of those nearly "incomparable" players, especially when they outlast everybody (maybe because they were juiced?), Pecota relaxes its rules on age/position matching, but it still seeks "comparables" using this relaxed set of criteria. And Pecota relies on this "relaxed" set of comparables for its forecasts. I don't think it completely defaults to a Marcel. But I suppose you could determine this empirically (de facto) by direct comparison of Marcels and Pecotas for such rare players. Not sure you'd have enough cases to do such a test.
As you know, Pecota isn't a Marcel-type system. It requires matching the baseline performance of a given player to a set of historical comparables. It uses some non-performance characteristics (body type, handedness, position) as well as performance stats to find the comparables, and it often uses peripheral stats in the process. It's the future performance of the comparables that represents the "projection" of the given player's performance.
I think it's possible that the match to comparables doesn't work as well for some types of players as it does for others -- e.g., for someone with an unusual combination of performance and nonperformance characteristics. In effect, Pecota may be "fooled" in such a case because the case doesn't fit a canonical pattern of relationships between the variables in the model. Pecota doesn't default to a Marcel if the similarity scores used to identify the comparables are really low (maybe it should -- thus becoming a hybrid model).
Marcel wouldn't be fooled so easily by any particular "type" of case because he's just a dumb monkey who always says that what a player's going to do this year is pretty much what he did the last few years. No peripheral stats, no speed scores, height, weight, or position to find "similar" players.
I can imagine Silver inspecting and probing certain individual cases, especially those of outstandingly talented players, as a way to evaluate his system, but once he's set up the formulas he doesn't "overrule" the system's projection for particular cases. I see him "disagreeing" with Pecotas for particular players in some of his articles, but he lets the system "speak for itself" in his statistical projections.
If Ichiro is as unique as Silver says he is, then no system is going to capture him well. Obviously, Ichiro has been a puzzle that he's considered and can't solve in any general way.
http://www.baseballprospectus.com/article.php?articleid=3497
IIRC Nate Silver announced plans to introduce a comments feature last year, or maybe even two years ago, but they never implemented it. Given what's been mentioned here (risk of spam, cost of monitoring), I bet this feature will be for subscribers only.
Great detective story, Dr. Tango.
Does anybody have a cache of "historical" MLB bats available for testing?
On the umpire theory, can you use Retrosheet to test for same pitcher-batter-umpire combinations? Or just for umpire effects (HR/contact) within and across seasons?
The purpose of the hearing was (IMO) two-fold: (1) To showcase some Congressmen in action (what else have they done recently that was memorable and casts a favorable light on the the House?); (2) to sucker somebody into perjury so that they could refer the case to the DOJ, since there was no way to prove or disprove the factual assertions from any available evidence or testimony.
It's an American style "show trial," meant more for show than for trial since this isn't a court (even a court of public opinion), and there are no charges on which anybody is being tried.
Re Phil Birnbaum's comments:
"—Get Dan Fox’s baserunning metrics up as a stat report particularly when there’s room for PAP, umpires, and RBI opps."
I think this is in process now. See here: http://www.baseballprospectus.com/unfiltered/?p=683
"[H]ow does PECOTA work? I mean, in detail, so I can code it myself. If that already exists, I haven’t seen it. If they want to keep it proprietary, that’s OK too, but then it’s useless to me as a researcher. . . ."
It's proprietary, as are most projection systems. All praise to Marcel, however.
I agree that even if they keep doors closed on most of their recent stuff they should open their archive, perhaps using something akin to what many journals do when the have their back issues on JSTOR. They use some kind of "moving wall," so that the journal content is available on JSTOR if it is, say, more than 1 year since printing, or more than 3 years (journals differ).
In the case of BP, it would be reasonable to take the restriction off of any article that they published more than 12 months age. Thus, any article published before December 2006 should be "open" today. And next month they just move that barrier ahead another month.
That would make a lot of their materials not only accessible but citable -- and would benefit BP's reputation on both accounts.
Tango, were you intending to brand the individual who uses "Lies, Damned Lies" as the header on his regular columns as a rank and arrogant number cruncher?
Whatever you intended, my view is that many blogsters and other regular columnists in the MSM who write about baseball -- whether from a sabermetric or nonsabermetric perspective -- are a distinctly opinionated bunch. But that's just what many readers want -- opinions, evaluations, not just descriptive analysis. Not "two-handed economists" ("on the one hand, on the other hand") but people able and willing to make their best expert judgments.
Dan Fox calls to our attention (and reprints) a BP analysis of his own from 2006 on platoon splits.
http://danagonistes.blogspot.com/2007/11/of-crowds-and-splits.html
Tango, you should also run a regression and see whether the intercept (constant) is significantly different from 0, and the slope significantly different from 1.0. You could have R=1.0 and still have different slopes, intercepts, and mean values.
Thanks, Tango. So (1) hitters are more patient," which in effect means that if pitchers throw 108 pitches they will throw one fewer inning on average today; (2) managers of yesteryear reached early decisions to pull pitchers who started poorly, thereby assuring that they had more starts than they might have if they were allowed to finish 5 or 6 innings (to what extent can this be shown in the data? are "modern managers" maybe realizing that back luck can lead to a poor first couple of innings?); (3) if a pitcher gets to 8 innings, the manager is likely to give him a chance to complete the game (contingent on size of lead?). Perhaps another factor is the number of pitchers in the basic rotation, and hence the number of starts the top pitchers can expect to have.
Seems plausible that these are explanatory factors, but perhaps begging the question about the changing use of the quick hook over time, and how much it may be contingent on the pitcher's performance and the score of the game (or the overall run-scoring environment in different eras).
Also, the drop in total CG's still seems rather more steep than might be accounted for in the terms explained above. Perhaps need to do both intertemporal and comparative cross-sectional analysis (within a season in 1960's and within on in the 1990's or 2000's).
Anyway, this is all very interesting. Thanks a lot.
I'm
Thanks, Tango for the link to your pitch count summary. (I wrote my previous message without seeing yours.) The figures for Koufax comport with my memory.
But it's also likely true that the average starter (or the average highly successful one?) was more likely to complete the game with that ca. 110 pitches, whereas the later-day pitchers do not.
So wouldn't a CG/Pitches ratio for starters would be far lower today than in Koufax's day? And if so, is it just a change of style of the game, or perhaps, as you are suggesting, the avoidance of extremely high pitch counts for ineffective pitchers? (Again, trying to avoid a circular argument here.)
Added note. It would be circular reasoning to argue that the reason there are fewer CG's in recent decades is the rise of relief specialists.
(It would also be ahistorical to attribute the decline in CG's to "PAP," since the latter is a more recent invention than the decline in CG's. Also, I think we started "saving the arms" of young pitchers, i.e., in Little League, at least a couple of decades ago not just by outlawing the curve ball; only more recently have strict rules been implemented about pitch counts but I'm not sure when that happened.)
So how do we come up with a non-circular and non-tautological explanation for the decline of CG's?
Leaving aside PAP, why is it that there are so few complete games these days compared to, say, three or four decades ago (even among those "top 25 in PAP")? Because managers believe in PAP (or are just counting pitches) and are making the stupid decision to 'save the pitcher' from risk of injury? Because pitchers decline in effectiveness at an accelerated rate after a certain number of pitches? Because batters are behaving differently? Or there some other reason or combination of reasons?
In short, what happened to the complete game, and why?
I do recall Dodger broadcasts back in the early 60's in which pitch counts were routinely reported, at least at the end of the game. I don't know how common that was, but I think Alan Roth must have been compiling such stats at that time and feeding them to Vin Scully and Jerry Doggett.
I hadn't heard of Scout School, but with a little Yahooing I found this link on MLB.com:
http://www.mlb.com/news/article.jsp?ymd=20030211&content_id=199087&vkey=news_mlb&fext=.jsp&c_id=null
How can anyone possibly support a guy like that for President? He uttered the m-f word on the air last night. It was quite wonderful.
I'm sick of the whole lot of current official aspirants on both sides. (Well, Ron Paul is definitely a breath of fresh air.) Bring on the interesting candidates like Colbert.
Where's Dick Gregory when we need him? I still have a "dollar" with Gregory's face on it -- recall that they used to work in dollar changer machines and so were banned. Even so they were more valuable than the Canadian dollar at the time. (Sorry about that one, Tom; though I'm pleased to note that the Canadian dollar's on a roll now, at least relative to U.S., so let those Canadians come south, or east or west to the States to shop til they drop).
How about Pat Paulsen? Another odd candidate of the past. Even Ross Perot would be a blessing in this marathon electoral race.
Should have added that in addition to having 7 inning games, they settle ties not by playing extra innings but with a shootout -- home run derby. Shades of international football and hockey.
There's a nice entry on the Israel Baseball League on Wikipedia, with a map and other interesting information. I believe they play 7-inning games, so this by itself raises potentially interesting questions about the effect of shorter games on the competitive environment, the possibility of using smaller pitching staffs (and shorter intervals between games, less concern about pitch counts), and so on.
http://en.wikipedia.org/wiki/Israel_Baseball_League
Without checking the calendar from way back, it looks like these results are consistent with my recollection. He did not pitch Game 1 in two of these three series.
In addition to Koufax and Drysdale the Dodgers also had, if I recall correctly, Johnny Podres and Stan Williams who started a large number of games (also Roger Craig for a while), and they had Clem Labine in relief. Koufax, in any case, would typically miss or delay his October starts because of Yom Kippur, so that would somewhat limit the number of games he could pitch in the WS.
[quote]Has anybody ever mentioned the possibility that Gagne just isn’t any good anymore?[/quote]
Lots of people, including Nate Silver in the article that was the stimulus of this thread:
[quote]In the two most reliable indicators of pitching performance — strikeout and walk rate — Gagne has been exactly the same pitcher in Boston that he was in Texas. And that is a pitcher with decent but unremarkable talent. The 2002-2004 Eric Gagne was incredibly good. The 2007 version is — I don’t know — a rich man’s Ryan Dempster.
Now, it’s not like Ryan Dempster is great shakes. But Gagne is not nearly as bad as his ERA suggests.[quote]
BP guys (just like guys out of other shops) have come and gone and come again in MLB organizations. Former BP'er Keith Law was with Toronto, and now is with ESPN.com. Two other former BP'ers (James Click and a staff guy) are with Tampa Bay now. Keith Woolner is with Cleveland, as you mentioned. There may have been others. The above information comes from Wikipedia and therefore may or may not be accurate.
There don't seem to be a lot of careers for real statistical analysts and theorem testers (as opposed to data managers and mere stats mongers) within those "small businesses," as Tango describes them. The larger baseball analysis business probably offers a lot more potential for careers in the media, fantasy baseball analysis, scouting and prospect reports, and so forth.
Most of the well known sabermetricians appear to be amateurs in the sense that although they may earn some money from their sabermetric work (writing books and articles for online or mainstream media), they earn their living mainly by holding down "real" jobs. And most do their sabermetric work simply for the love of the game, statistics, analysis, or who knows what -- but not for profit.
I read that column by Huckabay in the context of the last serious article I read by him, and that was in the 2006 Baseball Prospectus annual book, which I just pulled off the shelf: "Where Does Statistical Analysis Fall Down? Reality and Perception."
I don't know what it was that drove Huckabay away from BP.com for 3+ years, but I think he's been brooding over one issue in particular: that MLB organizations do not really want to pay for the quality of analysis that they could really use, and that to the extent that they recognize there is valuable research out there, such information is largely in the public domain. In fact, the findings ("The Book," for example) can be bought for pocket change. Because MLB's only open to a couple of big ideas at a time, and those ideas are a public good, it isn't going to hire the Tangotigers of the world on a full-time basis.
Now these conclusions are debatable. But I think from re-reading that BP 2006 article, which includes an interview with a cynical MLB executive, that Huckabay has been in a funk about the limited ability of analysis to improve the game.
Beyond this, however, anything else he may be saying about the value or validity of baseball analysis just doesn't make a lot of sense to me. (I also think, from reading some of his comments in his "chat" on BP yesterday, that he realizes how inarticulate he was in that article.) Nor could anyone on the BP staff other than Huckabay endorse the statement that analysis is dead.
I don't think a lot of batters make the decision whether to swing or take just on what they observe after the pitch is released. Obviously they need to figure out the velocity and location, but they also arrive in the batter's box with some prior expectations that may affect how they react to what they see. And they have an at-bat strategy that affects how and how quickly they make a decision whether and how to swing at the next pitch.
Consider the following as just some of the auxiliary considerations that come into play.
1) The count;
2) The man on base situation and who is on base;
3) Is there a "play" on -- e.g., hit-and-run, bunt sign;
4) What kind of pitches does a pitcher have in his repertoire (or what kinds does he have command of today), and what kind does he like to throw in certain situations (information that may help the batter know what to look for);
5) What kind of pitch and in what location the pitcher threw to him in his last plate appearance, or last time in a similar situation;
6) Whether the pitch is hittable (even if it's not a strike; but even if it looks like a strike, if it's not already 2 strikes against him, the hitter may pass on the pitch if he can't get good wood on it);
7) Whether to try to hit the ball in fair territory or just foul it off.
All of these things and more are in the batter's head before the pitcher winds up, and all may affect the decision to whether (and how) to swing as well as how long it takes him to reach that decision (for good or for ill);
Seems to me someone has to work through the batter's logic, not just his ability to pick up the pitch and get a bat on it.
I take back the term "Sabermetric School." I'd call it a "Baseball Strategy Academy," and focus on strategic decision making on the field and in games, in personnel management, and in finances. And draw "instructors" from the baseball business as well as from the baseball analysis business.
I don't see how that would work at all on a practical level, in part because the strategy decisions often involve assessments of the capabilities of the personnel on hand at the moment (which the field manager may have a better handle on than anyone in the front office).
A more general approach would be to organize a "sabermetrics school," in which you would use your sim, the product of research in "the book" and other sabermetric research (e.g., on player valuation and business decisions). A three-day intensive seminar for field managers and front office personnel. Paid for by MLBAM, perhaps. In other words, train them in what is known from a statistical standpoint, have discussions about strategic decisions and how different managers might handle them (use real, known decisions, lots of video, and build the stats and odds into the discussion, but don't just teach stats and odds).
My working assumption is that most people don't read -- not even these wonderful blogs -- but they might enjoy "going to school" for a few days in the off-season. You've gotta involve more than just statheads in the training, however. . . . Banish the powerpoints, or at least minimize them. Lots of videos and "scenarios" that lead to interpretation and learning.
anon, the sampling error with a sample N of 1000, and the proportions of responses roughly divided 50-50 is about 3.1%, not 1.6%.
Here's a sampling error calculator:
http://www.dssresearch.com/toolkit/secalc/error.asp
You are absolutely correct about the difference between sampling error and bias.
At the same time, you might note that a lot of the observed variance is due not just to sampling error but also to other factors (poorly worded or confusing questions, respondent disinterest, coding/recording errors by the interviewers, etc.). Sampling error is only one source of the error variance.
MGL: Thanks for reminding me that trades were for contracts and not players, and thus the valuation of such trades has to take into account more than just the future performance of the players on each side of the trade. (There are, of course, other arcane rules that may lead to such trades, e.g., to the Tigers' trade of Ledezma to Atlanta earlier this year.)
Pizza Cutter: What's the average number of pick-off throws that pitchers make? Other than the distraction factor, how much might a baserunner's ability to attract pick-off attempts or "throw-overs" wear down a pitcher?
"That does not bode well for his reputation in my book."
You should write a book about it.
Seriously, has anyone done a thorough analysis of the net gains-losses from trades? How many of these seemingly valuable properties are lemons (or go sour)? How often is there an essentially good trade, i.e., an even trade for value -- even if one team is seeking to exploit immediate value and another is seeking to benefit from that value in the future?
My hunch is that umps adjust for the type of pitch(er). One way that we all recognize (or think we recognize) is that certain pitchers are "given," the outside of the plate ("the black" and beyond?) and others aren't.
Another way is that an experienced ump knows if a certain pitcher likes to throw backdoor strikes. Even the catcher may cue the ump in, not just by framing the pitch but by a comment.
Not to take you too far off the main course, but it seems to me that some pitcher-specific analysis of sz's for pitchers with "known" tendencies, could be useful. Do they get "significantly" more called strikes in a given location than average?
One more question, related to front vs. back of the plate (or front of plate to catcher's mitt), do the data reflect whether the ball ever passed through the SZ or do they capture where it is at the leading edge of the plate?
What happens to "back door" sliders, etc.?
That may be one reason that Tango wants to see breaking ball splits.
Pretty interesting stuff. One thing to pursue for sure, given the asymmetrical relation between the handedness of the batters and the width of the zone, is the handedness of the pitchers, or more particularly the relationship between the shape of the zone and the combination (or interaction) between hitter and pitcher handedness.
Thaks. My bad.
Statistical Significance, or the reason that mathematician Ron Fisher is on MGL's "On Notice" Board (Fargo) —
Interesting article in Science News: "Odds Are It's Wrong: Science Fails to Face Its Statistical Shortcomings": http://www.sciencenews.org/view/feature/id/57091/title/Odds_are,_its_wrong
Analytics Conference (Fargo) —
Well I imagine that we'll only know for sure if he publishes his comment. I'd surmise that he was calculating this from a game theoretic perspective (he is a card player, after all). My own experience is more in statistical testing, in the sense that I described. But the point stands, I think, that taking the "safe" option will not get a coach in trouble. Deviating from it -- taking the risk of doing so -- requires that he have something more than just intuition to go on, but instead some other factor(s) that research or experience shows can shift the odds in a predictable way.
Analytics Conference (Fargo) —
Mike: take a look at the detailed summary of Nate's argument on the blog. If the "accepted wisdom" says you should choose A over B because A will work 2/3 of the time, are there any conditions when you should choose B? The accepted wisdom is safe, defensible, and won't get a coach in trouble if it fails 1/3 of the time. Nate calculates that choosing B will require information that increases its odds of success from the base 1/3 probability. So the question is how much better odds than the base 1/3 success rate do you need to justify choosing B? And more importantly (but not stated in the summary that is linked), what kind of information is that? To me this boils down to a standard issue in multivariate analysis, where you have to decide whether adding another variable to a regression model will improve the prediction enough (perhaps allowing for interaction among predictor variables) to justify the additional complexity -- and let the coach off the hook for defying the conventional wisdom. It can't just be a coach's intuition, but most likely an assessment of contextual information of some kind (e.g., injured key player on the other team; offense/defense mismatch of some kind) that is not just randomly related to the outcome. WHEN do you "go for it" on 4th down rather than punt? When do you pull your goalie? (me: when you know there's a 95% probability you will lose and only such a desperate move gives you a significantly better chance of winning. I don't know the real odds here, but Tango probably does.) So while I think Nate's logic is good, it doesn't answer the question when you should go against the conventional wisdom. Of course, that's what research is all about -- finding holes in the conventional wisdom by either improving measurement (of performance indicators) or looking for other predictive factors over outcomes that are also practical for managers/coaches to follow.
Era adjustments (Fargo) —
So those authors knew of other estimates but didn't even cite them. For example, in his "Ruth vs. Bonds" intro chapter in Baseball Between the Numbers, Nate Silver estimated that if Bonds had begun his career in 1916 (when Ruth did) he'd have had 444 career home runs. And if Ruth had begun his career in 1984 (when Bonds did), he'd have hit 913 homeruns in his career. This is a vastly different set of numbers from what the authors of this article came up with -- yet they cited the book without citing Silver's estimates. Not saying Nate's right and they're wrong, but the authors' had an obligation to read the lit.
Era adjustments (Fargo) —
Agree that the bibliography in the paper is weak, and misses some preceding studies specifically asking "What do statistics tell us about steroids"? Nate Silver has an article with just this title in Baseball Between the Numbers, a book that the authors of this study cite but strangely attribute to James Click. He also wrote a chapter comparing Bonds to Ruth in that same book.
Attendance and winning (Fargo) —
Doing it with # of fans is the most reasonable, it seems to me. When you use population size (in the metro area) as a predictor, you can also make an adjustment for whether there are competing ML teams in the region (LA, NY, Chicago, specifically). Don't split the population, just put in a dummy for "two teams".
Attendance and winning (Fargo) —
"Making the playoffs the year before raises a .500 team's attendance by about 3,000 fans per game - a major boost. Obviously making the playoffs raises hype around the team, and this appears to manifest itself in the form of increased attendance." Making the playoffs the previous year, winning the league, winning the world series all matter (in a study I did some years ago). But it's probably not just a matter of "raising hype." It's a matter of raising season ticket sales. Those sales are probably driven much more by previous season's record (specifically making playoffs, etc.) than by current. So a refined analysis -- if data is available -- would look for lagged effects of wins (making playoffs, etc.) on season ticket sales and for in-season effects of current wins on walk-up or game-day sales.
Era adjustments (Fargo) —
An interesting but somewhat curious study. I can't follow the math but I can follow the logic. Despite the title, it says nothing about steroids per se, as opposed to era in general. (I'll leave aside the fact that they chose RBI as one indicator.) I find it curious that they decided to scale the HR so that "The Babe" ended up with over 1200 HR. Maybe they could have anchored the baseline curve to the HR levels in the 20's and 30's and shrunk the more recent levels more than they did?
Clay full-time at BPro (Fargo) —
For the type of work that Clay has done with NOAA, there is no particular storm season for his job -- since he works on the entire globe not just the northern and western hemispheres.
Consistency is better than inconsistency? (Fargo) —
Still, this is looking better in the sense that higher variation (less consistency) in either scoring or being scored upon is associated with a lower win probability. Now you can try to add back one of the "enviromental variables," let's say RS only. That would capture the overall "level" of scoring. Or you could put in total runs in game or total runs per inning to capture the same effect.
Consistency is better than inconsistency? (Fargo) —
I wonder if either of the coefficients for the CV's would become significant if you omited the first two terms.
Consistency is better than inconsistency? (Fargo) —
I'm not even sure that you would end up needing the RG and RAG variables in a final equation of the type I've suggested, but they would in principle be interpretable.
Consistency is better than inconsistency? (Fargo) —
The coefficient on runs per game may fall because it's correlated with the standard deviation of runs per game. Another way to represent the variance is to use a coefficient of variation (CV), which is the standard deviation divided by the mean. In effect this "norms" the standard deviation -- its size is relative to the run environment. I think you could try substituting CVRG and CVRAG for the two std. deviation variables that you now have. And then I wouldn't be surprised if the coefficients for RG and RAG became insignificant in such a setup, but that would be perfectly fine, because the CVRG and CVRAG would capture the run-environment-normed effects of variation/consistency in runs scored and allowed.
Tangotiger = ESPN Blogger (Fargo) —
Selling out to the establishment! Going behind a pay-wall! Horrors! Seriously, this is a good move for you. Congratulations.
Consistency is better than inconsistency? (Fargo) —
@salb#17: I get your point about the interpretation of the Pythag calculation. But as you know many prognosticators rely on the Pythag projected RS and RA to predict the number of team wins. Further, they take the difference between the Pythag projected wins and the actual wins at any point in the season (including the end) as indicating how lucky or unlucky a team has been. However, this assumes that the difference between actual and Pythag wins is only due to luck. Suppose the ability to exceed the Pythag win projection were at least partly a "skill" or were predictable in some other systematic way. Perhaps a team's style (small ball vs. long ball) or its (unmeasured) baserunning skill (e.g., advancing on a ground out) were PERSISTENT. Suppose a particular team or manager showed an ability over several seasons to win more games than the simple Pythag projection. In all of those cases, the difference between the Pythag and actual number of wins is not just due to chance. Rather, it's a "skill" (perhaps managerial skill or team style). And if it is persistent and means that, say, an otherwise Pythag 85 win team can be expected to win 89 games, then that's very meaningful for making projections of team wins. So the question becomes, is the ability to exceed (or fall short of) a team's Pythag wins persistent? If so, then just like "consistency" (IF it is a consistent trait or habit of a team or manager), this ability (or inability) is not just a chance outcome. (Additional question: I wonder also whether the return to "consistency" differs depending on the overall run environment.)
Consistency is better than inconsistency? (Fargo) —
If there is no optimal strategy from a managerial standpoint, would there nonetheless be an advantage to taking a team's consistency in run production and run prevention into account as an adjustment factor in Pythaganpat estimates of W/L?
Consistency is better than inconsistency? (Fargo) —
Just a gut intuition here but I suspect there's an analogy to stock investing. Those who try to "time the market" don't do as well as those who dollar cost average. In the case of stocks another factor is the transaction cost, but I think the logic still applies. You can't always predict what your opponent (or the market) will do next inning, and what they did the previous inning may not tell you that much (except on defense: change pitchers!). So do the little things right: keep investing steadily through thick and thin; don't rely on perfect timing or the home run. This, BTW, is why I always feel when the Tigers are playing the Angels somehow the Angels are generating the extra run here or there (often via alert and aggressive baserunning). They accrete runs.
EqA renamed TAv (True Average) (Fargo) —
@Mike/#86. (1) Re Bradbury: Colin criticized. Pretty much everyone else who commented criticized, including MGL and others from the outside. It was a thorough thrashing. I say "thank you BP" for helping to show its audience that the emperor has no clothes. (2) Re did BP know there were problems with last year's PECOTA's? They read the criticisms; they know that PECOTA didn't perform nearly as well against the other systematic forecasts last year as it had in prior years. They don't have to have a big public ritual flagellation to expiate their sins. They just have to get to work to make sure things work better. There's plenty of evidence that they’ve tried to do that -- read Clay Davenport’s and Dave Pease’s comments, for example. (3) What's wrong with PECOTA now? I think PECOTA is probably a significantly more complicated system than most of the other major forecasts. For one thing it is extremely sensitive to how the raw performance data on the referent population -- the population of potential comparables -- is evaluated and transformed (park effects, league and era adjustments, MLE's). I suspect that's where most of 2009's problems may have came from -- including the Wieters projection. What I find least explicable is why the problems we are seeing now weren't addressed 6-8 months ago. If I were undertaking the translation of the Excel-based system into the new code then the first thing I would have done is to replicate past annual PECOTA projections using the new code -- while making no other changes in the methodology or he database. After late March 2009, Nate was no longer involved with PECOTA. They could have begun the code conversion any time after that. But it doesn't appear that they started this until after the 2009 season was over. Only if I could use the new code first to reproduce Nate's results from 2008 or 2009 would I introduce any changes in the procedures. Did Clay first attempt such a direct replication? We don’t know. At some point, Clay appears to have made some important changes -- he refers specifically to broadening the database of players from past seasons. Not only for the weighting of each players' baseline performance stats for the previous three seasons, but more importantly by adding a lot more player-seasons to the potential set of comparables. It seems likely that those added player-seasons -- ones that were not in Nate's database -- include mainly players with little or no ML experience. So these are players whose stats are highly subject to any assumptions and imputations made using the DT's and MLE's. In addition they've had certain problems that seem specific to the Player Forecast Manager (PFM): taking the PECOTA estimates and plugging them in to the PFM. Now this is more complicated than it might at first appear to be -- chiefly because the PFM introduces depth charts and playing time estimates as constraints on the PECOTA estimates for each player. And lineups and batting order and the projected performance of all players on a team 'interact' to affect the projected performance of each every player as well as the team as a whole. In that way the PECOTA's aren't simply input to the PFM; they are simultaneously dependent on the projected performance of other members of the team. Nonetheless, the PFM worked reasonably well in previous years (maybe not so much in 2009?). Why has it apparently become such a problem this year? Is it due to efforts to change the PFM this year? Or was this, too, transformed from an Excel spreadsheet macro to a new system of instructions for the first time this year? And then there's the issue of how similar players are identified. It requires another complicated piece of analysis to search the database of player seasons who are most comparable to each current player – on the set of physical characteristics and performance indicators that PECOTA employs. This matching probably didn’t use an Excel procedure but instead some kind of analysis of covariance. Maybe that's where some of the problem lies this year. We haven't heard anything about this. Finally there's the step of using the performance of the set of comparables to make projections for each current player's 2010 performance. More complicated code but probably fairly straightforward, if I understand the methodology. In sum, there are a lot of places where things could have gone awry. As Tango wrote, we have to wait for BP to explain what's going on. As a long-time BP subscriber, I mainly want them to solve the problem. But my confidence in the PECOTA system has been shaken, and it will have to withstand two tests to rebuild it. First, I want to know why it took so long to get the new projections ready. Second, and more importantly, I want to know that PECOTA is once again among the very best at what it tries to do.
EqA renamed TAv (True Average) (Fargo) —
I see a lot of dumping going on, people grinding again on issues they've raised before. But very little analysis of the comparative value of EqA (whatever it's called) and alternatives. Did EqA predate or postdate WOBA? If they are comparable (as Colin Wyers pointed out on THT), why was the "second" one, whichever one that is, created? Ragging on BP for publishing the Bradbury article is truly strange. By publishing it they didn't endorse it. And both BP writers and the vast majority of commenters on the article pretty much rubbished Bradbury's analysis. That's a mark in favor of BP, not against it. As for the comments on PECOTA, the strongest and, in my view, most relevant question is why this year's problems are occurring in getting the individual player forecasts stabilized and the forecast manager (hence also the team forecasts) working consistently. While it does seem there were some problems with the 2009 projections, it's less relevant that BP explains what went wrong last year than that they learned that there were some problems and worked to solve them for this year. I don't buy the argument that PECOTA -- the internal logic, the basic method of using comparables -- suddenly broke, after it had performed fairly well since 2003. But what I would like to know is how much of last year's problem was due to issues are are mainly external to PECOTA's logic, such as issues in the DT's and MLE's, i.e., in the data that is the input to PECOTA's forecasts. It appears from Clay's latest BP entries that this year there are definitely issues brought on by changes that Clay has made in defining the pool of players from which the comparables are drawn. And further there seem to be problems with how the Player Forecast Manager is cutting up playing time and filling out the rosters for each team. Both of these problem are issues that are largely external to the logic of PECOTA though they are critical to the player and team forecasts. And this brings up the question that was raised by at least one commentator above: why wasn't this addressed much earlier, both by replicating and improving the 2009 forecasts and by modelling the 2010 forecasts -- for example, by anticipating that expanding the player pool, which had the effect of making the comparables contain a larger number of players with little or no ML experience, might throw a lot of things out of whack. Why wasn't this worked on last summer and fall? Why was it only faced in December, January, and February?
Girls will be boys (Fargo) —
I think those are great photos. Those players worked so hard, they won, let 'em celebrate.
Hockey Day (Fargo) —
The streets and hiways of Canada will be empty from 7:30 til 10:00 -- well maybe a quick run to get some more beer around 8:45. (For the Russians, Moscow is 8 hours ahead of EST, so they're pretty much going to be vodka'd up the yingyang before the game starts at 3:30 AM.) Russia didn't start out so hot either. Can't hardly criticize the Canadians for this quarterfinals showdown. Guess: Canadians wear out Russian defense and win. USA, with 2 days rest, dominates Swiss. The Czechs came within an eyelash of getting beaten by the Latvians yesterday (3-2 OT win). Not clear whether Jagr is fit to play (thanks to Ovechkin?). I think they lose to the Finns tonight. Swedes should beat Slovaks.
You can only keep one sport... which would it be? (Fargo) —
Tango, another question to ask is if people could buy tickets to just one Summer Olympics competition, which one would it be? Same for Winter Olympics. For me, Summer would be track and field. (Next up: gymnastics.) Winter would be (men's) hockey. (Next up: downhill skiiing.)
You can only keep one sport... which would it be? (Fargo) —
Just one? Depends on the purpose. For everyday interest (especially in season): baseball. For betting interest: football. For occasional intense interest, in a sport that I was once pretty good in and have a feel for: track and field!
PECOTA Standings (Fargo) —
It wasn't clear whether Clay was saying this 75th percent fix applies only to the PFM settings to use when using the latest versions of the forecasts, or whether instead he was saying that the PECOTA's in the book also have this problem. The latter would be a more fundamental error. Time will tell.
PECOTA Standings (Fargo) —
I don't think BP is saying the projections in the book are wrong (they are, however, always preliminary, i.e., not based on final depth charts and playing times). Steve Goldman just posted that some of the problems seen in the weighted means spreadsheets and player forecast manager are not in the book. Right now it looks like they're having problems generating updates to the PECOTA estimates as well as getting the PFM to work as it should. And Clay is working on this. I suspect these are more data management problems than problems with the basic PECOTA algorithms. But since Clay had to take Nate's humongous spreadsheet macros and turn them into a different code, there's certainly room for slippage. And in the past, Clay's job basically involved getting the input data (translated stats, etc.) ready so that the PECOTA's could be run. Now he has the entire process in his hands. So now he could conceivably be checking the whole computational process from data input and translation to estimates to reporting. I agree that it's embarrassing to BP to have these problems. They should have beta tested and worked out the kinks before they released the projections online.
Best Stadiums you've been to (Fargo) —
I like creaky old stadiums and small upright ones (stadiums where you get relatively close to the diamond even if you're in the grandstand), but there's nothing like Dodger Stadium on a sunny day. 1. Dodger Stadium 2. Wrigley Field 3. Fenway Park 4. Progressive (Jacobs Field) 5. Old Tiger Stadium 6. Toronto (in sunshine) (don't like it with dome closed) 7. Anaheim 8. Oakland 9. Shea Old and dead one: Candlestick (sucked) Long gone: Wrigley Field in Los Angeles (where L.A. Angles PCL team played before Dodgers arrived). Loved this one as a kid.
SIERA (Fargo) —
Should point out that this article isn't behind the BPro pay wall -- anybody can read it. I think that's a good precedent for any new metric that they develop and incorporate into their reports.
Why do we need personal opinions about factual matters? (Fargo) —
Just a comment on the title of this thread. It was allegedly Senator Daniel Patrick Moynihan who coined the phrase, "Everyone is entitled to their own opinion, but not to their own facts."
Proofreading a book (Fargo) —
I proofread my manuscripts very carefully, using various methods -- not just reading word by word but also reading aloud, searching for common errors, etc. But once something has been published I don't dare read it, because I just KNOW I'm going to see an error somewhere.
Anti-virus software (Fargo) —
Depite the hype, Macs need AV software too. Once Snow Leopard disabled my Norton AV, I went looking and tried Avast! but now have installed ClamXAV, which found a couple of things that both Avast and NAV had overlooked. But I'm really not sure this is the last word. If anybody has other suggestions, I'd appreciate them.
How to track a flyball (Fargo) —
In agreement with Peter Jensen, more from observation than playing: an outfielder uses all his senses, including hearing (cf. Adair's article), and getting a jump means paying attention to the pitcher's and hitter's tendencies, the base-out situation, the count, the elements (wind, sun location, the cut and dryness of the grass), the catcher's setup, etc. Further an outfielder chooses an initial location before the pitch is thrown and may also "shade" his stance or lean in a particular way. He can't just be reactive to the flight of the ball.
How to track a flyball (Fargo) —
Getting a good jump on the ball apparently depends on accoustics -- the "crack of the bat." Here's an article by Adair about this. http://www.acoustics.org/press/141st/adair.html
Psst... wanna work at Baseball Prospectus? (Fargo) —
You have to figure that a lot of subs are at much less than the $40 that Tango assumed. If they really have 20,000 subscribers, then many are re-subscribers who pay $35 per year and many are fantasy subscribers who pay $20 per year.
Psst... wanna work at Baseball Prospectus? (Fargo) —
BPro's revenue stream has multiple sources. 1. Premium subscriptions. 2. Advertising on the website. 3. Book sales (mainly the BPro annual). 4. ESPN sponsorship (most of which probably goes to the authors of insider articles, and some to PBro as an organization). I would guess that BPro's annual gross revenue is 1.5 to 2 times what Tango calculated from the subscriptions alone.
Trademarking "sabermetrics"? (Fargo) —
Wikipedia has had a "Sabermetrics" article since 2001. It gives a definition and the origin of the term. But according to Hahvahd Law School, "4. How do you acquire rights in a trademark? Assuming that a trademark qualifies for protection, rights to a trademark can be acquired in one of two ways: (1) by being the first to use the mark in commerce; or (2) by being the first to register the mark with the U.S. Patent and Trademark Office ("PTO"). 15 U.S.C. � 1127(a)." http://cyber.law.harvard.edu/metaschool/fisher/domain/tm.htm
BPro's cover (Fargo) —
"In your eyes therefore, it would be IMPOSSIBLE for me to post that picture without my beating a dead horse." You're probably right because you've been beating that horse for years. I don't think the hype is smart either, BTW. But who takes it seriously?
BPro's cover (Fargo) —
Tango: don't pretend you didn't post the cover to make the same point you've made many times before. As for the issue of using Nate's name, I wonder whether they cleared that with him. If I were him I wouldn't want my name on the forecasts themselves unless I did them. However, they could have worded that promo differently on the cover while still drawing on Nate's name recognition: "Featuring projections for 1600 players based on Nate Silver's deadly accurate PECOTA."
The latest (and last I hope?) JC thread on this site (Fargo) —
re #2: "At what age do economists start to decline?" You mean economists who never make it to a major league university? Kennesaw State University is Carolina League level.
BPro's cover (Fargo) —
Tango, you're beating a dead horse. It's hype for a book cover. B.F.D. Get over it already. You could make a far better criticism of this promotional cover if you wanted to: Nate Silver invented PECOTA but has had nothing to do with the 2010 projections, as Kevin Goldstein made clear the other day. This is basically in the hands of Clay Davenport now, and no doubt he will be tweaking the procedure some more, as he did last mid-season.
IOC (Fargo) —
While I was rooting for Chicago, Rio should be a great venue for the games. Let the samba begin! The point about whether a country's delegates can vote for their own country (or at all if their country is a contender) is not necessarily how many votes they have in a given "selection" but their bargaining power on procedural and planning issues, including decisions such as whether baseball is excluded from the Olympic sports. Also, there is logrolling on votes. I recall being in China when their first bid was being prepared. They were sooo disappointed that they didn't get the millenial games. But they were a lock for getting one down the road in part because they went along on that decision calmly, and methodolically prepared for the future anyway.
Unintended consequences or the price that some people pay for something that is good? (Fargo) —
Two comments. The Cash for Clunkers program is a clunker of a design. Agree completely that the size of the incentive should have been fit to the MPG differential. I got rid of a 20 mpg 13 y.o. Taurus for pennies. Bought a new car getting 38 mpg. Did get hybrid tax credit, which was far smaller than the $4,500 for the clunkers. Would have made sense to offer $600 for a 6 mpg improvement, and $100 for each additional 1 mpg improvement up to some limit (say up to 40 mpg). On MGL's OP: the core issue of one of "empathy," something that the right tends to denigrate. To me the concept has to do with the a sense of understanding how other people feel and considering not just their feelings but their actual well-being. Many people tend to think that whatever they have is theirs, even if earned largely through the labor of others or through inheritance or luck. They feel no sense of social responsibility. A practical aspect of this is that we've totally lost the sense of the purpose of progressive income taxes, or, for that matter, the dangers of inherited wealth.
Poll: Strasburg or Verlander? (Fargo) —
Why not go with B-R's spelling (Jurrjens)? That's pretty much where people are going to be looking up info on him. That's also the spelling by MLB.com.
Poll: Strasburg or Jurggens? (Fargo) —
Jurrjens has been underestimated his "whole" (very short) career. The Tigers gave him away and had nothing much left on their pitching staff last year. He's not impressive to watch. He's not a big K guy. But he's incredibly heady for a young pitcher.
Blogspot blocked (Fargo) —
With your own laptop and a mifi in your pocket or a USB modem you would be able to log in any time anywhere. About $60 per month. http://www.usatoday.com/tech/columnist/edwardbaig/2009-05-13-mifi-wireless-mobile-hotspot_N.htm
Diamond Mind baseball + leading forecasting systems = standings (Fargo) —
As I have commented on before, some systems come up with updated versions as the depth charts change during the preseason. The DM projection for BP (PECOTA) is obviously based on an early version of the numbers and perhaps also not even using a BP-based depth chart but instead only the early individual player projections as published in the annual or the first weighted means spreadsheet.
YouTube of the day (Fargo) —
And I was watching the guy in the white shirt who got creamed by the dunker coming over his back.
The Studes interview (Fargo) —
@Dackle: I've heard of something called the "rank-size principle" in urban settlements; maybe that's what you have in mind. On this whole issue I think the revenue has to be thought of as all-sources: subscription (if any), advertising (if any), publication sales. If you broaden your concept of "baseball sites" beyond those that have a major focus on sabermetrics, clearly some others such as Baseball America are drawing revenue from all three streams. And I suspect that BP's revenues are coming to them partly for the expanded coverage they've given in recent years to prospects (read: Goldstein's columns), and partly from adding more "common" coverage (read: Perotto and others). IOW, the audience and potential subscriber base for "sabermetric" analysis is probably smaller than the subscriber base of BP. Perhaps most importantly, you have to figure that the audience of fantasy baseball information is much much greater than the audience of sabermetric analysis. And the one product that BP has that feeds that audience is PECOTA. Without PECOTA I bet BP would lose half of its subscribers. I don't now how much money the other fantasy sites make, or the fantasy services offered by some of the big news sites. But right now you have most of the best sabermetically based "forecasts" being given away on their non-subscription websites! PECOTA and Shandler are prominent exceptions, and their sites and books appear to make enough money to sustain a fairly large stable of writers and analysts. (BP also gets what appears to be decent advertising revenue as well as revenue from subs and book sales.) At the very least, I think you all need to give some thought to what products you have that can appeal to a bigger paying audience -- and this has to be fantasy players, and the product has to be preseason forecasts and tout sheets, coupled with in-season updates, player injuries reports, prospects, trades, etc. Fantasy players feed on that stuff. And then you have to question why you're giving it away -- other than your devotion to baseball stats -- and whether it's this part of your product that you most need to try to sell for cash.
The state of fielding sabermetrics in MLB (Fargo) —
I agree the the idea that at base the grading has to be observational (but some combination of this and "objective" metrics is also reasonable). But what I would propose as a way to reduce the costs would be that a sample of trained observers. 100 seems reasonable but that gives a +/- of about 10% for any given observation/play; however, but if they observe all plays in a game the event log grows a lot so your number of observers can probably be much smaller. Also I'd have those observers grade the fielder's actions even when they aren't "in" on a play (nowhere near the ball, for example, but are expected to move to cover a base, e.g., the pitcher covering the plate when the catcher has to go after a nubber or a PB/WP). The "grading" can be done on the basis of video, which could be rerun, and the video could even be enhanced by putting electronic markings on the ground to indicate location on a grid. The "observers," then, could grade one player for a game, or one position for an entire game -- that player's movements would be captured in video, with a "pic in pic" perhaps for larger simultaneous views of what's happening on the given pitch or play (not just what that player is doing). (An occasional cutaway to a particularly luscious fan in the stands could be a welcome tension breaker! Sorry guys, no beaver shots!) Grade each player for ca. 16 random games per year, i.e., about 10% of the regular season games -- but grade them on every pitch/play for those games. The observer panel can do this at their leisure and submit their grades online. It would be a heckuva interesting experiment. You'd definitely need to develop rubrics for grading and to train the observers. I can't calculate the statistical "power" you'd need -- you might not need 100 observers of each play(er) to get what you want if the different plays over the entire game are pooled together as observations. Maybe 10 trained observers would be needed for each position. A competent sampling design expert would be able to figure this out.
The state of fielding sabermetrics in MLB (Fargo) —
To add to a possibly defunct thread, one thing that occurred to me is a term from football teams. After a game, the coaches "grade" each player. I don't think they do this by reviewing a lot (or even any) pbp film. But it's SOP in football. Imagine in baseball if a team could do this for each defensive player. Maybe not every game. But imagine that they literally watched a player play an entire game -- where he positions himself, what he does on every pitch and batted ball. Do the infielders move to cover the base, do they move to the right cutoff location, how quickly do they get off the throw, do they throw to the right base, how accurate is the throw, how strong? Imagine that by truly tracking a player's every move on every play you could learn whether he's doing the things that are going to put him in a position to make a play, to back up the play, etc. Would a baseball player accept the idea of coaches "grading" him based on this -- giving him a score for a game even on the kinds of things that Tom has asked us fans to do (but only) for a player's entire season? The player wouldn't be terribly receptive, I imagine if he hadn't made any errors, hadn't been out of position on any play he was expected to make, and so on. But the smartest team players are going to be in position to make a play on every pitch, and many of the things he does do not show up in any of the standard stats. And a smart team might evaluate and train thair players by doing such careful grading. Why couldn't this be part of the assessment of how well players play defense?
Joining SABR? (Fargo) —
Just a comment on why people join professional associations. Lots of academics belong to more than one association. Most of those associations offer (a) a journal, (b) an annual convention. Analogous to SABR. For some years now, if you're affiliated with a university that has a good library, you've been able to get instant online access to thousands of professional journals without being a member of the sponsoring association. So why join the association? Answer: for the meetings. You often still have to pay a registration fee (maybe less than nonmembers, or maybe nonmembers can't attend at all). When some professional associations at first hesitated to make their (recent) journal issues available online -- thinking that this would discourage people from paying dues -- they learned that in fact this isn't the case. Perhaps the opposite was true: if you can "sell your journal" (by getting libraries to buy access to online versions of it, perhaps through some kind of publishing consortium), then you may sell the value of your association -- people will come to present papers, get feedback, meet other experts and up-and-comers, have fun. Following this logic, only if you're interested in the meet-ups are you likely to join SABR.
O.J. thread (Fargo) —
I used to be a real fan of OJ the football player. After that -- as actor, promoter of Hertz -- not so much. I loved one thing he once said about why he always seemed just a bit slow -- taking an extra couple of seconds to get up from the ground -- after a play. "I don't want them to know that they hurt me. I get up slow whether they hurt me or not." That was a signature of his. I think he was guilty as hell of that double murder. A jury of his peers found him not guilty. And the state can't appeal a not guilty verdict. That's the system. It's not perfect. It's supposed to favor the accused against an overreaching government. I agree with you that the crime for which he has just been sentenced was over-charged or shouldn't have been charged at all. The sentence is way out of line with any harm he did in this instance. The only explanation is that this is the law enforcement "system" getting even for its earlier failure to convict OJ, plus the fact that OJ has no friends in high places. Scooter Libby will probably be pardoned by Bush, for committing several felonies including perjury and obstruction of justice. And OJ will rot in jail for at least 9 years for stealing back his own stolen property from some of the most unsavory characters around.
The FairTax and other "flat tax" schemes in lieu of income and payroll taxes (Fargo) —
Every version of a simple or so-called flat tax that I've seen turns out to be deceptive for one basic reason: either it's a tax on consumption or it's a tax on earnings. "Unearned" income (read: income from investments, savings, inheritance, real estate gains, etc.) is totally excluded. Thus such a tax is ultimately regressive by excluding from taxation a very substantial part of the income of the rich. You have to read the fine print to ask what kind of income is in the base that is subject to tax. Does the print say "earned income?" It's not truly a flat tax.
Baseball Prospectus free for all, for now (Fargo) —
Tango, Looking that Nate's list, I'm reminded about the Elo application that he made to professional baseball (and that Clay Davenport also incorporated into team season projecttions). Looking up Elo on Wikipedia I see that such ratings exist for some other major sports including basketball (Sagarin uses them in part). But I don't see one for hockey. I'd love to see one for both NHL and D1 college hockey (which maybe USCHO could add to their power rankings?). Any chance you might take this on -- just for fun and for us hockey fans?
Tiger Woods? Tiger Wuss! (Fargo) —
Dull? The last two games were electric, IMO. I only had NBC to watch. We used to get CBC on our basic cable but no longer, so didn't have the option to compare the two coverages. I'm going to miss CBC when the Beijing Olympics come on because they always do a better job, and broadcast for more hours, as well as more events in "real time," than do the American networks. No insult intended here but I think one reason CBC has better coverage is that they pay more attention to the preliminary rounds and the heats while the Canadians are still competing, whereas the U.S. networks seem only to be interested in the finals and often don't even stay tuned to the entire event (say if it's a long track race).
Good and bad pitchers' hitting (Fargo) —
"Why Pitchers Can't Hit," an article published by Silver on Slate a few years ago: http://www.slate.com/id/2099612/
Letting pitchers bat: It never ceases to amaze me... (Fargo) —
Thanks, guys. I guess I was just thinking that WE or other "expectancies" are based on what are the general or overall or "large number" average probabilities. I'm thinking that while certainly it's an important baseline for evaluating managerial decisions, a more explicit modeling of some of the complexity might be worthwhile. For example, if you were to systematically study managerial decisions in various areas (let's say you were trying to come up with a standardized metric for evaluating managerial competence), then you might test for the effects of various factors in accounting for the "mistakes" (cases where their decisions appeared to be counter to what was seemingly rational). Then you could perhaps understand that that there is a method (perhaps an accepted "rule of thumb") or perhaps a "psychology" to such decisions, just like that study of football decision-making in which coaches were found to take the field goal rather than go for a first down far more often than was "rational." In training managers to make "correct" decisions, then, you wouldn't just drill a WE matrix into their heads, but you would also address which other factors might be worth taking into account in certain game situations.
Letting pitchers bat: It never ceases to amaze me... (Fargo) —
Thanks, Ryan. I guess what I'm thinking is that WE is a very static kind of factor to consider; that is, it presents an "average" probability (of winning) that, while situationally accurate doesn't allow for the multiple factors that are involved in decisions at any point in the game. Just those considerations that you mentioned -- remaining outs available in the game, available pitching resources and pinch hitters -- could be explicitly modeled, instead of being subsumed by the "average" WE associated with the given situation. To take this a step further, there are also "next game" considerations or "series" considerations. Do I burn a relief pitcher today who I may need tomorrow (or who I used yesterday)?
Letting pitchers bat: It never ceases to amaze me... (Fargo) —
Thanks. Has anyone tabulated "runs lost" due to allowing pitchers to bat with RISP? Following the original post, this could be used as a systematic measure of the quality of managerial decision-making. (Along with other quantifiable questionable managerial decisions, e.g., bunting, steal attempts.) It's probably a lot easier just to look at NL data and calculate PA's by lineup order and then, since pitchers are almost always in #9 slot (except if there's been a double switch during the game) figuring out how many PA's are saved for "better hitters" by comparing total PA's by pitchers with total PA's for #9. But getting such a calculation for cases with RISP would be pretty hard, I suppose. And wouldn't you have to look at the RS in the inning after the PH occurred, and compare it with the expected RS without the PH to net out the gain from removing the pitcher? (Not a data maven here.)
Letting pitchers bat: It never ceases to amaze me... (Fargo) —
This isn't a comment on the particular decision by Cox to let Reyes hit. But how much, if it all, should one weigh in both the run/win expectancy of pinch hitting for the pitcher in a given situation at a given point in the ball game and the assessment of how well the pitcher is performing on the mound, and the availability of better substitutes at that time? If the manager thinks the pitcher will hold the opponents for a couple more innings, and the chances are decent that his own hitters will score another run or two, it's probably still wrong to let the pitcher hit but the calculus is more complicated than has been allowed for. Another way of putting this is if the situation were similar in terms of runners on base and outs, should the manager make the same decision whether the batter is a pitcher or a position player? Does the expectation of runs gained by pinch hitting always trump in that situation? And wouldn't that be just as true in the 2nd inning as in the 6th? (Should a .117 BA pitcher ever be allowed to hit with RISP? When and when not?)
Make me a deal for Leverage Index (Fargo) —
Is "Leverage Index" a copyrighted name? If it's in the public domain, why do they need to offer to pay you for it? Answer (I think): you created it, and they want to keep other game manufacturers or any other commercial venture from using the term. Presumably that would also keep MLB from adopting the term as an official statistic as well, unless they paid a royalty to whoever owned the name. (I assume they're not going to call it "Tango's Leverage Index" and therefor want to use your name as well.)
Can we be frank about the Democratic candidate for POTUS? (Fargo) —
Re #14: The results in IN and NC were very much as expected if you had the right model, which (a) almost nobody did, and (b) didn't depend too much on the polls. To use a perhaps faulty analogy, those who relied on polls and the image of a horse race with late deciders, etc., are like baseball analysts who rely on short-term random factors to determine the underlying quality and skill of the players. There was one forecasster who relied more on fundamentals (demographic analysis) who has nailed the last three major primaries within a few points. In yesterday's primaries, he had projected +2 for Clinton in Indiana and +17 for Obama in North Carolina. Here's the site: http://www.fivethirtyeight.com "538" is also known as "poblano" on DailyKos.com. He lays out his methodology in great detail on his 538 site.
Forecasting Standings (Fargo) —
Re #7 and other comments on PECOTA projections. The full projections by team posted above were the preliminary ones before defense was considered AND before depth charts were set. There were many changes in the predicted wins in the revised projections based on establishing depth charts. If last year is a precedent, those team projections will be revised a couple more times before the season starts but further revisions will be quite minor. If anybody wants to see the "latest" (and they have asubscription) then they should go to the "Depth Charts" here: http://baseballprospectus.com/fantasy/dc/
Are Consistent Batters Easier to Project than Inconsistent Ones? (Fargo) —
MGL, I have no inside knowledge of PECOTA methods. I've tried to figure out how it works, and I agree that it's an amalgam of "Marcel" with its reliance on last 3 years for a baseline (plus aging and regression adjustments), and "similarity scores" with their reliance on comparable players for the forecast. The biggest mystery about it, it seems to me, is the method of matching individual players to comparables in the database. The most thorough discussion of this that I'm aware of is in the BP 2003 annual, but there's not enough information there for anyone to mimic it or reverse engineer it, it seems to me.
Time: Bill James (Fargo) —
Tom, a question for you. Is the regular season vs. playoff scoring environment as different in most ML sports as it is in baseball (as percent of base)? Maybe not in soccer. What's the situation in hockey, American football, or the NBA and NCAA basketball? In any case, don't you have to correct for the quality of the teams or competition in the playoffs (i.e., selectivity or survival, especially the further along you go in the playoffs, say from first to second to third rounds), and not only look at aggregate differences between regular season and playoffs? Maybe the playoffs in baseball are more selective for quality because 8 of 30 teams qualify, while in other sports a larger proportion of teams qualify?
Are Consistent Batters Easier to Project than Inconsistent Ones? (Fargo) —
On some of those nearly "incomparable" players, especially when they outlast everybody (maybe because they were juiced?), Pecota relaxes its rules on age/position matching, but it still seeks "comparables" using this relaxed set of criteria. And Pecota relies on this "relaxed" set of comparables for its forecasts. I don't think it completely defaults to a Marcel. But I suppose you could determine this empirically (de facto) by direct comparison of Marcels and Pecotas for such rare players. Not sure you'd have enough cases to do such a test.
Are Consistent Batters Easier to Project than Inconsistent Ones? (Fargo) —
As you know, Pecota isn't a Marcel-type system. It requires matching the baseline performance of a given player to a set of historical comparables. It uses some non-performance characteristics (body type, handedness, position) as well as performance stats to find the comparables, and it often uses peripheral stats in the process. It's the future performance of the comparables that represents the "projection" of the given player's performance. I think it's possible that the match to comparables doesn't work as well for some types of players as it does for others -- e.g., for someone with an unusual combination of performance and nonperformance characteristics. In effect, Pecota may be "fooled" in such a case because the case doesn't fit a canonical pattern of relationships between the variables in the model. Pecota doesn't default to a Marcel if the similarity scores used to identify the comparables are really low (maybe it should -- thus becoming a hybrid model). Marcel wouldn't be fooled so easily by any particular "type" of case because he's just a dumb monkey who always says that what a player's going to do this year is pretty much what he did the last few years. No peripheral stats, no speed scores, height, weight, or position to find "similar" players. I can imagine Silver inspecting and probing certain individual cases, especially those of outstandingly talented players, as a way to evaluate his system, but once he's set up the formulas he doesn't "overrule" the system's projection for particular cases. I see him "disagreeing" with Pecotas for particular players in some of his articles, but he lets the system "speak for itself" in his statistical projections.
Are Consistent Batters Easier to Project than Inconsistent Ones? (Fargo) —
If Ichiro is as unique as Silver says he is, then no system is going to capture him well. Obviously, Ichiro has been a puzzle that he's considered and can't solve in any general way. http://www.baseballprospectus.com/article.php?articleid=3497
PECOTA chat at SOSH (Fargo) —
IIRC Nate Silver announced plans to introduce a comments feature last year, or maybe even two years ago, but they never implemented it. Given what's been mentioned here (risk of spam, cost of monitoring), I bet this feature will be for subscribers only.
Changes in home run rates during the Retrosheet years (Fargo) —
The article by Jay Jaffe of Baseball Prospectus addresses changes in the ball: http://baseballprospectus.com/article.php?articleid=7167
Changes in home run rates during the Retrosheet years (Fargo) —
Great detective story, Dr. Tango. Does anybody have a cache of "historical" MLB bats available for testing? On the umpire theory, can you use Retrosheet to test for same pitcher-batter-umpire combinations? Or just for umpire effects (HR/contact) within and across seasons?
Is Congress Doing its Due Diligence (wrt Clemens and co.)? (Fargo) —
The purpose of the hearing was (IMO) two-fold: (1) To showcase some Congressmen in action (what else have they done recently that was memorable and casts a favorable light on the the House?); (2) to sucker somebody into perjury so that they could refer the case to the DOJ, since there was no way to prove or disprove the factual assertions from any available evidence or testimony. It's an American style "show trial," meant more for show than for trial since this isn't a court (even a court of public opinion), and there are no charges on which anybody is being tried.
Quantum Leap or Wiseguy? (Fargo) —
Re Phil Birnbaum's comments: "—Get Dan Fox’s baserunning metrics up as a stat report particularly when there’s room for PAP, umpires, and RBI opps." I think this is in process now. See here: http://www.baseballprospectus.com/unfiltered/?p=683 "[H]ow does PECOTA work? I mean, in detail, so I can code it myself. If that already exists, I haven’t seen it. If they want to keep it proprietary, that’s OK too, but then it’s useless to me as a researcher. . . ." It's proprietary, as are most projection systems. All praise to Marcel, however.
Free Dan Fox! (Fargo) —
I agree that even if they keep doors closed on most of their recent stuff they should open their archive, perhaps using something akin to what many journals do when the have their back issues on JSTOR. They use some kind of "moving wall," so that the journal content is available on JSTOR if it is, say, more than 1 year since printing, or more than 3 years (journals differ). In the case of BP, it would be reasonable to take the restriction off of any article that they published more than 12 months age. Thus, any article published before December 2006 should be "open" today. And next month they just move that barrier ahead another month. That would make a lot of their materials not only accessible but citable -- and would benefit BP's reputation on both accounts.
Goalie Masks (Fargo) —
Then again there was the face that Terry Sawchuk used instead of a mask. Those were amazing days.
Keith Law interview (Fargo) —
Tango, were you intending to brand the individual who uses "Lies, Damned Lies" as the header on his regular columns as a rank and arrogant number cruncher? Whatever you intended, my view is that many blogsters and other regular columnists in the MSM who write about baseball -- whether from a sabermetric or nonsabermetric perspective -- are a distinctly opinionated bunch. But that's just what many readers want -- opinions, evaluations, not just descriptive analysis. Not "two-handed economists" ("on the one hand, on the other hand") but people able and willing to make their best expert judgments.
Personal Platoon Splits (Fargo) —
Dan Fox calls to our attention (and reprints) a BP analysis of his own from 2006 on platoon splits. http://danagonistes.blogspot.com/2007/11/of-crowds-and-splits.html
Bill James, Online (Fargo) —
Tango, you should also run a regression and see whether the intercept (constant) is significantly different from 0, and the slope significantly different from 1.0. You could have R=1.0 and still have different slopes, intercepts, and mean values.
PAP (Fargo) —
Thanks, Tango. So (1) hitters are more patient," which in effect means that if pitchers throw 108 pitches they will throw one fewer inning on average today; (2) managers of yesteryear reached early decisions to pull pitchers who started poorly, thereby assuring that they had more starts than they might have if they were allowed to finish 5 or 6 innings (to what extent can this be shown in the data? are "modern managers" maybe realizing that back luck can lead to a poor first couple of innings?); (3) if a pitcher gets to 8 innings, the manager is likely to give him a chance to complete the game (contingent on size of lead?). Perhaps another factor is the number of pitchers in the basic rotation, and hence the number of starts the top pitchers can expect to have. Seems plausible that these are explanatory factors, but perhaps begging the question about the changing use of the quick hook over time, and how much it may be contingent on the pitcher's performance and the score of the game (or the overall run-scoring environment in different eras). Also, the drop in total CG's still seems rather more steep than might be accounted for in the terms explained above. Perhaps need to do both intertemporal and comparative cross-sectional analysis (within a season in 1960's and within on in the 1990's or 2000's). Anyway, this is all very interesting. Thanks a lot. I'm
PAP (Fargo) —
Thanks, Tango for the link to your pitch count summary. (I wrote my previous message without seeing yours.) The figures for Koufax comport with my memory. But it's also likely true that the average starter (or the average highly successful one?) was more likely to complete the game with that ca. 110 pitches, whereas the later-day pitchers do not. So wouldn't a CG/Pitches ratio for starters would be far lower today than in Koufax's day? And if so, is it just a change of style of the game, or perhaps, as you are suggesting, the avoidance of extremely high pitch counts for ineffective pitchers? (Again, trying to avoid a circular argument here.)
PAP (Fargo) —
Added note. It would be circular reasoning to argue that the reason there are fewer CG's in recent decades is the rise of relief specialists. (It would also be ahistorical to attribute the decline in CG's to "PAP," since the latter is a more recent invention than the decline in CG's. Also, I think we started "saving the arms" of young pitchers, i.e., in Little League, at least a couple of decades ago not just by outlawing the curve ball; only more recently have strict rules been implemented about pitch counts but I'm not sure when that happened.) So how do we come up with a non-circular and non-tautological explanation for the decline of CG's?
PAP (Fargo) —
Leaving aside PAP, why is it that there are so few complete games these days compared to, say, three or four decades ago (even among those "top 25 in PAP")? Because managers believe in PAP (or are just counting pitches) and are making the stupid decision to 'save the pitcher' from risk of injury? Because pitchers decline in effectiveness at an accelerated rate after a certain number of pitches? Because batters are behaving differently? Or there some other reason or combination of reasons? In short, what happened to the complete game, and why? I do recall Dodger broadcasts back in the early 60's in which pitch counts were routinely reported, at least at the end of the game. I don't know how common that was, but I think Alan Roth must have been compiling such stats at that time and feeding them to Vin Scully and Jerry Doggett.
The Stats Man (Fargo) —
I hadn't heard of Scout School, but with a little Yahooing I found this link on MLB.com: http://www.mlb.com/news/article.jsp?ymd=20030211&content_id=199087&vkey=news_mlb&fext=.jsp&c_id=null
Stephen Colbert for President (Fargo) —
How can anyone possibly support a guy like that for President? He uttered the m-f word on the air last night. It was quite wonderful. I'm sick of the whole lot of current official aspirants on both sides. (Well, Ron Paul is definitely a breath of fresh air.) Bring on the interesting candidates like Colbert. Where's Dick Gregory when we need him? I still have a "dollar" with Gregory's face on it -- recall that they used to work in dollar changer machines and so were banned. Even so they were more valuable than the Canadian dollar at the time. (Sorry about that one, Tom; though I'm pleased to note that the Canadian dollar's on a roll now, at least relative to U.S., so let those Canadians come south, or east or west to the States to shop til they drop). How about Pat Paulsen? Another odd candidate of the past. Even Ross Perot would be a blessing in this marathon electoral race.
Why the Phillies, Cubs, Yanks, and Angels lost the DS (Fargo) —
They stunk up the place?
Israeli baseball league play-by-play data (Fargo) —
Should have added that in addition to having 7 inning games, they settle ties not by playing extra innings but with a shootout -- home run derby. Shades of international football and hockey.
Israeli baseball league play-by-play data (Fargo) —
There's a nice entry on the Israel Baseball League on Wikipedia, with a map and other interesting information. I believe they play 7-inning games, so this by itself raises potentially interesting questions about the effect of shorter games on the competitive environment, the possibility of using smaller pitching staffs (and shorter intervals between games, less concern about pitch counts), and so on. http://en.wikipedia.org/wiki/Israel_Baseball_League
All you need is 4 or 5 pitchers (Fargo) —
Without checking the calendar from way back, it looks like these results are consistent with my recollection. He did not pitch Game 1 in two of these three series.
All you need is 4 or 5 pitchers (Fargo) —
In addition to Koufax and Drysdale the Dodgers also had, if I recall correctly, Johnny Podres and Stan Williams who started a large number of games (also Roger Craig for a while), and they had Clem Labine in relief. Koufax, in any case, would typically miss or delay his October starts because of Yom Kippur, so that would somewhat limit the number of games he could pitch in the WS.
Eric Gagne (Fargo) —
[quote]Has anybody ever mentioned the possibility that Gagne just isn’t any good anymore?[/quote] Lots of people, including Nate Silver in the article that was the stimulus of this thread: [quote]In the two most reliable indicators of pitching performance — strikeout and walk rate — Gagne has been exactly the same pitcher in Boston that he was in Texas. And that is a pitcher with decent but unremarkable talent. The 2002-2004 Eric Gagne was incredibly good. The 2007 version is — I don’t know — a rich man’s Ryan Dempster. Now, it’s not like Ryan Dempster is great shakes. But Gagne is not nearly as bad as his ERA suggests.[quote]
Baseball Prospectus is dead. (Fargo) —
BP guys (just like guys out of other shops) have come and gone and come again in MLB organizations. Former BP'er Keith Law was with Toronto, and now is with ESPN.com. Two other former BP'ers (James Click and a staff guy) are with Tampa Bay now. Keith Woolner is with Cleveland, as you mentioned. There may have been others. The above information comes from Wikipedia and therefore may or may not be accurate. There don't seem to be a lot of careers for real statistical analysts and theorem testers (as opposed to data managers and mere stats mongers) within those "small businesses," as Tango describes them. The larger baseball analysis business probably offers a lot more potential for careers in the media, fantasy baseball analysis, scouting and prospect reports, and so forth. Most of the well known sabermetricians appear to be amateurs in the sense that although they may earn some money from their sabermetric work (writing books and articles for online or mainstream media), they earn their living mainly by holding down "real" jobs. And most do their sabermetric work simply for the love of the game, statistics, analysis, or who knows what -- but not for profit.
Baseball Prospectus is dead. (Fargo) —
I read that column by Huckabay in the context of the last serious article I read by him, and that was in the 2006 Baseball Prospectus annual book, which I just pulled off the shelf: "Where Does Statistical Analysis Fall Down? Reality and Perception." I don't know what it was that drove Huckabay away from BP.com for 3+ years, but I think he's been brooding over one issue in particular: that MLB organizations do not really want to pay for the quality of analysis that they could really use, and that to the extent that they recognize there is valuable research out there, such information is largely in the public domain. In fact, the findings ("The Book," for example) can be bought for pocket change. Because MLB's only open to a couple of big ideas at a time, and those ideas are a public good, it isn't going to hire the Tangotigers of the world on a full-time basis. Now these conclusions are debatable. But I think from re-reading that BP 2006 article, which includes an interview with a cynical MLB executive, that Huckabay has been in a funk about the limited ability of analysis to improve the game. Beyond this, however, anything else he may be saying about the value or validity of baseball analysis just doesn't make a lot of sense to me. (I also think, from reading some of his comments in his "chat" on BP yesterday, that he realizes how inarticulate he was in that article.) Nor could anyone on the BP staff other than Huckabay endorse the statement that analysis is dead.
Pitch Trajectory (Fargo) —
I don't think a lot of batters make the decision whether to swing or take just on what they observe after the pitch is released. Obviously they need to figure out the velocity and location, but they also arrive in the batter's box with some prior expectations that may affect how they react to what they see. And they have an at-bat strategy that affects how and how quickly they make a decision whether and how to swing at the next pitch. Consider the following as just some of the auxiliary considerations that come into play. 1) The count; 2) The man on base situation and who is on base; 3) Is there a "play" on -- e.g., hit-and-run, bunt sign; 4) What kind of pitches does a pitcher have in his repertoire (or what kinds does he have command of today), and what kind does he like to throw in certain situations (information that may help the batter know what to look for); 5) What kind of pitch and in what location the pitcher threw to him in his last plate appearance, or last time in a similar situation; 6) Whether the pitch is hittable (even if it's not a strike; but even if it looks like a strike, if it's not already 2 strikes against him, the hitter may pass on the pitch if he can't get good wood on it); 7) Whether to try to hit the ball in fair territory or just foul it off. All of these things and more are in the batter's head before the pitcher winds up, and all may affect the decision to whether (and how) to swing as well as how long it takes him to reach that decision (for good or for ill); Seems to me someone has to work through the batter's logic, not just his ability to pick up the pitch and get a bat on it.
Did the Padres get rid of Bochy because they are smart and he is not? (Fargo) —
I take back the term "Sabermetric School." I'd call it a "Baseball Strategy Academy," and focus on strategic decision making on the field and in games, in personnel management, and in finances. And draw "instructors" from the baseball business as well as from the baseball analysis business.
Did the Padres get rid of Bochy because they are smart and he is not? (Fargo) —
I don't see how that would work at all on a practical level, in part because the strategy decisions often involve assessments of the capabilities of the personnel on hand at the moment (which the field manager may have a better handle on than anyone in the front office). A more general approach would be to organize a "sabermetrics school," in which you would use your sim, the product of research in "the book" and other sabermetric research (e.g., on player valuation and business decisions). A three-day intensive seminar for field managers and front office personnel. Paid for by MLBAM, perhaps. In other words, train them in what is known from a statistical standpoint, have discussions about strategic decisions and how different managers might handle them (use real, known decisions, lots of video, and build the stats and odds into the discussion, but don't just teach stats and odds). My working assumption is that most people don't read -- not even these wonderful blogs -- but they might enjoy "going to school" for a few days in the off-season. You've gotta involve more than just statheads in the training, however. . . . Banish the powerpoints, or at least minimize them. Lots of videos and "scenarios" that lead to interpretation and learning.
Did the Padres get rid of Bochy because they are smart and he is not? (Fargo) —
anon, the sampling error with a sample N of 1000, and the proportions of responses roughly divided 50-50 is about 3.1%, not 1.6%. Here's a sampling error calculator: http://www.dssresearch.com/toolkit/secalc/error.asp You are absolutely correct about the difference between sampling error and bias. At the same time, you might note that a lot of the observed variance is due not just to sampling error but also to other factors (poorly worded or confusing questions, respondent disinterest, coding/recording errors by the interviewers, etc.). Sampling error is only one source of the error variance.
Did the Padres get rid of Bochy because they are smart and he is not? (Fargo) —
MGL: Thanks for reminding me that trades were for contracts and not players, and thus the valuation of such trades has to take into account more than just the future performance of the players on each side of the trade. (There are, of course, other arcane rules that may lead to such trades, e.g., to the Tigers' trade of Ledezma to Atlanta earlier this year.) Pizza Cutter: What's the average number of pick-off throws that pitchers make? Other than the distraction factor, how much might a baserunner's ability to attract pick-off attempts or "throw-overs" wear down a pitcher?
Did the Padres get rid of Bochy because they are smart and he is not? (Fargo) —
"That does not bode well for his reputation in my book." You should write a book about it. Seriously, has anyone done a thorough analysis of the net gains-losses from trades? How many of these seemingly valuable properties are lemons (or go sour)? How often is there an essentially good trade, i.e., an even trade for value -- even if one team is seeking to exploit immediate value and another is seeking to benefit from that value in the future?
Where is the strike zone exactly? (Fargo) —
My hunch is that umps adjust for the type of pitch(er). One way that we all recognize (or think we recognize) is that certain pitchers are "given," the outside of the plate ("the black" and beyond?) and others aren't. Another way is that an experienced ump knows if a certain pitcher likes to throw backdoor strikes. Even the catcher may cue the ump in, not just by framing the pitch but by a comment. Not to take you too far off the main course, but it seems to me that some pitcher-specific analysis of sz's for pitchers with "known" tendencies, could be useful. Do they get "significantly" more called strikes in a given location than average?
Where is the strike zone exactly? (Fargo) —
One more question, related to front vs. back of the plate (or front of plate to catcher's mitt), do the data reflect whether the ball ever passed through the SZ or do they capture where it is at the leading edge of the plate? What happens to "back door" sliders, etc.? That may be one reason that Tango wants to see breaking ball splits.
Where is the strike zone exactly? (Fargo) —
Pretty interesting stuff. One thing to pursue for sure, given the asymmetrical relation between the handedness of the batters and the width of the zone, is the handedness of the pitchers, or more particularly the relationship between the shape of the zone and the combination (or interaction) between hitter and pitcher handedness.