THE BOOK--Playing The Percentages In Baseball

What percentage of all PA are over 2? Over 3?

Barring injuries, the playoff LI shouldn’t be too big of a deal since we’re only comparing teammates. Beltran and Wright would both presumably play in any important late-season contest.

The thing you need to answer is: What is more tension-filled, the 10.0 LI in a game that won’t affect the standings to a 2.0 LI in a game that will affect the standings?

Is that the right question? I suppose it is if you are asking which scale do we want to measure clutch on, ‘Game Leverage Index’ or ‘Season Leverage Index.’ Since season leverage index can be affected by other games, perhaps being played simultaneously, the question becomes what did the hitter know and when did he know it. If your season is on the line in the bottom of the ninth, but meanwhile across the country a miraculous comeback puts you in the postseason regardless of the outcome, is the clutchiness of the AB lost?

I don’t doubt that a player with a clutch skill should conceivably be able to raise their hitting skills when the game is on the line, and the degree to which they raise their hitting skills should likely be affected not only by how important the outcome of the AB is to the game, but the season as well. However, the latter requires knowledge of game states outside of the current game.

I was going to go with a straight Leverage Index (say above 2.5 or 3.0 or some such). Do I still stick with this, or do we want to factor in the games behind and games left?

For this simple, probably extreme example, I would vote for sticking with straight leverage index. However, if there were a framework for ‘Season Leverage Index’, I would definitely go that route.

Fun project Tango, looking forward to choosing hitters.

I had the inspiration in a thread at SOSH:
http://sonsofsamhorn.net/index.php?showtopic=26762&st=20

I wrote this that would apply here:

Ah, meeting a double-threshhold, rather than multiplying, is more appealing.

After all, game 163 will have virtually all the PA more tension-filled, even if the LI is below 1.0, and some random game in the season has an LI of 10.0.

But, if we set certain thresholds, like any LI over 4.0, regardless of standings, all the way down to an LI of 0.5 for a game 163 situation, that will likely mimic how a person will perceive the tension, since not every player is solely focused on just making the playoffs (otherwise, why would Royals players show up?). There is a certain amount of tension a player perceives simply based on the game being played, and then an added tension based on the playoffs.

In order to mimic how a player feels, we’d really need to know how a player feels.  How much of the tension is game-based, as opposed to standings-based?

I’d have to hope you could come up with a way that doesn’t make the 10.0 PA and the 2.0-playoff PA an either/or proposition.  I’d be inclined to say the 10 is more likely to color people’s votes, especially the “down ballot” guys that may end up being chosen as the #2.  But I’d also think that playoff-weighted should probably put most teammates on a fairly level LI.  That’d probably be better for the study, though again, I’m not so sure it doesn’t put you at odds with how a fan’s memory works—though perhaps that’s part of the point.

PS: Did you see the Pete Palmer interview on THT where he seems to suggest he’s doing his own clutchiness project?  Seasonal LWTS vs. PBP LWTS vs. WPA.  Good idea, that… surprised nobody thought of it sooner.

Fangraphs did it already last year.  And as Palmer said, Cramer did it 30 years ago.  Nothing new there.

Vlad is the man on the Angels.  After him, I’d take Hunter as the best hitter and Garret Anderson as the clutch guy.

Maybe im still affected by the way the season ended but as far as mets go I would have to say that the person I feel most confident in a big spot would be David Wright.

I was going to offer this bet to King Kaufman, a long time ago. 

Pick any number of players you think are clutch.  Pick any number of players you think are NON-clutch.  Pick any situations you consider clutch situations.  (There is no restriction on AB or anything, pick ‘em the way you want.) I will even let you weight the situations differently, within reason. 

At the end of the year, we compare the clutch performance of both groups.  For each point of batting average in the expected direction by which the clutch BA exceeds the non-clutch BA, I pay you a dollar.  For every point in the NON-expected direction, you pay me $1.25.

So if your clutch guys hit .300 clutch, .290 non-clutch, you win $10.  If your anti-clutch guys hit .250 clutch, .265 non-clutch, you win another $15.  But if it goes the other way, you owe me $15 and $22.50, respectively.

The larger amount for me is because you get absolute discretion to pick the players and the situations.

If you believe in clutch, but you don’t think this is a fair bet because of the $1.50, what DO you think the right price is for a fair bet?  If you think the only fair price is $1.00 to $1.00, then you don’t believe in clutch at all.

(I am actually willing to make this bet, if anyone is interested, at any reasonable stake.)

Ooooohhhhh..... I like that, too. 

I like mine, because it is deathly simple: who do you want?  The good hitter who moves it up a step, or the even better hitter who can’t seem to raise his game when it counts?

In your case, it’s more involved, but probably more powerful.  You are giving absolute discretion to the other guy to define everything.  In this particular case, it works best with a one-on-one.  Getting a consensus on crucialness will not be very strong (and likely simply result with the highest leverage in LI).

If 10 yapping clutchers (who are bigshots in the mainstream) would like to step up and work out a bet, post here, and we can discuss that too.  Put your money where your mouth is and speak now, or forever hold your peace.

I like yours too, but I think there will be some people who would argue that, say, Eckstein isn’t better than Pujols in the clutch, but he closes the gap a fair bit.  For those people, try my bet!

(I need to add one thing to my bet: all the differences need to be adjusted by the league difference.  So if your clutch guys outhit by 10 points, but the league outhits by 2 points, you only get 8.

Also, I meant $1.50, but mistyped $1.25 above in one place.)

Perhaps I will post both bets on my blog later, see if anyone bites.

Before you go making all these bets, I’d make the period of performance in the future.  If the perception of clutchiness is based on 2007 performance, then the stats of 2007 might reflect that.

Yes, that was implied.  You pick the players before the start of 2008, and we settle based on the 2008 stats at the end of the season.

How about a listing of “choke” hitters as well?

It won’t work for my bet, since I’m looking at actual production, rather than a difference in production of the same player in two different game contexts.

And if the question is: “Who would you least like to see there”, it’ll be a fight between Endy Chavez and whoever the backup middle infielder is, neither of which will probably get enough plate appearances.

I think it’s a great project, and hopefully I can send a fair number of Reds watchers over here when the voting commences.

I am curious about what you plan to do about injuries.  If a player is hurt or otherwise has limited PA’s, that would almost certainly make his top-50 PA’s less crucial than the other players, simply due to the small sample you’re choosing from.  And if clutch guys do better in higher-leverage situations, then lower leverage situations might not be “exciting enough” (or whatever) for them to perform like their most clutchiest selves.  So should there be a minimum number of PA’s required (e.g. 500?) before you bump down to the next guy on the list?
-j

Definitely a minimum.  Probably at least 30 PA. Somewhere around the LI = 2.0 or 2.5 mark.

About 20% of a player’s PA occur at LI of 1.5 or higher.

to truly know this you not only need to define and observe clutch situations but you need some insight into the minds of the players.  impossible i’m sure, but it would be really interesting to know how players react to different situations (such as by increasing heart rate, or sweaty palms, or some other physical or psychological symptom).  some players are cool as a cucumber, others may see a spike in adrenalin.  let’s see if their is a noticeable difference in performance for the guys whose vitals stay the same all the time.  and let’s see if the guys who physically or psychologically react in some way see different performance results.  could be interesting.

I have a theory, does it make sense?

The Clutch guy will have a slight increase in BA, slight decrease in OBP, and slight decrease in SLG.
The best overall hitter will see a slight decrease in AVG, slight increase OBP, and slight increase in SLG.

I don’t think “clutch” exists, but pitching around batters do exist. The better hitter will get pitched around more often in high leverage situations, thus increasing OBP. He’ll have less to hit, and his AVG will suffer very very sightly, but when pitcher makes a mistake, it’ll be a big mistake.

The clutch guy is looking for a hit. The reason why people think “clutchiness” exists is because people remember hits, not walks. The clutch guy will get more hits, but probably more singles because he’s looking to make contact. He won’t BB as much and his OBP will go down. Also his SLG is going to decrease because he’s looking to make contact.

This is all assuming that the clutch guy who’s picked isn’t already the best hitter picked, and the clutch guy is a lesser hitter in general. Picking Pujols as a clutch guy doesn’t really apply to my theory, since he’s already the best hitter.

As for picking players, I think you need to pick 2 players of each per team. You should only count 1, but what if that guy gets injured in April and is out for the season. There has to be some backup plan just in case.

MLBfan - that’s an important confound.  We can’t reward Bonds or Ortiz because pitchers are irrationally scared of them.

rob, you hit on what I think is the BEST way to measure this.  Heart rate, sweaty palms, etc., in clutch and non-clutch situations.  We can’t get in their heads, but with a few electrodes we’d be able to get a great glimpse at what chemicals their heads are telling their body to produce.

And unlike baseball stats, you could theoretically identify “clutch” players by their biology in their very first season!

I definitely think that measuring the medical side is the most important way to identify whether or not “pressure” situations exist. You would think they do, but hitters may not feel any different. In my sport I’ve been in some pressure filled situations and sometimes did great, others just average. I would bet that the more times a player has been in those situations, the less they physically affect them. Experience probably does play a role to some extent, which might go against the SABR view.

But think about it… when a guy is nervous, the muscles will tighten up and it will affect his reaction time, and probably some power (still muscles will already be more contracted so when it’s time to contract them, there will be less potential enegery than if they were relaxed). The guy who’s young and really cares might get nervous, while the veteran might not. If this is the case, then what I said above probably holds true.

It would be interesting to see if players would allow small moniters/machines to be hooked up to them to record this stuff. I would think most players would view this as a distraction, but some might be accepting. Another aspect is that if they were hooked up, they would most likely be nervous because of it, at least initially.

(I’m just speculating above, but it makes sense from my athletic history and educational knowledge and experience)

I hope you guys aren’t serious about seeing the heart rate of a player during the game. Even if they allow it for some odd reason, why would it be useful? Do I really care if Aaron Boone’s heart was racing or if he was nervous or calm when he stepped into the box against Wakefield? No. Whether he was nervous or not, all that matters is that he performed in the clutch.

I love it!  How about we do it for lots of other things that people claim they can “see”?  Can we get some sportswiters and/or baseball execs involved?  You’ll have to get someone like Neyer involved, who can put in e-mails or calls to baseball insiders.  If he can’t, he can surely get some people who can (like Olney).

Tango, I think you KNOW who is going to win this one, so it is kind of a moot exercise, is it not?

And you know that all the people who took players who were not the best, are going to have some excuse as to why their picks (collectively) lost, or they will criticize some part of the methodology - or something.

Phil, I’ll back your offer for ANY amount of money if anyone wants to wager!

Remember, it is one of MGL’s rules.  To paraphrase - you can usually tell whether someone is serious about a (fact-based) conviction.  Simply offer to bet them money.  If they don’t really believe in their own conviction, they will give you some excuse as to why they won’t bet on it.  Of course, there are many “legitimate” reasons not to bet, even if you think your convictions are accurate.  But for those who do, they will take a reasonable bet maybe 10-20% of the time.  If they don’t believe in their own convictions, they will take a (non-trivial) bet almost 0% of the time.  I’m not talking about a $5 fun bet or a steak dinner of course.  I am talking about a bet that hurts a little if you lose.

I have offered many, many bets on BTF, etc., over the years.  Never had one taken.  And I am always quite serious.  And my offers are always for ANY amount of money.

BTW, I LOVE they way you couched your offer (YOU pick the players, you pick the criteria, you even pick what you think is a fair price, etc.).

MGL, I think it will be moot, but not to the people participating, which is the point.  Same deal for Phil’s bet.

And I agree, I can’t wait for the excuses.

dan - Boone nailed it.  Does that mean he was calm in the situation, and it was inevitable that he was going to hit a HR?  It’s entirely possible he was scared shitless and just guessed at location and swung hard.

People who are calm in the clutch can still hit screaming line drives directly at fielders.  And people who are nervous can quite easily get a Luis Gonzalez-type blooper that just falls in there, and wins a world series.

I don’t understand your issue.  If all you care about is results… fine.  My contention is simply that I cannot predict clutch results… but with heart rate data and a few similar data points, I think I might be able to.

The A-Rod - Jeter - NY sportswriters dynamic was born for this study.

Bobby/27: u kidding me?  After ths season ARod had last year, no way ARod doesn’t win the balloting for the guy the Yanks fans most likely want at the plate.  Basically, I’m going to have to go to the #2 guys in both sets.

Of course, if this was the playoffs…

Tango, 3 things:
1) How are you going to decide the “best” player? I mean, Pujols or Guerrero are pretty easy, but what do you do for a team like the White Sox, where 4 players (Swisher, Thome, Konerko and Dye) are pretty even and all of them have a reasonable chance of being the best player in any given year? Do you only use 2007 data or career data while giving more weight to 2007?
2) How will you handle injuries that seriously affect playing time?
3) Finally, career years and slump years can take place. I don’t believe in clutchness and I think that the “best” player is also the one who’s most likely to do well in the clutch, but the “best” player can vary from year to year. You go with Travis Hafner for the Indians and he completely flops the year. He also flops in clutch situations, but it is only due to his overall poor season in the end. I think you ought to make yourself a little concession here and just pick the best player at the end of 2008 by using whichever qualifier you decide beforehand (highest OPS, highest VORP, highest EqA, highest whatever you want). This way, if Hafner flops and Lowell has a career year, you’re covered, and you’re not actually cheating anybody because the qualifier rigidly covers whoever you define to be the best overall (and fans still get to pick their own man, regardless - if the top guy is the same, you still go down to number 2, except you do this after the season rather than before).

I also note a serious issue here, but maybe it’s just me. Sabermetrically-inclined types aren’t going to bite, but maybe some old-school guys who want to prove you wrong will. However if you start using LI or WPA at the end of the season they might start claiming you’re using newfangled meaningless stats designed to prove them wrong rather than “neutral” ones such as BA/RISP or game-winning RBIs or whatever. How are you going to solve this problem? I’m not so sure there will be many takers who will both accept this and fully accept the results at the end of the season…

Reds fans are lining themselves up to vote against Adam Dunn:
http://www.redszone.com/forums/showthread.php?t=64873&highlight=tango

***

Phil, backed by MGL, is taking your bets:
http://sabermetricresearch.blogspot.com/2008/01/i-will-bet-real-money-that-you-cant.html

***

Rene/29:
1. I’ll use Marcel.  If it’s close, I’ll defer to the clutchers.  As an illustration, say that Thome is just a shade ahead of the other 3 guys according to Marcel.  But Thome also leads as the votegetter for clutcher.  I’ll give the clutchers Thome, and I’ll just pick the #2 guy as per Marcel.  This was really to prevent a case like Pujols or Vlad, who are just much better than their teammates.

2. I don’t know yet.

3. No way will I choose the best player after the fact!  There’s no way the clutchers will win.  Ever.  As for career years, etc, that’s why I’m aggregating 30 players.  And in any case, a guy can have a career year in the clutch too, right?  It works both ways.

4. I’m not going to just use WPA, etc.  I’ll present the entire spectrum of data. The goal of this project is to shut people up, and I’ve got to do that by making them eat their own words, not mine.

Tango/30:

Ok, that clears it up. You sound like a man on a mission, and I appreciate that you’ll give the clutchers their man if it’s a close call. I know there’s no chance that they’ll win as a group, but the rules have to be strict as there’s going to be a whole lot of mirror climbing. Having said that, I wonder how many are going to go with the Ecksteins of this world. Good luck with the project, it sounds fun.

Tango/30, don’t be too quick to judge us Reds fans based on the opinions of those on the Sun Deck board....

http://www.redszone.com/forums/showthread.php?t=64885

Though certainly the topic ruffles feathers for everybody involved.

Between my project and Phil’s bet, we should have something for every clutch yapper.

There are relatively few players who performed noticeably worse than their norms over long careers in low leverage situations.  Barry Bonds and Sandy Koufax both did.  I am not aware of any who performed noticeably better than their norms over long careers in high leverage situations.  Phil is not in danger of losing all his worldly possessions!

I am quite sure that you would find greater deviation from career norms in low leverage situations than in high ones.  Koufax would undoubtedly throw fewer painful curveballs when the game was out of reach, whereas a healthier pitcher (say Bert Blyleven) might continue to pitch more or less the same way.

Nobody biting on my end yet.  If anyone sees any clutch advocacy over the next while, from sportswriters or whatever, let me know and I’ll solicit them directly.

You should be able to make good money.  $100 a bet, 10 bets.  You pay out 500$, you take in 1000$ (presuming you win 5 and lose 5).  You’ll make 500$ for every $1000 bet.

Yes, but nobody’s betting!! 

I just posted at SOSH.  You should be getting tons of hits very soon.  And I would hope lots of Ortiz bets.

I also posted on the Reds forum about Phil’s bets, and I’ll make a blog post about it tonight.  Maybe you’ll get an unclutchy Dunn bet or three. -j

Phil/Tango, do we know that clutch versus non-clutch “difference” as opposed to ratio is the way to go?

If I pick good overall hitters as my clutchers and bad ones as my chokers or vice versa, might the ratio be the same for both, but not the difference?

The basic age-old problem with trying to get “yappers” to eat their words, even their own words, is that the “yappers” rarely understand the methodology and thus rarely concede that you won.  Also yappers tend to be - I’ll delicately put it - a little “off,” or at the least, not the sharpest intellects in the tool box.  Thus they tendto proffer pretty much any excuse in the book, even bizarre ones, after losing the proposition.  IOW, these things tend to be a no-win situation, at least from the standpoint of getting people to eat their words.

Just take any of these late-night infomercial products, and present solid evidence right to the face of their inventors (the Dr. Franks of the world) and purveyors, that they don’t work (which, not surprisingly, most of them don’t).  If you could do that, you would just get some gobbledeygook rebuttal.  You are NEVER going to hear, “Gee, it looks like you’re right!” Ego (looking good and not being wrong) and money are far too powerful forces.

But, interesting project nonetheless.

For any sportscaster or journalist who takes the bet, the price should be that if he loses overall, he is never allowed to mention “clutch hitting” as if it were true, and in fact, he has to mention, once a week or month, that there is no evidence that it exists, beyond a “de minimus” amount, other than as random fluctuations among all players, or something like that.

BTW, in the interest of full disclosure, we did, in the research mentioned in “The Book,” find a small clutch skill among players.  Not a lot, but a skill nonetheless.

So SOME of us at least, are NOT claiming that a clutch skill does NOT exist.  Only that it is small and extrememly difficult to “detect” over less than an enormous number of PA, such that the better overall player will almost always come out on top, on the average, versus a similar, but not quite as good overall, player, who has demonstrated a large clutch split in the past.

If two players are exactly equal overall, however, the one who has demonstrated the largest positive clutch differential in the past will likely (greater than 50%) come out ahead in clutch differential in the future.  Again, This effect is likely (we don’t know for sure - 100%) very small.

Ratio probably would be better ... but I figure at 2:1 odds, I can afford give the bettor the slight advantage of picking a .340 hitter over a .240 and pocketing the extra point or two.

Actually, I wonder ... do you think it’s actually ratio?  Maybe it’s ratio based on the odds ratio instead of the raw numbers?

Even if it’s not the best way to go, using the difference is more straightforward than the ratio for most people.  So I’d definitely recommend going with the difference.

Also, I’d just like to suggest that if we want some big name individuals to actually step up and take Phil’s bet seriously, we probably want to avoid critiquing their intellect in these threads.
-j

Well, should be easy enough to “check.” Take all those with OPS above .750 (or BA, OBP or whatever) and look at their clutch to non-clutch ratio and difference and do the same for players with OPS of less than .750 (or whatever).

No takers yet on either front? I’m disappointed. Why do they all chicken out?

You can read the SOSH here to see how people are thinking:
http://sonsofsamhorn.net/index.php?s=&showtopic=26762&view=findpost&p=1340169

If I can summarize: they don’t like the playing field Phil is offering, and don’t know what kind of playing field they’d need.

You can also jump down to post 60 to see what else I am saying.

Ah, I got a friend to bite. Giants fan, believes in Bengie Molina and Aaron Rowand. He didn’t sound too confident, but still decided to take the bet.

This series of statements by Tango, from the SOSH thread, is spot on and should be the end of the discussion until and unless someone wants to take up Phil’s offer and hammer out the details:

So, how much do you believe in Clutch? Ask Phil for 60% odds (1.5 to 1). Maybe he’ll bite at that.

But, if you insist on claiming great insight into Clutch, then step up to the plate, and put your money where your mouth is. If you don’t like the odds, then negotiate.

I will add that if you are not more certain than 60% (3-2 odds) that your clutch player will “outperform” your choker (remember you get a built in “swing” or “double counting” by being able to compare a clutcher to a choker and not just a clutcher to an average player) over the course of an entire season, then you are barred from EVER talking about how so-and-so is such a great clutch player, and that’s the guy you want at the plate when the game, season, or series is on the line, etc.

I can’t speak for Phil, but I’ll take 3-2 odds for ANY amount of money…

Amazing how the excuses are flying BEFORE the contest actually takes place.  Perfect example of one of MGL’s rules (about what happens when an “expert” challenges an “amateur naysayer” to put their money where their mouth is).  What you almost always get is, “well, uhm, ehh, hmmm, sorry, I have another call right now...”

The “problem” with the bet (among many things) seems to be that anyone who has touted the “virtues” of clutch (whatever that means) thinks that 2-1 or even 3-2 is too much to fade.  They will argue till they are blue in the face that even if clutch exists as a skill to a large (again, whatever that means) degree, the true odds should be somewhere around 55-50 (or something like that, I am assuming).

What they don’t realize is that odds of around 55-50 for an entire season and for a pool of several (10 or 20) clutch and choke players, represents a chance of detecting clutch of near zero.

And for those of you who keep saying that I, Tango, and other analysts are saying that clutch as a skill does not exist, please read the chapter on clutch hitting in The Book.  That FALSE premise (that serious analysts do not think that clutch as a skill exists) is getting old.

BTW, why not throw in streak hitting and batter/pitcher matchups into the fray?  I think those are believed in (at least spoken of) more strongly than clutch hitting.

I’ll take the same bet, at 3-2 odds (maybe less), for best/worst batter/pitcher matchups AND I’ll take the same bet for anyone who during the season wants to put a player’s game AB’s or PA’s into a “cold” or “hot” bucket as the season progresses.  IOW, on any given day, you can designate that day’s game as either hot or cold for any player.  At the end of the season, we’ll tally the hot and cold PA’s for any metric you want.  If that player, or a group of players, has a higher BA, OBP, OPS, or whatever, in their hot bucket than their cold bucket, after accounting for the pitchers faced, stadium, and weather, you win $1 (or $10, $100, whatever you want).  If the other way around, I win $1.50.  You can even use injury status if you want!  I could probably be negotiated down a little, although not too much because of the injury thing and because people might be able to choose favorable or non-favorable situations that cannot adequately be accounted for after the fact.

The batter/pitcher matchup should be a no-brainer.  That would be before the season starts.  I’m even willing to go with 1.25 odds on that one (1.25/1).

MGL/50: Agreed, 100%.  I would make those bets too, although, for now, I’m trying to keep things simple.

How would you handle batter/pitcher matchups?  The direct results of the matchup would have to beat log5 for the bettor to win?

How would you handle batter/pitcher matchups?  The direct results of the matchup would have to beat log5 for the bettor to win?

Exactly.  That is an easy one.  The details would have to be worked out.  We would want people to choose say 10 batters and then choose 5 pitchers who they should do real well against and 5 pitchers they should not do well against.  Of course, we measure “well” or “not well” against what they should do based on the batter and pitcher’s projection or based on their full season stats.  The reason why we have to use good and bad matchups for every batter is that if we don’t, it is going to be hard to figure out the baseline.  If we use pairs of pitchers (good and bad) for every batter, we don’t care much about the baseline, at least for the batter.

Is there any traditional baseball person, including the most respected “insiders” in the game, who doesn’t think that they know or can find out which pitchers or at least type of pitchers a batter should do well or not so well against?

In the baselines, you would have to include GB/FB platoon matchups. One reason why certain batters do match up well (or not) against certain pitchers is because of each player’s repsective GB/FB tendencies.

For example, one reason (maybe the only reason) why a “good highball hitter may match up well against a high fastball pitcher” is that the batter swings down on the ball and is thus a ground ball hitter, and the pitcher is a fly ball pitcher, which high fastball pitchers tend to be.  We know that there is a significant platoon advantage with respect to GB/FB tendencies, so that has to be included in the baseline.  In fact only 3 things need to be included in the log5 or odds ratio formula to establish the baseline.  One, the overall pitcher and batter OPS (or whatever), two, the batter and pitcher’s true platoon ratio, and three, the aformentioned GB/FB tendencies of the batter and pitcher.