THE BOOK--Playing The Percentages In Baseball

Don’t laugh.  Downside variance actually has a history of use by practitioners.

Some financial professionals to this day use the Sortino ratio.  For skewed deviations like those that characterize security returns, this actually could make a tiny bit of sense, though in practice it boils down to brokers saying, ‘volatility is good on the upside,’ then throwing away half their data.

How skewed are distributions of baseball performance?

Prior to that at bat, Crisp was 2 for 4 with a BB against Bradford lifetime. Playing devil’s advocate, perhaps Trembley was thinking that the rest of his relievers were so dreadful that Bradford may be the best option even against a LHB. That consideration makes the decision more reasonable, if still not correct.
Also, Bradford got a huge out against a LHB (Drew) out the other night, so that may have factored in as well. Of course, that fact should absolutely have not factored in.

wcw, whatever you are talking about (over my head), I have a feeling that is not what Trembley means.  Maybe I am underestimating him though.

I am sure that Trembley had a “good reason” in his mind.  Does that make the decision “reasonable?” I would have to look up the definition of “reasonable.” With “reason?”

While I would be curious to hear his reason, it would not matter at all (unless he came up with some reason that I had not thought of that actually had merit, which I doubt 99.9%) in terms of the utility of the decision.  Bradford is dreadful (going forward, and even in reverse in this case) versus LHB and should not go near one.  The only time he should EVER face one is in very low leverage situations when the manager just wants to save his pen, or when he comes in to face a RHB and the opposing manager pinch hits with a LHB and even then the pinch hitter versus Bradford would have to be worse than the original batter versus another reliever, assuming the pinch hitter could have been anticipated (and would come in almost 100% of the time).

As far his other pitchers, as I said, it doesn’t matter who they are.  Bradford verus a LHB is worse than replacement level.  You could bring in a AAA pitcher who would be better versus a LHB.

My guess is that Trembley would have mumbled something along the lines of what you said - one, he did fine against Drew the other day (why was he pitching against him?) and that his other relievers have not been pitching well, both of which being irrelevant information in this case of course.

In any case, the situation screamed for one of your better relievers, if not your ace, a high leverage situation with 2 outs in the 8th.  Since Crisp is a switcher and about the same from both sides, you also want a reliever with a low platoon split himself, preferably a lefty if possible (since Crisp is a little bit better from the left side I think).  But that is getting really too complicated for a manager, even one who took statistics in college.

Baseball performance distributions are pretty much symmetrical and normal, AFAIK.  Run scoring obviously is not.  Estimates of true talent are definitely skewed (because there are many more below average players) and are probably part of a normal distribution, but definitely not normal themselves.

I wasn’t watching the game.  Was Jamie Walker available?  If not, Bradford may have been the “best” option. 

The thing about the 2007 Orioles is that EVERY option coming out of your bullpen is a bad option.  It is a disaster of epic proportion.  And now Guthrie and Bedard are done for the year.  At this point the season can’t end early enough for the O’s.

Bradford had relieved Walker.  Of all the relievers who’ve pitched for the Orioles, only 4 of them have ERA’s under 5 - Walker and Bradford, then Ray who is on the DL, and Williamson who’s no longer on the team.

More than half of them have ERA’s over 7.  I know that small sample size ERA is not indicative of true talent level, but in this case, this is just a horrible bullpen.  These are not major league caliber pitchers.  The only ones who are good enough to be part of a decent team’s bullpen are Bradford (in a very specialized role) and Walker as a loogy.

Decent relievers are supposed to be easy to find, in theory, but what the Orioles need is a new front office that can actually find such pitchers.

From Trembley’s Wikipedia entry.

Trembley has a bachelor’s degree in physical education and a master’s degree in education, both from the State University of New York at Brockport. He also did graduate work in sports psychology at Penn State.

Not sure when he would have attended Penn State, but speaking as someone who recently took a graduate level stats class in a psychology department (and taught an undergrad level one), Trembley’s answer has all the markings of “I took the class 20 years ago and remember a few terms, but I probably couldn’t run a t-test now if my life depended on it.” The thing is, given that he does have a decent educational background, if someone gave him a quick referesher course, he might just be able to pick up on some of the Sabermetric issues out there.

"In any case, the situation screamed for one of your better relievers, if not your ace, a high leverage situation with 2 outs in the 8th.”

The problem is that with Walker already having been used, Bradford is the ace, horrid platoon splits or not.  Bradford is replacement level against lefties, but Rob Bell and Danys Baez have pretty much been replacement level period.  Not to mention that bullpen has had to log so many innings that he didn’t have a full compliment of options.

As Rally mentions, individual bullpen pitcher ERA’s for the season are next to worthless (well, not quite, but they definitely suffer from “small sampleitis").  Managers and GM’s live by bullpen pitchers’ season-to-date ERA’s (they lived by season-to-date everythings, as do most fans, commentators, and journalists), which is one of the major ubiquitous errors in bullpen and roster managment.  That is why you see a revolving door between minors and majors for your back of the bullpen, which is ridiculous.

It is true that the Orioles do not have a very good bullpen. Here are my current projections, in normalized, park neutral, component ERA, for their entire bullpen:  Keep in mind that the average bullpen pitcher is by definition, around 3.90.  The average closer is around 3.30 and the average “second best ace” (set-up man) is around 3.50.  I will put an asterisk next to the pitchers who have a better projection versus a LHB than Bradford.  Also keep in mind that you cannot compare a short to a long reliever/swing man.  The short reliever will have a lower ERA (and ERC or whatever) than a long one, even for the same talent, much like a starter who turns into a short reliever will improve his ERA by almost a run.  Also, Bradford’s projection versus RHB is 5.04, which is amazing, given that his overall projection is in the low 3’s (his projection versus RHB is 2.70).

Baez 3.82 *
Bradford 3.21
Bell 4.68 *
Birkins 4.47 *
Burres 4.85 *
F. Cabrera 4.01 *
Cherry 3.97 *
Hoey 3.75 *
Santos 4.39 *
Walker 3.49 *
Zambrano 4.60 *

As you can see, all the pitchers in the pen have a better projection than Bradford versus a LHB, although Bell is close.  Any LH pitcher is better, no matter how bad they are, although in the case in question, the batter was a switch hitter, so, as I said, it would have to be a LHP with a small platoon split himself.

It is probably not that big a deal, given that the Orioles truly have a terrible pen (although Cherry, Hoey, Cabrera, and Walker are decent, with Cherry and Hoey’s projections based mostly on minor league MLE’s), but as I also said, you NEVER want to willingly pitch Bradford against a LHB in a situation other than a blowout (low leverage) or perhaps 2 outs and no one on base in a not so close game.

Not that big a deal I guess, but I thought I would link that situation with the interview, which I definitely found amusing (the part about statistical analysis).

To me, asking a manager about the use of statistical analysis and getting an answer like the one from Trembley would be EXACTLY like if I were on a nuclear submarine and someone asked me if I could run the ship, and I said, “Sure, no problem, I took a little physics and engineering in college.”

Why can’t managers just admit that they know NOTHING (credible) about statistical analysis and in-game strategy, but that they would love to learn?

During the Cubs game on radio today, Ron Santo said a couple of really stupid and amusing things.  Of course, anyone who listens to him knows that he is a complete idiot (with all due respect to him, a great player, and apparently a very nice guy and a pleasant announcer, IMO).  The Cubs were leading 6-1 in the bottom of the 6th inning.  They scored a run and Santo said, “That was big.  Really huge.” After he had been going on (and his colleague was agreeing with him) about how because of the rivalry the Cubs and Cardinals have, if one of the teams is losing by alot, they always seem to catch up and make it a game.

Of course, when you are up by 5 runs in the 6th inning, another run is not big or huge.  It is small. Very small.

But that was not the funniest (stupidist) comment.  With the Cubs up 7-1 in the next inning, 1 out and a runner on, the count went to 3-1 to Taguchi (one of the worst hitters in the lineup of course) after Lilly through an up and in fastball.  Anyway, Lilly threw another fastball and Taguchi hit a rare home run.  Santo immediately said, I don’t understand that pitch.  after that high and tight fastball, you have to throw him a curveball.  He then went on for like 5 minutes saying, I can’t believe he threw him a fastball.

I wish I were there in the booth.  When he said, “How can you throw him a fasball?” I would have said, “Uh, Ron, because the count was 3-1, the Cubs were up 6 runs, it was the 7th (or 6th, I can’t remember) and Taguchi has no power.”

Not to mention the fact that in ANY situation there is no reason to ALWAYS throw a curve ball after a fastball up and in.

Anyway, just thought I would throw that in. 

Bertrand Russell said, “The trouble with the world is that the stupid are cocksure and the intelligent are full of doubt.”

To me, that sums up baseball managment in a nutshell!

There is not a more worthless and phony “business” than sports commentating, at least baseball (I don’t listen to enough of the other sports or know enough about them to comment on them).  If you know anything about sabermetrics or bona fide sports analysis, listening to Kennedy and Dibble on the radio every day, or BT on TV, is a complete waste of time.  And for the average fan, whether they find it entertaining and informative or not, they are being deluged with 99% worthless or wrong information.  It is no wonder that we live in an ignorant society and we have so much trouble solving important things in our country and on the planet.  Whereas almost everyone knows that proffesional wrestling is not “real,” almost everyone thinks that sports commentators actually disseminate useful and correct information, and they consider guys like Dibble, Kennedy, and the like, to be the “experts.” If I ever showed up on a baseball show on TV, I would be laughed off the air, and if I were ever a partner with an ex-player commentator during a baseball game, I would either be fired after one game or beat up and thrown out of the booth DURING the first game (after the first time I said to Morgan or Santo or Sutcliffe, “Have you ever actually played this game before?").

Great stuff, this thread.

As someone who has taken several graduate level statistics courses, I humbly submit, I know nothing of stats…

Thanks for the numbers, MGL. I guess I have to amend my previous statement that leaving Bradford in was reasonable, but likely wrong. After looking at Bradford’s platoon splits, it was completely wrong to bring him in. And yes I think everyone who has the inclination to read and opine at a blog like this one has had moments where they want to throw a brick through their TV screens or radios. One of my biggest pet peeves is when a guy is recently up from the minors and 1) the announcers have no clue who he is, frequently mispronouncing names 2) they make no mention of his minor league performance. And on the rare time when they do make mention of his minor league stats, it’s all AVG, RBI and W-L.

Bochy, whom I have criticized before - in fact, devoted a whole thread to - makes another egregious blunder tonight that ultimately costs him the game.

He has his closer in the 9th in a 1 run game and the D-Backs end up with runners on 1st and 3rd with 2 outs.  They put in Jeff Salazar, a lefty, to pinch hit.  Kline, the lefty reliever, is ready in the pen.  But, Hennessey is their closer, so God forbid you should take him out BEFORE he blows a lead.  Keep in mind that Hennesy is one of the worst closers in baseball.  He doesn’t have good stuff so there isn’t even an illusion that he is a good closer. He is one of those pitchers who happens to have a good ERA for the 60 IP he has thrown so far this year (but with his usual poor K rate).  His career ERA is 4.27 with a career OPS against of .763 (granted, some of that is as a starter). He is NOT closer material and that will eventually catch up to him.  He is not a bad reliever, just not closer material.  It is a no-brainer to bring in the lefty Kline.  Now, I am not sure that the D-Backs would not have countered with a RH pinch hitter, but I don’t think they would have.  In any case, Kline is NOT horrendous against RHB.  But, then again, the Giants are going nowhere, so who cares, right?

I care.  I just finished watching the game and am a marginal Giants fan.  Or at least I thought I was.  Oh well, at least it’s always fun to see what Bonds will do.

I was curious to see the WE difference between pitching Hennesy and Kline, according to my projections for them of course, using my sim.

With Hennesy still pitching to Salazar, SF wins 84.30% of the time (in 250,000 simmed games).

With Kline coming in, it is 85.41, a difference of 1.11%, about what I would have expected.

An egregious error like that generally costs on the order of 1%.  But, these things add up. I always have to remind pissed off fans (not that most fans would realize this error) that even the worst of managerial decisions does not “cost the team the game.” In fact, it costs so little that you would never even notice it.

While I was at it, I wanted to see if taking Bonds out (for defense) in a close game was correct.  I have never been sure.  My sim does a good job of using each player’s UZR projection of which Bonds’ is not very good of course.

With Raj Davis in for Bonds in the 9th in a 1 run game and the Giants starting with the #8 batter if they need to bat in the bottom of the 9th, they win the game 83.81% of the time.  With Bonds still in the game, they win 82.91%, so taking him out seems to be the correct play, although it is close.

What if Bonds were to definitely bat in the 9th inning (say, he bats 3rd in the 9th inning) if the other team ties the game or goes ahead?

With him in the game, SF wins 82.89% of the time.  With him out of the game (and Davis hitting 4th, in his spot), SF wins 83.57.  So it is still correct to remove him.

Since recently we have a lot of discussion about managers and their in-game strategies. I think this article by HBT’s Chris Jaffe is relevant and deserve some notice. (Is Bob Melvin a good manager?)

http://www.hardballtimes.com/main/article/no-mirage-in-arizona/

I read the article a while back and it uses incorrect logic and is a really stupid article.  If a team has a bad record in blowouts, OF COURSE IT WILL LOOK LIKE THEY ARE USING BAD RELIEVERS IN THOSE BLOWOUTS, since those relievers obvious did poorly in those games!

And if a team has a good record in blowouts (i.e., they are the ones doing the blowing out), it will look like they are using good relievers in those blowouts, since by definition the relievers must have done well in those blowout games.

The whole concept of why teams under or overperform their pythag records does not interest me at all.  Looking at a pythag record is only (marginally) interesting if you don’t know anything about the players.  If I want to know how good a team is likely to be going foreward, I simply look at each player’s projection (pitching, defense, offense, baserunning) and add them all up.  That tells me everything I want to know and I could not care less about that team’s pythag or actual record.  If they have a good closer, then that will show up in his projection and the fact that you double his innings value (because of his average LI) is a proxy for “winning close games.” If they don’t have a good closer, like the aforementioned Hennessey, that will also show up as if you expect them NOT to do so well in close games.

Now if we go back and look at each player’s projection before the season started, adjusted that for injuries, and then prorate those projections for actual playing time, and then add all the numbers up, we can likewise come up with a projected win total for the team to-date.  If a team under or overperforms THAT, that is mildly interesting and can be used as a starting point for whether a manager has done a good job or not.

Using a team’s w/l record, independent of anything else is a very poor (read: worthless) starting point for managerial evaluation.  Using a team’s w/l record compared to conventional expectations for a team is also a poor starting point.  For example, I heard today on the radio that Freddy Gonzalez must have done a wonderful job because everyone expected the Nats to lose 120 games.  What idiot projected them to lose 120 games?  Same thing with the WHite Sox for example.  People might criticize Guillen because the White Sox underperformed so much this year, given that the conventional win expectation for them was in the mid to upper 80’s.  Well that was a lousy (wrong) expectation.  All of the analysts had them in the mid 70’s.  And of course evaluating a manager based on pre-season projections and actual w/l is ridiculous without adjusting for things the manager had no control over, like injuries and personnel moves.  Finally, even if you do the most rigorous thing, which is, as I juist explained, take the pre-season projections (by competent forecasters), pro-rate by actual playing time, adjust for known injuries, and then compare that expected w/l to actual w/l, there are still lots of things out of control of the manager, not the least of which is good old-fashioned random fluctuation.

IOW, without literally going through each game with a fine toothed comb (and that would just be for in-game strategies, not teaching, motivation, etc.), there is simply absolutely no way of evaluating a manager using any of the things that either people traditionally use (most of those are just plain stupid) or even the things that I have seen and read analysts use, like the article by Jaffee.  If there is one thing in baseball that is next to impossible to analyze and evaluate, especially in teh short-term, it is the impact of a manager, good or bad.  The most ridiculous award in baseball is the MOY award, froma sabermetric perspective (by that I don’t just mean from an in-game strategy perspective - I mean from a separating fact from fiction perspective).  It makes the GG awards look like the work of Bill James, Tango Toger, and Tom Tippet combined (names are for illustrative purposes only).  There are a few things that really, really irk me (many things just irk me) when listening to people talk about baseball.  One of them is listening to people try and evaluate managers.  It just can’t be done, at least it hasn’t yet, and as I said, the traditional “methods,” if you want to even call them that, are really stupid.

I didn’t watch that game but I can assure you Arizona would have pinch hit for Salazar if Kline had been brought in.  Unless they are complete idiots.  The only reason not to if a lefty was brought in would be that you just don’t have many people on the bench. 

In these days of 12 man pitching staffs, if you’ve already used a few bench guys (I see AZ had used Cirillo, Callaspo, and Clark) you might be stuck with your first choice of pinch hitter.  But this is September, and roster restraints no longer apply.  They should have plenty of RH options to choose from among the callups.

Whether Hennessy should be closing instead of Kline is another issue.

Well, the brick just went through my TV again. Resident buffoon Jerry Remy was just going on and on about how great the Red Sox lineup was because it was balanced now. About how they now have guys who can bunt (always never a beneficial play, occasionally a neutral one. But when it is, it is because the bunter is such an unproductive player otherwise.), run, hit and run and so forth. He even seemed to imply, though he never said this directly, that this lineup was better than previous incarnations. For a guy who watched the Red Sox put out some amazing offenses in 2003 and 2004 (averaging 955 runs per year) using remarkably “unbalanced” lineups, this is profoundly stupid.
So many people paid enormous salaries to cover and work for baseball still have no clue how runs are created when the framework to answer that question has been around for decades. Little do Red Sox fans and Remy know that a lineup full of Wily Mo Penas (about a .280 EqA guy) would be definitively better than the current Sox lineup and most of baseball’s. A team full of Adam Dunns (or Ichiros for that matter) would be historically great. I’d almost rather listen to Remy shamelessly shill his website or do dreadful comedy with Don Orsillo during crucial moments of a game than analyze. Almost.

That should read “almost never” not “always never” in the above post.

"Having a balanced lineup,” like “being consistent” is just one of those announcerisms and cliches that “sounds like” it should be good but doesn’t really mean anything.  Of course, it is true that to optimize power you want guys on base and vice versa, so there is SOME merit to the concept of “balancing” your lineup (as an example, if everyone in your lineup only hit home runs or made outs, a home run would only be worth around 1 run rather than the 1.4 it is typically worth.  On the other hand, the “balance” of speed, power, small ball large ball, etc., is a mostly worthless concept.

Try explaining your position, Phil, to a group of broadcasters…

I’d bet Trembly is referring to the fact that matchup stats are usually based on very small sample sizes. The small sample’s create a large standard error where the ‘true’ matchup rate resides. The “plus and minus standard deviation” line is just his way of stumbling around that thought. Maybe he is one of the brighter managers out there.