THE BOOK--Playing The Percentages In Baseball

I think “consistency” means that you don’t act as a random distribution. That is, if you have a 10% K rate, then after getting 1 K, you will have 9 PAs w/o a K. At least that’s the idea I get from how it’s generally used when referring to baseball players.

Of course such a skill is impossible…

Andy:

Why is it impossible?  Just because we treat plate appearances as independent events doesn’t mean they are.  Why couldn’t a player’s true talent involve more than just a mean probability of a hit?  Couldn’t it involve correlations (or anti-correlations) from appearance to appearance?  I haven’t seen any studies that show plate appearances as independent events.  Probably you couldn’t prove that, since the starting pitcher does induce correlations on short scales (1-3 plate appearances).

Studies in basketball have shown independence, the probability of making a shot after a make is the same as after a miss.  But, I’ve never seen the same sort of study for baseball (not that I’ve looked).

Somewhere Joe Morgan is weeping. Consistently, of course.

Why exactly would consistency be a positive thing anyway? Why is it better to play “consistently” than to have peaks and valleys, assuming you come out with the same numbers overall?

"Consistency” is just something that someone wants to say to explain something that he happens to see.

It isn’t necessarily a positive thing of course, yet it is treated as such by the media and by the fans.  In general, if you are above average, you benefit from being consistent and if you are below average you benefit from being inconsistent. But that is another story.

LarryinLA, that is the point I am making. That is virtually impossible to be any more “consistent” than having a consistent true talent mean for any particular component of performance (K, BB, HR, etc.).  That is the best anyone can do (and we know that is impossible itself).  And given that, your actual performance will vary around that mean, looking basically like a binomial distribution.  That is a given.  There is nothing a player can do about that.  Plate appearances are almost certainly NOT independent events, but that has little to do with whether or not the performance is consistent or not.

True, consistency is treated as a positive - ironically, often by the same people who speak admiringly of a player’s ability to “get hot.” Aren’t both ideas just the result of people looking for order in randomness?

Usually explaining this with a coin flip works well. Nothing is more consistent than a coin, yet it sometimes lands on heads 70 times out of 100 (or 10 in a row, or something like that) just by random chance.

This guy is a computer simulation, and I can guarantee you his true talent changes only marginally from year to year, but his performance sure swings wildly:

http://www.baseballprojection.com/apba/encyc/player_encarjo01.html

Yup, some people would be shocked if they saw a typical or even a slightly atypical seasonal or career performance of a simulation.  And a simulation, assuming that it has a fixed talent level per PA, which they usually do, at least within a season, is as consistent as you can possibly get.  That is what I was trying to say and still have not articulated very well.  Basically that the ceiling for consistency which no player can surpass or even approach (much like the speed of light) is a computer simulation - and a computer simulation is going to typically have lots of streaks and not appear particularly consistent to a casual observer as the above link illustrates.

As many of you know, consistency is a word that is bantied about by the media in really silly ways. For example, how many times have you heard exactly this:

“He is inconsistent, but that is because he is still young.  As he gets experience in the major leagues, he’ll learn to be more consistent.”

First of all, I am not sure that we know that young players are more inconsistent than older ones.  Since we know their true talent less than we do of the older player, their performance may appear to be more inconsistent across players, but individually, I don’t know if I have EVER seen a study on the consistency (basically, the variance of performance) of young versus older players.  But even if that were true, as we have discussed a little already, do we want players to “get more consistent” as they mature?  Clearly the media implies that to be the case.

Basically, when the media talk about a player being consistent or inconsistent, they mean “bad” and “good.” There is no reason to talk about whether a player is consistent or not - since even to them (the media), it should mean nothing.  The other thing is that bad players often appear more inconsistent that good ones.  For example, suppose you had a pitcher who has a true talent level of 3 runs a game with a SD of 1.5 runs.  He’ll give up between 0-6 runs a game 95% of the time.  And that distribution is not symmetrical because he can’t give up less than 0 runs.  That will appear to be fairly consistent.  Say a pitcher is a true 5 runs a game with the same SD. He’ll give up between 2 and 8 runs a game 95% of the time with an occasional (less than 2.5% of the time) 0 or 1. He will appear to be more inconsistent I think.  Basically any pitcher, at least, who is bad will appear to be more inconsistent, I think, because he will give up a lot of run more often than a good pitcher, and he will occasionally pitch really well, as all pitchers will do.  That will look inconsistent.  The good or great pitcher will not very often give up a lot of runs and when they do, it will be excused.  And they will have a lot of good games of course.  They will appear to be more consistent. So in the media’s and fan’s eyes and minds, bad = inconsistent and good = consistent.

If anything, defense should be LESS important in the post-season, as there will be pitchers who tend to K more batters and perhaps even batters who tend to hit more HR and K more (I don’t know about that). Either way, the idea that defense is more important than other way of adding or subtracting runs, during the post-season as opposed to the regular season, is nonsense, as Tango says.  I have written about this before.  This is an example of how regression analysis can really get you in trouble.  The results of a regression analysis better match what we know about reality. If it doesn’t, it is likely that we are making a Type I error in the analysis regardless of the p values…

I generally agree with the point, but there are a few very subtle points that I believe cause the argument to break down in extreme cases.  Independent of a purported “consistency skill,” players of different abilities will have demonstrate different levels of “consistency.” What I mean is that the standard deviation of the result of a plate appearance (or string of plate appearances) will be different for players of different skill levels.

Consider two robotic players with unchanging true talent levels.  Their results give us perfect binomial samples.  If RoboPlayer 1 has a true talent batting average of .250 and RoboPlayer 2 has a true average of .350, then the standard deviation of RoboPlayer 1’s ABs will be about .433 while 2’s will be about .477.  Maybe this all seems like semantics, but 2 could be said to be 10% more “streaky” and thus less “consistent.”

Even two robotic players with identical true talents can have different standard deviations in their results.  Consider two roboplayers that each have a .340 wOBA.  They get to that .340 wOBA in very different ways.  Robot A does it purely by singles and his batting average is thus about .378.  Robot B can only hit home runs or get out, so B gets his .340 wOBA by homering in about 17.4% of his at-bats.  So both robots have the same true talent and that true talent never changes.  Even so, because their talent is composed differently, their performance level shows different levels of “consistency.” In a single at-bat, A will have a standard deviation of about .436 while B will have a standard deviation of about .740 (using run values of .9 for the single, 1.95 for the HR, and 0 for the outs).  So in this hyper-extreme example, B will be about 70% more “streaky” and thus much less “consistent,” even though he has identical true talent to A.

Perhaps this comes off as just a bunch of semantics though; I don’t know.  If one player’s performance tends to stay in a tighter band around his mean performance, I think it would be fair to call him more consistent even if it’s not the result of some consistency skill.  Of course this is obviously not what the mainstream media generally refer to, and it in fact leads to the seemingly paradoxical result that good hitters are less consistent than bad ones (and the reverse for pitchers).  I don’t really want to take away from the OP; I certainly agree with it except for the nuances.  I hate when members of the mainstream media throw around terms like consistency as they so often do.  This does make me wonder though if there are less extreme examples of players whose skills give them a much different standard deviation than their peers of equal talent.  It very well might be trivial and not worth discussing though...hard to say without doing some more research…

"What I mean is that the standard deviation of the result of a plate appearance (or string of plate appearances) will be different for players of different skill levels.”

Yes, of course.  That doesn’t cause my argument to break down, because my argument has nothing to do with that.  Of course I know that. In fact, I don’t really have an “argument.” My only point, which is difficult to articulate, is that the ceiling of any player’s consistency is merely the mathematical SD of his fixed true talent level and that is probably not what people would consider “very consistent.” Every player, because their true talent level changes from PA to PA, if for nothing else because of the competition and the environment, HAS to be less consistent than that.

Why is that interesting?  Because I think that the MSM and the public think that a player actually has MUCH more control over his consistency or lack thereof than is possible.  Again, it is possible that below that ceiling, players have some control over how much their true talent varied from PA to PA and from day to day or month to month.  But if that were the case, I don’t think that would be visible to the naked eye because the random, binomial variance around those true talent means would dwarf any variance in true talent vis a vis their true talent variance (if that makes any sense).  IOW, let’s say that the roboplayer had a true talent variance from PA to PA of zero. And that for the average player, it was 10 points in BA.  Now let’s say that the ability (skill) to vary from that 10 point variance was another 3 points in variance (IOW 95% of all players were between 4 and 16 points). I don’t think that would be “visible.”

BTW, in your example of roboplayer 1 and 2, you can’t give a SD without specifying in how many AB. You obviously mean in 1 AB or per AB.

For example, for roboplayer 1, his SD for 80 AB would be about 44 points of BA.  If you told the average fan that ALL players with a true .250 BA would have around a month every season where their BA was less than around .250 and that there was nothing they could possibly do about that, they would not believe you.  They would tell you that a veteran player or a professional hitter could somehow make sure that that didn’t happen.

I hope this thread does not evolve into the regression one, where people start chiming in with things like:

“I disagree, because dogs are actually smarter than cats...”

I believe what people who talk about consistency actually mean is for a player to learn not to dwell on past failures, trying to make each plate appearance an independent event. Dice don’t have memories, players do.

In my late teens I was in a couple bowling leagues for several years. Being a stat nerd even then, I copied my scores for awhile, keeping track of the results of each ball thrown, as in which pin combinations did I leave on the first ball, and what rate did I convert spares on each.

I then designed a simple dice game based on those probabilities (no computers yet). The simulation always had a higher average score than I did in real life, because sometimes I was still pissed over a bad throw, and it carried over to the next frame, causing another bad throw. The probability at n was somewhat dependant on the result of n-1.

A player who ‘learns to be consistent’ I believe is the one who can put the bad results of the past out of his head, and concentrate on what he did when he had good results.

My problem with consistency is always this: What is the sample size being used?  Say a player hits .250 (ignore the lack of BA importance and everything else, I’m just going for easy math here). Say the player is a true talent .250 hitter. I think “consistent” would be going 1 for 4 every single game, right? But clearly that will never happen. So what makes him consistent? Going 5 for every 20? 25 for every 100? I don’t think the people who talk about this truly realize what they’re saying.

Brian Cartwright/#14 summarized my feelings perfectly. It’s possible (not definitely, but certainly possible) that some players carry a bad plate appearance into the next, and into the one after that, and so on. This would cause them to underperform their true talent level for an extended period of time, and not simply because of the random fluctuations of chance. I certainly don’t think it’s out of the realm of possibility.

My real question is, how would you even test for “consistency” if you wanted to? I don’t have nearly the statistical know-how of MGL or Tango, but perhaps you could do it similarly to how they showed that clutch hitting does exist… Take samples of a player’s wOBA (perhaps weekly, maybe for every X number of PAs), measure the fluctuations, repeat for lots and lots of players, and if the distributions of fluctuations is different from what you would expect by chance, then consistency exists. Would that work?

"It’s possible (not definitely, but certainly possible) that some players carry a bad plate appearance into the next, and into the one after that, and so on.”

Now we’re talking about hot and cold streaks I think, which we have shown to basically be no more than random fluctuation, which suggests that that (having a bad PA carry over into the next one) is not significantly manifested in MLB.

Sure, you can test for a consistency skill exactly as you are suggesting Tom.  Compare the distribution of variances of performance for each player to what would be expected by chance.  As with anything else, you are limited in the certainty of your results by sample size and without an enormous sample, you are limited in the amount of signal that you can detect (sort of like James’ “fog” article).

"My only point, which is difficult to articulate, is that the ceiling of any player’s consistency is merely the mathematical SD of his fixed true talent level and that is probably not what people would consider “very consistent.” Every player, because their true talent level changes from PA to PA, if for nothing else because of the competition and the environment, HAS to be less consistent than that.”

I fully agree, but it was more interesting for me to discuss the subtle related point about how dogs are smarter than cats.

I think the main issue in getting the MSM or casual fan to agree is not in getting them to believe that plate appearances are independent but rather that they are even random.  I don’t know how to convince them that plate appearances can be treated as though they were random.  Frankly most people don’t understand randomness anyway.

I have often heard the term that someone is consistent used (In sports but also in other fields such as in the workplace) to describe someone who can be “counted on” to deliver at least a (usually relativly arbitrary) minimum acceptable performance.

If the word is used in this way, obviously better players will be considered more consistent.

Also here I think it is important to mention that what we are discussing here is consistency in performance.

But it is also possible for players to have more or less consistent mechanics. I would suspect (though I have no proof) that players that have mechanicly consistent swings (i.e. that look the same from pitch to pitch and PA to PA) are better hitters than those who don’t.

There can also be consistency on a process level. Always throwing the same number of warm-up pitches, taking practice swings the same number of times etc etc. Some athletes (not just in baseball) takes this type of consitency to an almost ritual level. (I.e. always put on one sock first, always step onto the turf with a certain foot etc etc.)

I think the phallacy here to some extent comes by mixing these different levels of consistency up.

Every player, because their true talent level changes from PA to PA, if for nothing else because of the competition and the environment, HAS to be less consistent than that.

MGL - Either you misspoke in this statement or your definition of true talent differs from mine.  A player’s true talent should reflect some skill within him and be indifferent to factors outside of his control like the competition and environment.  It should be his performance that varies with those factors.  Of course “true talent” doesn’t really exist, it is a construct that we invent to be able to better project a player’s future performance, so I guess you can define it in any way that is helpful to that end.

"I think the phallacy here”

Anybody else laugh at the misspelling here?  So apropos.

***

“True talent”

Peter, I think you and MGL are both right.  If, for example, you are playing at Coors, Juan Pierre’s “true talent level” might decrease relative to everyone else, because everyone else can take advantage of something Pierre cannot.

It’s a nuanced point that is really not in play most of the time to any great degree, and so, from Peter’s standpoint, the true talent level does not change by environment.

> hot and cold streaks ... which we have shown ...

MGL, so could please once again formulate (briefly, in about three sentences or so) what exactly you have shown with that analysis, and what real like inference it made.

For instance, did you claim that ‘hot hand’ in baseball didn’t exist?

Something I’ve learned to do here: whenever you read “___ does not exist”, before you protest, think about it a minute and decide if the person saying that might mean “the “footprint” of ___ is not apparent in any statistically significant way, and/or is indistinguishable from statistical noise, in which case it may as well not exist”.

I’ve found this helps bridge the gap between my way of looking at things and others’ way in a few key instances…

Sergei, it’s all in The Book.  Click on the Amazon link at the top left, and once there click “Look Inside” and read it for free online.

I think some people see some consistency in some sports and then look to baseball for the some results. 

I know in Fantasy Football, some receivers are between 80-120 yards and average 0-1 TD a game.  You know how the receiver will generally do.  While others go weeks with out going over 100 yards and then catch for 150 and 3 TDs. 

On the baseball front, I think pitchers and hitters need to be split up.  With hitters it is impossible to measure consistency on a game by game basis and they could be measured in maybe 10 game streaks.  I totally know athletes that let factors effect there play, especially how the team is performing.  They won’t do as well when losing, but will try much harder if they are on a team that is winning.  I will think if I can see if I can find hitters who give up while team losing.  Alot of variables to look at here so not sure if it can be done.

Pitchers.  I actually have done some work on starting pitcher consistency and believe that there is some possibilities here. I always figured this could useful in that pitcher will give certain level of performance and knowing the team’s offensive ability, predict the number of games the team should win when that pitcher starts.  A single measure, like Standard Deviation, of game score (or game FIPS) for each start measure’s little.  I have made some progress by putting Game Scores into bins of 90% winning and losing chance and the area where the game is up for grabs depending on the teams offensive ability.

> Sergei, it’s all in The Book ...

Thank you. But I’ve read about it a zillion times already on this blog.

What I had in mind, frankly, was kind of lashing out at it. But to do so I needed to have the exact thing briefly formulated and right in front of me, say in this thread.

Actually, it’s been quite some time I was thinking about criticizing it. You, guys, just keep bringing it on and on as if it’s a decided thing.

How about that? :D

Sergei, if you want to rip their work, rip it and be prepared for them to respond (maybe in kind).  But it seems weird that you’re requesting a three sentence synopsis to rip.  Sort of like lining up someone in front of a firing squad, and then asking them to move a bit to the left so you don’t have to work so hard aiming…

I must be missing something Sergei.  Are you saying that you are unwilling to read, for free, the chapter in The Book in question?

Instead, you want us to summarize the findings?

And then, you will criticize the summary because you will… I dunno, point out something we didn’t consider in the summary?

And then, we will tell you that we did consider it in the chapter, and all you have to do is read it?

And then, we’ll be right where we wanted you to be: read the darn chapter first because being critical?

"because being critical? “

=

“before being critical?

> Instead, you want us to summarize the findings?

Of course, I want YOU to summarize the findings.

What do you want, ME summarize the findings?

And then (using your words), you will criticize the summary because you will… I dunno, point out something I didn’t consider in the summary?

Sergei: you are seriously asking me to do work?  Not only did we write the book, but now you want us to write the cliff’s notes version of it too?  Seriously?

Am I asking for too much that before you criticize something that you read, by yourself the thing you are intent on criticizing?

Tango,

First, you say I go read the book for free online.

Now you say I’d better buy it. What next?

I can’t understand what the fuss is all about.

I just wanted to know the up-to-date take on the issue at hand.

Haven’t you written that already and why can’t you just copy and paste it here?

I never said for you to buy it.  I presume you read “by yourself” and mistranslated it to “buy yourself”.

Read the book, for free, on your own.  Make the comments as you see fit.  Don’t ask me for anything else.

Sergei:  YOU want to write about something, YOU do the necessary research to find out.  Pretty simple stuff.

OK, no problem then.

It’s just that the extra comma in “read, by yourself” confused me and I got the impression that you actually meant “bUy yourself” but omitted the U.

All right. But then I’m not gonna make any comments either. What, “you are seriously asking me to do work?” LOL

Match ends in a draw.

I never thought I’d say so on this blog, but I feel dumber for reading what I’ve read.  Should have stopped before the posts got into the mid 20’s.

You never should feel dumber for reading something, although I know what you mean.  As I always tell my son:

“Read and listen to everything you can.  Keep your ears and mind open and your mouth shut (as much as possible) unless you are asking questions, and even then, keep them to a minimum.  You will learn something from everyone and everything, even if you think you learned nothing.  In fact, if you expect to learn something, even from an idiot or a 6 year old child, you will. Even that which you “discard” or “reject” is a learning experience. And you will eventually know what to keep and discard.  But be careful.  What you think you should discard NOW may become useful or handy some day in the future. So put even your discards in a file somewhere in your brain.”

Or something like that…

So should I.

Revisiting the tread in the morning, I came up with a very short question. In two sentences.

MGL and Tangotiger,

An mlb manager says that he bases his strategic/tactical decisions on, among other things, player’s RECENT FORM.

Do you [still] think he is dumb because of that?

Just say DUMB or NOT DUMB.

Also, I want to apologize to everyone, and Rally in particular, for any harm that the conversation above might have caused.

And trying to atone for any losses Rally suffered, I’d like to provide him and also Greg, whose comment #23 I certainly appreciated, with a following link:

http://www.cricinfo.com/magazine/content/story/417934.html

Faster, stronger, uglier: Why has cricket deteriorated aesthetically even as it has improved as a spectacle?

A pure piece about the game and nothing but the game. All in beautiful English language.

You know, one can start to forget just how good the language can be if he reads too much of a certain blog.

Where too often one can find words like: idiots, whores, stupid, and, god forbid, her majesty the regression to the mean.

Sergei

Sergei, I looked at batters and pitchers (separately of course) performances throughout the year.  If they had a 2 week spate of performance that was above a certain threshold or below a certain threshold that performance went into a “hot” or “cold” bin.  I then looked at their performance over the next day, 3 days, and 7 days.  So basically, we have batters who were “hot” or “cold” and we are able to see if that has any predictive value over their performance in the near future (next day, 3 days, or 7 days), beyond what their last 3 years stats would tell us.

The results for batters were that unequivocably (as much as sample stats can lead to an unequivocable conclusion of course) there was no predictive value.  For pitchers there was some.

For example, let’s use BA.  If a batter has a BA of over .350 for two weeks straight and his career (3 years actually) BA is .278, he hits .278 1 day, for 3 days, and for 7 days “after” the 2 week streak.  If he hits .180 for 14 days, he still hits .278.  Nothing.  No effect.  No predictive value.  You can’t even say that a batter who is “cold” may be hurt and will continue to perform a little worse than he usually does.  Nothing.  Forget about saying that a batter is “seeing the ball well” or “not seeing the ball well.” Nada. Zip.  Nothing.

As with the clutch issue, we can’t say that NO batters don’t have periods of time where their mechanics or psyche is “clicking” or not.  But we CAN say with pretty good certainty that the prevalence and magnitude of that, if it exists to any non-de minimus degree, is small such that we cannot identify that kind of thing by looking at a batter’s stats for a week or two (or three, or whatever).

As I said, for pitchers, I found some predictive value, maybe on the order of 2 tenths of a run in ERA or so.  IOW, if a pitcher is very hot or cold for 2 weeks, his ERA over the next couple of appearances will be a tenth or two of a run higher or lower than would be expected from their last few years’ stats.

So yes, for batters, managers would be “stupid” to make any decisions, like lineup changes, IBB’s, benchings, etc., based on recent performance.  At least as far as we know according to the research…

"on recent performance”

As long as this implies the traditional performance lines (OBP, SLG, etc). 

It is possible, I suppose, that you can look at swing and contact rates.  i.e., a scouting-level look.

What I would *love* is for a manager to
a) look at the scouting-level data, see a guy who swings alot at outside pitches (more than he normally would), or swings and misses alot (more than he normally would), leading him to think that maybe something is wrong with the hitter
BUT
b) that guy to also be above his career OBP, SLG numbers while he’s doing all that

That is, I would love for the manager to be faced with a conundrum, and in the face of actual high performance, to sit the guy’s a$$ down, because his gut is telling him he can’t possibly keep it up based on his fundamentals.

I’d love that.

However, in no way will this ever happen.  Because when it comes down to it, a manager does not rely on his gut, but looks at short-term performance as if it means something.

Given the choice, I’d want the manager to do, in order:
1. look at long-term performance
2. look at his gut
3. look at short-term performance

However, I would bet that most managers go in the exact opposite order.

Rally, we’ll win you over yet

Having seen several managers get burned by moving “cold” or “hot” players in the lineup, just in time for their luck to change, it seems obvious that many managers react too soon. On the other hand, when does prolonged bad (or good) performance force a manager to make a change?

I was thinking about this last night while watching the Red Sox/Rays game. I couldn’t figure out whether I agreed with Maddon’s decision to drop Upton to 7th. I mean, obviously you can’t have a leadoff guy OBPing .319, but his career to now suggests he’s much better than that. I guess I’m just struggling with where to draw the line. Clearly, you shouldn’t be managing based on the last few weeks, but what about several months of a guy seriously underperforming projections? I guess this gets into defining “short-term performance” vs. “long-term performance”, to borrow from Tom’s last post.

Right, there is no “line”.

You simply keep reupdating your evaluation every day.  It’s small movements every day, even as a guy is going through 0-fer slumps.

Unless you can identify a fundamental change (or injury) to which that change is persistent (not transient), you have no choice at all.

That said, moving guys around in the batting order really has little to do with how many runs a team will score.  If the players thinks this is good, then that’s fine.

More important is that Upton remains somewhere, anywhere, in the lineup.

Tango #45:

What you said about identifying an injury is important, in that this point recognizes that there is a difference in how you ought to consider hot vs. cold. 

There is no known injury out there that can make a player’s performance dramatically improve, and skill development occurs gradually, not instantaneously.  Therefore hot streaks should universally be considered statistical artifacts.

Not so cold streaks.  There are plenty of injuries or temporary conditions that can make a player’s performance suffer.  Furthermore, players frequently conceal injuries, or perhaps try to convince themselves as well that they are fine, in order to keep playing.  So a cold streak could be a statistical artifact, or it could be a material change in a player’s capability (typically a temporary one). 

Part of what a manager is paid for is to distinguish the two, and since either one can produce the same (poor) stats, stats are not very helpful in making that distinction…

Greg, as long as he doesn’t quote that he was 1 for 16 when he makes his decision, I’m ok with that.  Similarly, I’d also like the manager to say that “even though he’s 10 for his last 18, he’s been battling the flu all week, so I have to sit him down”.

Did you see this?

http://www.baseballthinkfactory.org/files/newsstand/discussion/mlb_hinch_gives_reynolds_rare_day_off/

I did not, but my glove-slap to AJ Hinch!  Ballsy.

MGL, thank you very much for writing it out one more time (post #41).

In fact, it was completely in line with what I could remember from your previous posts on the matter (except I missed your similar study on pitchers).

I also particularly liked what Tango (#42) and Greg (#46) said.

When I was asking, although in a less than graceful way, for a short description, here’s what I wanted to see in particular:

1) [Forget about saying that a batter is “seeing the ball well” or “not seeing the ball well.”]

and this

2) [So yes, for batters, managers would be “stupid” to make any decisions ... based on recent performance.]

First off, I was talking about recent FORM, not PERFORMANCE.

And it is point 1) what I can’t agree with, even though MGL has already pointed out this generalization’s possible shortcoming: “ ... we can’t say that NO batters don’t have periods of time where their mechanics or psyche is “clicking” or not ...”

And then, “we cannot identify that kind of thing by looking at a batter’s stats for a week or two (or three, or whatever).”

That’s exactly my main point: we can not identify good or bad FORM by looking at the batter’s recent PERFORMANCE, as shown by [traditional] stats.

IMHO, the problem with unusual FORM could be not that it’s “magnitude is small”, but also that
the player’s FORM itself can hardly manifest itself in producing visible changes in PERFORMANCE.

IMHO, I’d say that in baseball, a player’s TALENT has a particularly hard time manifesting itself in RESULTS.

So what else should we expect for CHANGES in talent then - small or no effect at all.

(Not so in tennis, for example. The rules of the game, equipment are all different)

In other words, we can’t say if the batter is in the zone (“seeing the ball well”, mechanics or psyche is “clicking”) by looking at any traditional PERFORMANCE stats.

Will probably add something later.

- Sergei

Ditto exactly what Tango just said in 45 and 47.  Greg, when I did my streak studies, I expected to find some predictive value for cold streaks for the reason you state:  A cold streak suggests injury which suggests that a player will continue to play a little worse than his career stats, at least until he is better.  I found ZERO predictive value to cold streaks, which tells us that you CANNOT tell whether a player is injured or not by whether he has been cold or not.  Which tells us that a player is just about as likely to be injured and have a hot streak as he is to be injured and have a cold streak (assuming the same magnitude).

As Tango says, I’d love for a manager to bench a guy or move him down in the order because he in a little injured or tired yet has just gone 8 for his last 17!  Now that would be a manager I would let my daughter marry!  Of course that ain’t never going to happen as people would think he was crazy and the player would be pissed.

Matt, BJ is a very good hitter, so he probably should not bat 7th or 8th.  In general though, it does not matter much in what order you have your lineup and even if it does, a manager is not going to know the nuances of an optimal order anyway (no one does unless they do some analysis).  And Maddon is an idiot, in my opinion.

One of the things on my to do list:

Look at the performance of all players who are normally not good hitters but were “hot” for some period of time or the manager thought he matched up well against a certain pitcher, and moved him up in the order.  And look at good hitters who have been cold or who the manager does not think matches up well against the pitcher who were moved down in the order.

I will guarantee that the bad hitters who were moved up in the order performed badly (their career norms, adjusted for the pitcher and the platoon situation of course) and that the good hitters performed well even after being moved down in the order. If that is true, and I am just guessing that it is, I don’t EVER want to hear something like, “Yeah, but the manager knows the injuries and the matchups better than we do, and he knows who is swinging the bat well and who isn’t.” Ever!  If I am wrong, well, I’ll take back every bad thing I have ever said about all managers…

Sounds like we’re all in agreement: there’s too much random variation in (short-term) performance stats to be able to pick up real changes in true talent level, and so, a manager has to rely on his gut (not the stat sheets) to make the call.

> moving guys around in the batting order really has little to do with how many runs a team will score.

As I said in #50, not only player’s talent struggles to manifest itself in performance in baseball, but the very same is true for manager’s talent , I think. At least with pre- and ingame decisions.

To put in together - baseball, especially at pro level, is just a one long meaningless thing to watch and to participate in.

As someone brilliantly pointed elsewhere, the typical game recap should read: “Two teams played today. Each one made 27 outs. Everithing else was statistically insignificant.”

In fact, I consider baseball manager’s job (during the game, at least) as one of the most boring ways to spend time. You simply have very little control over things. All you really have to do is to just let your players play while pretending to be very busy.

BTW, I still can’t understand how one can watch the game looking from the side(from the dugout).

The two seasons that I worked with a team here in Russia, I watched exclusively from behind the backstop while filming games on camera, as usual. Not anymore, though.

- Sergei

“Two teams played today. Each one made 27 outs. Everything else was statistically insignificant.”

PLEASE, tell me where this was said.  It’s brilliant!

Tango,

Sorry, I don’t remember where exactly I saw it.

It was more than a year ago. In comments to a thread discussing casual fan’s perception of 9th inning hitting heroics.

Whether the one with a game-winning hit is a hero, forever labeled as clutch.

Was he really the game’s MVP.

Something like that.

It was not on this blog or on BTF.

Maybe ballhype (comments to Hardball Times), but I’m not sure.

But the text above is close to the original. Tried to google, no luck though.

I wrote something like that a while back:

http://otherfifteen.blogspot.com/2008/03/how-important-is-narrative.html

The world wouldn’t be a better place if newspaper articles all read “Today the Cubs and Brewers recorded 27 outs apiece in a contest at Wrigley Field, which revealed almost nothing about the two teams due to the small sample size involved.”

Great!

”...a manager has to rely on his gut (not the stat sheets) to make the call.”

I am guessing that that “gut” is worth nothing.  If anyone wants to take the other side and posit that we would find significant differences in performance when players are moved up or down in the lineup, as always, I’m up for a bet!

“The world wouldn’t be a better place if newspaper articles all read “Today the Cubs and Brewers recorded 27 outs apiece in a contest at Wrigley Field, which revealed almost nothing about the two teams due to the small sample size involved.”

I don’t necessarily agree with that (that the world would not be a better place).  Not that those are the only 2 alternatives of course (the way it is now, and that description above).  If I read that in the newspaper, I could discuss that with my children, if I were a teacher I could explain what that meant to my class, etc.  Thankfully there are few things in the world that are either/or.  There is plenty of room for all kinds of perspectives and viewpoints.  What makes the world a better place, IMO, is that we have ESPN’s BT, we have Poz on the web, and we have MGL, Tango, and all the other boring saberists…

I was curious as to how much consistency robot players would show. (Well, they aren’t all robots, only the Portland Decepticons)

So for the league I linked to in #8 above, I took all players who had at least 200 at bats through about the mid point of 2008 (the closest backup I had saved was 88 games) and also had at least 160 at bats in the second half.  These players have not had any change in true talent level during the season, being computer simulations.

I took the WOBA for each player in the first half, and for the second half.  The r is .442, and the RMSE is 0.042.

It would be interesting if somebody compared that to MLB players in first and second halves, I’ll do it eventually unless somebody beats me to it.

OK, I did it.  MLB players in 2008, comparing 1st half woba to 2nd half:

r = .388
rmse = .044

So they are a little bit more variable than computers.

Rally, the computer’s batters are facing different pitchers in different parks, with different weather, etc.?  IOW, in real life because the first and second halves don’t include the same opponents and weather, we would expect to see more variability even if players’ true talents did not change.

# 55, 56

Oh yeah, that was the article. Thanks.

“Recorded 27 out a piece”, the phrase I’ve been recalling many times later.

Funny thing, that it was Tango who sent the link to the story to Hardball Times, of course.

http://www.hardballtimes.com/main/article/how-important-is-the-narrative/

But he somehow missed or didn’t remember that particular sentence. But now, thanks to this forum, the article has made a full circle back to Tango. :D

~~~

“recorded 27 outs apiece in a contest ... which revealed almost nothing about the two teams ...”

Yes, and that’s the real problem with baseball, I think. As opposed to soccer.

And, IMHO, cricket, too, though it might sound strange at first.

You know, both sports are bat and ball games, and pretty similar in many ways. So where’s the difference?

To reiterate what was said in “What if… TWO strikes and yer out?” thread

http://www.insidethebook.com/ee/index.php/site/comments/what_if_two_strikes_and_yer_out/

I can see too major reasons:

1) the rules of the game: three outs - change, set batting order, very few at-bats for a batter, etc.

BTW, why a batter is not allowed to keep hitting until he is retired (this will call for tricky pinch-running rules, of course)

Imagine, Albert Pujols or Barry hitting for about 15 minutes uninterrupted and getting, say 5, walks and several home runs in a row in the.

Now that would be what people would remember forever! A real triumph for the batter and a real embarrassment for the pitcher.

Not a silly single accidental event like Fukudome’s homer mentioned in Colyn’s piece.

And no more silly quotes in wire reports like: “He pitched very well, but made one mistake ... if only we could take that one hanging slider back”

Taking a breather ... and oh, deadline to download yesterday’s japanese data nearing ...

Couple corrections. Should read

Imagine, Albert Pujols or Barry hitting for about 15 minutes uninterrupted and getting, say, 5 walks and several home runs in a row in the PROCESS.

(btw, that will do away with IBB problem :D)

Not a silly single accidental event like Fukudome’s homer mentioned in Colin’s piece.

No need for the word ‘silly’. Accidental is enough.
Meaning, accidental to a certain, HIGH degree.

Would Fukudome insist that he intended to hit nothing but a home run? No.

All what he could do was making a good swing.

Or maybe a good full swing (full swing - japanese lingo, to differentiate from contact swing).  Though many players and coaches preach you always swing the same way, of course.

OK, making a good swing.

That’s exactly what Hideki Matsui is always quoted as saying after hitting a homer: just a good swing, just an easy pitch to hit in a comfortable location. Just was able to see that ball well.

He sounds more like as if the homer was not his feat at all, but rather a happy occurrence.

Hence the very subdued celebration when rounding the bases, If only to acknowledge the fans and teammates.

As if he says: don’t really expect me to do it next time, even if it’s another easy pitch in good location.

I just might not be able to read it that well again or will swing just a tad earlier/later, contact the ball millimeters higher/lower, at an angle just a bit less steeper. And the result would be anything from a screaming liner to a harmless grounder back to the pitcher, from a hard foul to a downright miss.

That’s baseball after all, not tennis. We are hitting with a bat, not a tennis racket.

And here we come to the second ‘major’ reason ...

As a fan of pretty much every (honest) sport/competition I can lay my eyes upon,I must tentatively agree with Sergei’s latest few posts.  There is a certain, shall we say, “satisfaction”, achieved upon witnessing professional athletes perform at the height of their abilities, in high leverage situations, in which their performance strongly correlates with their abilities/talent.  In truth, (except for the fact that I am anything but a wordsmith) this is the most rewarding thing i have grown to expect from the mighty spectacle of “sports”.

I may be in the minority, but i feel that for all my love of baseball, in the above regard it largely fails as a sport.  It rarely delivers such satisfying moments.  And thus over the last several years I have found myself drawn more to other sports such as football and soccer, where leverage is almost always high, and talent is usually correlated strongly with results (though not always).

... The IBB was said to be an equalizer. And in baseball we have many more equalizers: three outs - bases cleared, number 9th hitter having to hit with 2outs and bases loaded in the 9th, etc.

But, IMHO, the biggest equalizer is, of course the bat and the ball.

So another major reason (of the difference between cricket and baseball):

2) The thin round bat, the small wildly and unpredictably moving ball and the fact that pitcher is 60’ 6” away.

And why it’s 60’ 6”?

It was just 50 feet at first, but someone decided to move it back to 60’ 6” about 120 years ago.

And why 60’ 6” and not 70 feet.

What, whoever that was could envision how baseball would evolve a century later?

With all the specialized training, wicked breaking pitches, fastballs with late action. That the athletes would simply become bigger and stronger.

And it was not like the situation in soccer. You know, modern goalkeepers cover bigger area of the goal then their 19th century predecessors. Just a bit bigger.

But then soccer keepers do not necessarily act at the extreme human reaction limits too often.

It is high-level baseball players who always perform at the edge of their reactionary capabilities.

And then on top of that we have the ever growing pace of the game, cut fastballs, fork balls, etc.

Of course, baseball has already tried to do something with it. Several times. The lowering of the mound, smaller strike zone (10-20 years ago).

But you can hardly make the mound any lower. Somewhere along the line you migth have to do exactly what women’s fast-pitch softball has done several years ago.

They moved the pitcher’s plate one full meter back!

Originally softball was so poorly designed that, for example, the best US hurlers had as much as 50 career no-hitters or perfect games each, or so. If I recall correctly.

So badly it was designed, that even at lower levels, the only way for the hitter to succeed was to slash
down through the ball (shortest possible bat path to the ball).

Thanks god, baseball’s 1890 redesign proved to be much better. But is it perfect? Does it let the player’s talent to manifest itself in results?

In lower levels- possibly yes. You are able to guide the ball where you want as you have enough time.

But at pro level? Hardly so. You just make a good swing and that’s all what you can do.

But how cricket is different then?

The difference is not that big, but very vital.

A cricket bat is just wider and, also, flat.

Thus giving the batsman a better control, a bigger license for error.

Had they hit with tennis rackets, it would’ve been even easier.

Besides, the timing of events in cricker is a little different (on a bit slower side, I think).

I’ll probably exagerrate, but what it all leads to is the fact that the very best batsmen in cricket CAN hit sixes almost at will (provided the bowler’s deliveries are not in very tough areas). To a pretty high degree.

How about six sixes off six straight deliveries? Done!

~~~

Guys, pardon me for a rush of graphomania, but there’s another thing.

Who in the world came with the idea that baseball field shouldn’t be a ROUND SECTOR.

I mean why it’s 400 feet in center and 333 feet at the poles? What was the idea?

Let’s see. What is the most difficult thing for the batter and if present considered a huge asset? An ability to hit with power to straight away and opposite field (opposite alley). Imagine Carlos Delgado here.

And, of course, we should not forget that the biggest single play in baseball is a home run.

So, what do we actually achieve by moving the CF fence 400 feet away as opposed to having the fence at 350 away from one foul pole to another?

Don’t you think we just rob the VERY best players of home runs and ourselves from an entertainment? And from appreciating the really outstanding talent, too.

Why create more opportunities - as center-fielder has to play deeper - for hitters with mechanically incorrect swings, who always slash the ball, hit bloopers and all?

See, it’s just another equalizer among many other equalizers in baseball.

- Sergei

Sorry for too many words! :D

Finally, for now at least :D

MGL says,

> I am guessing that that “gut” is worth nothing.  If anyone wants to take the other side and posit that we would find significant differences in performance when players are moved up or down in the lineup, as always, I’m up for a bet!

I might want to ‘posit’ that. Though it has little to do with that “gut”.

But then do any of you, guys, want to get buried into realms of either russian baseball? Or japanese yakyu?

I just doubt it.

Especially the Russian game, ha-ha.

Another thing is that I’ll have to talk mostly about my former team and myself. Can I be objective then and why someone should really care?

Well, when we played in 2006 Russian Series, there were seven (!) former and future professional players (in America) on the field, but still ...

OK, at least I can suggest you watch the Seibu Lions in NPB. Just a bit (use TVAnts - channel NPB-something - or Justin.tv).

No, not because it’s my favorite team there, but there’s something particularly interesting about the way they operate.

For instance, they often swing at 3-0. Even when leading off!

But again, does anyone here care?

- Sergei

"Rally, the computer’s batters are facing different pitchers in different parks, with different weather, etc.?”

Weather and parks are not factors in the APBA game (at least I have park factors turned off).

The schedule is not any more uniform than MLB.  You might have a few cases where teams play stronger opponents in the 2nd half than 1st, but it’s probably a minor factor in the game and real life.  I have trades and callups as the season goes on, so the change in opponents is similar to what MLB players would see.

But even then, I think this data supports your original point.  MLB players are only slightly more variable between halves than computer players, and you need to run an RMSE to see the difference.  If I put the two datasets next to each other without names, I don’t think anybody could tell which set was the humans and which set was the computers.

"If I put the two datasets next to each other without names, I don’t think anybody could tell which set was the humans and which set was the computers.”

THAT is a beautiful statement…

If the unenlightened were to see Rally’s statement, his personal character would come into question.

Sure, computers are capable of generating results very similar to real life.  For as long as their programmed assumptions match real life.  This is a very useful thing, I completely agree.  Trouble is, you never know when the two diverge for any particular player.

I bought Out of The Park Baseball a number of years ago, and had Rick Ankiel on my team.  He was a great pitcher.  Dominated the league season after season.  If I recall, 2009 in the league I was running was his finest pitching year of all…

Just sayin’

Trouble is, you never know when the two diverge for any particular player.

Right, which is why you always state a player’s true talent level with an uncertainty level, even moreso if you are trying to estimate his true talent level six years from today.

The point is that we can model baseball players very well, as a group.  We can obviously not model any individual player.  But, that is irrelevant.

If you are swimming, you want to know how choppy the waves are.  You obviously are not going to know when the big one is going to hit you until it’s too late.  You can state something like I am 10% sure that the big wave will hit me in the next 30 minutes.  Whether it happens or not is irrelevant.  However, your behaviour, how you decide to swim, is based on that estimate.

This is true of every decision you make, be it crossing a highway, or walking your dog.  You can model a whole town, and figure out how many crimes are going to be committed.  All you need are valid assumptions. 

You run a billion simulations, and out of those billion, you just choose one randomly.  That’s life.

I think I’d have to force an Ankiel transformation in my sim league, that’s not going to happen in a program left alone.

My players do not change talent levels during season, and it seems that real players don’t change very much in season, other than a few spectacular outliers.

They do change a bit between seasons, sort of a marcel-type update that goes back before marcel or chone were invented I think (early 90’s), and an aging adjustment.

It’s a subtle thing, a .260 hitter who hits gets really lucky and hits .350 might be a .275 or .280 the next year.  In a few cases a player with ordinary ability will keep that sort of thing up until he’s got a hall of fame career.  We credit that to his exceptional work ethic and hustle (even a sim league - especially a sim league - needs the narrative element).

And likewise, a few first round picks have underplayed consistently and wound up as major busts.  We can credit that to the player consuming too many virtual hot dogs and beers, and not pumping enough virtual iron.

In the end I’ve got a product that imitates life, except when something new happens in the sim league first and then life imitates the art.

> MLB players are only slightly more variable between halves than computer players

I’m just wondering how well a simulation would model the following - take a look at some amusing graphs about Karim Garcia’s debut season in Japan [2005] being full of constant adjustments.

http://www30.brinkster.com/sborisov/jb2005/garcia.html

(Without comments, and incomlete, though)

Start with two graphs at the bottom: “the August 10 (first opposite-field homer) - August 21” and
“August 23 - September 8”.

Will tell you, the japanese advanced scouts were certainly not snoozing somewhere.

And look at Garcia himself, too

red - hits
black - outs
blue - foul
yellow - swing and miss

gree - called strike
white - ball

- Sergei

The graphs above are from catcher’s point of view.

And speaking of attempting a swing at 3-0 count in NPB ...

Looked up the 2008 numbers yesterday

3-0 count, Fastballs IN THE ZONE

NPB swing% average - 89%

Interesting that only three teams were actually LOWER than the average:

Club (Manager)

1. Seibu (H. Watanabe) 77% !!!
2-3. Lotte (B. Valentine) and Hiroshima (M. Brown)

80 and 83 percent (or vice versa, forgot already)

...

12. Yokohama (some Japanese kantoku) - 96%

In 2009, though, Seibu looks like trying to break all kind of records in this regard.

The rationale being that, overall, the 3-0 count in NPB is the only time when you can realistically expect nothing but the fastball (in the zone that is).