THE BOOK--Playing The Percentages In Baseball

Great, Tango.  I’ve been wanting to do that exercise.  Do you really mean it when you say that Jeter may be having the “clutchiest” season ever?  at least so far?  Have you compared his relative WPA stats to other historical WPA stats?

Also, is there an easy way to calculate OPS wins +/-?

I did not compare, but pretty much, anything more than two wins is phenomenal.  For example, when I looked at 1999-2002, Tejada and Giambi were the clutch hitters, and I think they averaged an extra win each, meaning they each might have had a season of +2 clutch wins.

Ortiz last year was almost +3 wins of clutch performance, if you work out the numbers here:
http://www.users.nac.net/thegangof4/rhoids2005/fastDump.htm

Jeter has already performed at almost +2 clutch wins after 64 games.

***

As for figuring OPS wins +/-, easiest thing is to do is:

.03 * ( 1.7*OBP+SLG-1 )

That’ll give you wins above average per PA.  So, a guy with a .430 OBP, .460 SLG, and 290 PA comes in at +1.66 wins.

(Instead of .03, maybe .027 works better.  Just a matter of trying it out.)

What intrigues me right now is Papelbon.  As of a day or two ago, I think he had passed Jeter for the AL (and non-Pujols) lead in WPA.  A strong argument could be made for him to be the MVP at this moment as well.

IF he were to keep up this pace, rack up about 50 saves, maintain his .30 ERA, etc, I’d think he’d be a lock for the Cy Young and should likely win the MVP.  But at what point does he fall out of contention for these awards?  I would think - and this is totally arbitrary, and simply trying to pick the minds of the voters - that any ERA in the double digits (below 1.00) is likely to win him the Cy Young simply for the “wow” factor… unless Santana goes and fires off another 13-0 second half.

But, sad as it is, I think Papelbon wouldn’t win the MVP with 40 saves and a 1.50 ERA - even if Mauer and Jeter fall back closer to their baseline as well.  Any thoughts on this?

Mariano Rivera, 2005
1.38 ERA, 43 SV, .868 WHIP

No Cy, no MVP

I will be unbelievably sad if Papelbon doesn’t win the MVP with a 1.50 ERA and 40 saves.  Awwwwwww Mike

He’ll get a lot of recognition for an ERA under 1.00 though, although I don’t think it will happen

WPA is only one way of measuring “clutchness.” There are so many equally legitimate ways, it is not even funny.  I don’t know off the top of my head if they would all yield similar results most of the time.  If yes, then it doesn’t really matter what method we use.  If not, then it matters a lot and renders the excercise kind of silly.  IOW, we could cherry pick a certain method of measuring clutchness and find all kinds and sundry of clutch and anti-clutch players (if in fact the methods were not very co-related).

I kind of like the WPA method, but I would have to think about it a little more.  Since we are trying to measure a skill, at least we think we are, any method that measures “clutch performance” should mimic as much as possible what a reasonable player would try and do if he thought that he could vary his performance at will.

Another “problem” is that there is a difference between “pure clutch hitting” which might be defined as an ability to simply do “better” or “worse” depending on the game leverage, and an ability to tailor your approach to the game situation, but not necessarily the leverage.  It is more likely that the latter is a true skill than the former.  Some of these metrics, and WPA may be one of them, conflate the two, which is problematic.

To me, there’s a big difference between a high WPA for a closer and a high WPA for a hitter (or starter).  A hitter invariably has a LI close to 1.0 over a season—to elevate his WPA, he has to perform better in his high-LI opportunities.  A closer gets to perform in high-LI situations all the time, because his team can choose when to deploy him (unlike the hitter).  The closer did nothing to create the high-LI—that was done by his teammates, often the preceding starter—but his WPA is inflated as a result.  And since ALL teams can allocate good pitchers to high-LI situations (though they vary in how well they do this), it’s not quite right to measure a closer against the “average” pitcher as WPA does. 

I’m not sure what the solution is here, but I’m convinced that straight WPA overvalues closers.  Papelbon won’t get the MVP, and he shouldn’t.

WPA is pretty much the only way to do it.  The other approaches are essentially a poor-man’s WPA. 

WPA is a function of two things:
1 - the impact that a typical event has in this game situation (which I call Leverage Index), and
2 - the impact that a particular event has relative to the other events, in this particular game situation.  As discussed with the Bonds/Maybe bases loaded situation, the walk was very less damaging that a hit from Bonds, and that a single and a HR were equally damaging from Mayne (which you can call Custom Linear Weights or situational hitting)

For #1, I give a discrete value to every possible game situation.  It stands to reason though that a player doesn’t necessarily respond to an LI of 3.6 any differently than he would to a 5.4 situation.  The player knows it’s very high leverage either way.  So, to do a poor-man’s LI, you could lump all PA into the “late and close” pile, without making a further distinction.  At the very least, it’s easier to understand, and aggregate.  This is what Andy did in The Book.

For #2, and as BBTN discusses, perhaps situational hitting is what drives the player, more than the crucialness of the situation (though the two are somewhat related).

There’s no question that players adapt to their context.  As Andy showed in The Book, protection does affect how a batter/pitcher approach the PA, and changes their style.  As I talked about with basestealing, a LHH with a runner on 1B will be in a better position than a RHH.

In the end, our overall decision-making process won’t change much, since we’ll never have enough information to make a change in our evaluation of a player.  Even if you want to “prove” that ARod is a choke artist, if he’s on your bench, then that’s the guy you’re going to use as a pinch hitter.  We may talk about clutch and choke as a persistent skill, but when push comes to shove, we’re going to base our decision otherwise.

Guy, I agree with you to a point.  While a reliever does nothing to create the situation of the high LI, it is his ability that allows the manager to leverage his skills.

Ozzie Smith and Gary Pettis really did nothing to make a batter hit a ball to SS or CF.  And poor Minky and Helton would love to show off some more at 1B.  But, their particular skillsets make it so that the manager gets to leverage their skills when and where it matters most.  Ozzie may be +20 relative to an average fielder at an average position, but put him at SS, and suddenly he’s a +25 player, because a SS gets 25% more balls than an average fielder.

So, fielders do have a leverage score as well, and they earned it in the way a reliever does: by being better than the next guy.

Studes writes what I consider the single most interesting sports article around: Ten Things I Didn’t Know Last Week.  It’s brilliant in its simplicity, and, like all information-readers, gives me something new to learn or think about.  Like this:

http://www.livescience.com/humanbiology/050209_under_pressure.html

Which reminds me of the Billy Beane / Lenny Dykstra bit in Moneyball.  Beane, the smart guy, always thinking, was befuddled by the pitchers, as he tried to outthink them, only to outsmart himself.  The old “paralysis-by-analysis”.  Dykstra on the other hand just told Beane that he goes to bat, and hits the ball.  We know who had the longer career.

On the other hand, I consider Tim Raines to be one of the smartest baseball players, as well as clutch as they come.  So, who knows.

"So, fielders do have a leverage score as well, and they earned it in the way a reliever does: by being better than the next guy.”

You make a fair point—good closers “earn” their LI to some extent.  But here’s the difference:  We measure Ozzie’s value by comparing him to other SSs—i.e. we measure him against other good fielders.  But WPA (and WS for that matter) compares closers to what an average pitcher would do in that situation.  In fact, teams don’t generally use average pitchers in those situations, so it isn’t the right benchmark.

A related point:  WPA also assumes (implicitly) an average pitcher on the other team.  But in high-LI situtations, both teams have an incentive to use good pitchers.  If all teams leverage their pitchers effectively, no one really comes out ahead, right?  It’s like standing on tiptoes to get a better view of the parade—if everyone does it, no one gets a better view (just tired feet).

In Tom Ruane’s article at retrosheet on the Value Added approach to run creation, he calculates the runs created based on the situation. This is obviously similar to WPA, but takes out the context of the situation (the score of the game). If you had Linear Weight Runs, Value Added Runs, and WPA, you could calculate a Situational Clutchness (Value - Linear) and a Context Clutchness (WPA - Value Wins).

I would think fangraphs should be able to also calculate value added runs with their model along with WPA.

WPA also assumes (implicitly) an average pitcher on the other team

WPA uses what an average pitcher in that situation has done in the past. So Papelbon is being compared to the average closer (and any non closers that are used) when he is brought into a closer situation.

Any thoughts on Torre’s management of the bullpen?  Looking at the WPA of those relievers, particularly the top three and bottom three, and then recalling their usage in varying situations has lead me to believe that Torre has hurt this team tremendously, perhaps even more than I imagined, and I’m a huge critic, and I love run-on sentences.

We measure Ozzie’s value by comparing him to other SSs—i.e. we measure him against other good fielders. 

I don’t.  I put in an average fielder, not an average fielding-SS.  So, if the average SS is say +4 runs above average (+5 if given the SS opportunities), then I bump everyone by that.

***

As for WPA, it assumes that both sides are always equals.  So, if you have Mo on the mound, it assumes that Papelbon is on the other side.

***

Chris: right, that’s the Linear Weights by the 24 base-out states.  You can compute a straight linear weights, a linear weights by the 24 base-out states, or linear weights by the game state (inning, score, base, out).  To keep everything in synch, you should use win values, and not run values.

***

Jody: when you have someone like Rivera, it’s hard to mess up.  Coming into the season, you expected Farnsworth to be much better than Villone, so Torre is using his bullpen then way we think he should haev used it.  But, with Villone pitching so much better, it’s easy to see that Torre should be using him better.  But, maybe he’s pitching so well because his LI is only 0.5.  Who knows at this point?

Has anyone gone back and calculated the WPA from previous seasons using the retrosheet event files?

Woolner did it.

tangotiger, was Woolner’s data published anywhere?

You should check out his work in BP2006 and BP2005.  As well, James Click did some stuff as well in there on baserunning.

Over at BTF, Chris tried to interpret what I said.  Here is my response.

Chris,

You said the following on a post:

He also says that using the clutch rating he’s showing in selecting players to use in a high WPA sitatuion is a bad idea. WTF?

I actually never said it was a bad idea.  I actually said the following:

In the end, our overall decision-making process won’t change much, since we’ll never have enough information to make a change in our evaluation of a player. Even if you want to prove that ARod is a choke artist, if he’s on your bench, then that’s the guy you’re going to use as a pinch hitter. We may talk about clutch and choke as a persistent skill, but when push comes to shove, we’re going to base our decision otherwise.

In other words, one will never get the evidence they want by trying to figure out if a true .360 wOBA hitter is actually .340 or .380 in clutch.  They certainly will never find it if they are looking to see if he’s truly a .300 or .420 hitter in the clutch.  If you’ve got two guys that are close, say both are righties, and both have a wOBA of .360-.365, then sure, I’d use the clutch as the tie-breaker.  But, it’s way down the list of things to look at.

So, if ARod is a true .420 wOBA hitter, and if he is truly a choke artist, the most that someone can reasonably thinks he will choke is down to a .400 wOBA hitter.  Since no one can be that certain that ARod has suddenly become a choke artist, or that someone will persist in chokeness, you might drop him from .420 to .415.  Like I said, you wouldn’t use it to base your decision.  (But, come tiebreaker, you might.)

Hope that clears up what I was thinking.

Tom

Has anyone ever tried to determine “WPA Opportunity”?  Jeter bats before A-Rod in the lineup, so if Jeter comes through in a “clutch” situation (i.e. one that has a huge impact on win probability), then A-Rod’s plate appearance will have less leverage.  It might be interesting to compare a player’s WPA to the “max” WPA he might have accumulated (e.g. if he hit a HR in every plate appearance).

As a general rule, I think any “perfect” type of viewpoint is not a good one to have, and I’m not even sure what the point of it is.  And, it’s not a necessity that a high leverage situation is followed by a small one.  It could in fact be followed by an even higher one.  Jeter’s LI is 1.08 and ARod is 1.09.  Giambi is 1.10, Posada is 1.03.  Damon is 1.02.  There’s no general rule to follow here, other than a leadoff hitter should be lower, since the high LI situations are more with men on base than not.

Click the link on my name for a new “Clutchiness” site.  This is another one of those 1% inspiration, 99% perspiration sites.  I think I’ll start calling them the “Edison sites”.

One minor change I would suggest, for precision, is to take the context-neutral wins above average and multiply it by the LI.  That’ll give you how much his performance would be given the leverage.  Anything about that level is his “clutchiness”.  But, it’s a very minor point, and not worth implementing, but worth bringing up.

I just did a quick test and this equation works out better:
.025 * ( 1.7*OBP+SLG-1 )

(I originally had .03)

So, the example of a guy with a .430 OBP, .460 SLG, and 290 PA comes in at +1.38 wins, not +1.66.

The clutch stars are…

http://clutchiness.blogspot.com/2006/07/first-half-clutch-stars.html

...Pujols, Jeter, Ortiz.

And the chokers are…

http://clutchiness.blogspot.com/2006/07/your-first-half-chokers-all-of-them.html

...tough to narrow down.

Great job Dan!  Interestingly, I had Tejada as one of the two best clutch hitters of 1999-2002, and he’s in the running for top choke hitter of 2006.

Here’s one person that I can confirm does not read this site, fangraphs, or the Clutchiness site:

1. Mister High Standards Posted: July 10, 2006 at 10:45 PM (#2094918)

Hafner, has been great in the clutch too. Not as clutch as Ortiz or Jeter or Jermaine Dye, but very clutch.

Travis Hafner has a WPA of +2.5 wins, which is a great total.  But his OBP/SLG is .461/.650, which is very high.  A guy with that kind of performance, and realizing that he doesn’t have that many clutch opps to begin with, should be at +3.6 wins or so.  He’s been very UN-clutch.

Looking at his ESPN page, and it looks like he’s been performing very clutchy.  Kinda strange, but that’s why we do this, to find these cases.

In terms of someone saying that Hafner has been great in the clutch, I think you have to start with the visceral reaction to the five grand slams.  The perception has to be that any bases-loaded situation is a clutch situation, and thus any grand slam, a huge clutch hit.  But looking at the play-by-plays and Fangraphs for those five slams, only one has really seemed particularly high leverage, while he’s had one each with a major lead and a major defecit.  But just the thought of “five grand slams in a halfseason” has to strike people as clutch.

As for Hafner’s runners-on, close-and-late, scoring position type stats, I suspect, just from a few weeks of tracking and seeing the occasional large spike or drop, there’s probably one or two extremely high leverage outs in there.  Just as a shot in the dark, I searched “Hafner had a chance” and found this game.  Bottom 9, two outs, two on, down one, fly out.  A couple of those could really do some damage to the ol WPA.

Here is Hafner and Thome:

Bat....G....AB....H...2B...3B...HR...R...RBI...BB...SO...BA… OBP ... SLG
Haf...83...286...92...17...1...25...68...74...71...60...322....461....650
Tho...82...289...86...12...0...30...69...77...56...89...298...414....651

Extrabase hits are about even, with Thome getting 5 extra HR, but 5 less doubles.  Haf has 6 more singles and an extra triple.  Haf has the lead with 15 extra walks, and 9 fewer outs.  Overall, in neutral performances, Hafner is ahead by about +0.8 wins, which is also explained by this quick OPS to wins converter:

.025 * (1.7 * OBP + SLG - 1)

Because Hafner had fewer high leverage opps, we expect him to be at about +0.7 wins ahead of Thome.

But, WPA says that he’s only ahead by +0.1 wins.  Essentially even.  In essence, Hafner’s impact has been as if Thome were hitting in the lineup instead of Hafner.  Hafner is getting less bang for his buck.

***

FIVE grand slams?!?  Wow!

Jeter’s clutchiness is at +1.339 and Ortiz is at +1.338!  Pujols is leading the league.

http://homepage.mac.com/dls308/Sites/clutchiness/clutchiness.htm

They are all in the top 15.  Funny, isn’t it?  If I asked everyone to compile their list of the three clutchiest clutchy players, these three would undoubtedly be at the top.  And there they sit.

While it doesn’t prove anything, it’s an extra tough thing for the anti-clutchers to argue against.

Btw, “clutchiness” is a great word, in fact, the perfect word.

For those who don’t follow Stephen Colbert (silent T), he coined the term “truthiness” as a way to explain something that feels right in your heart, even though your brain has no factual basis for it.

Please read all of this:

http://en.wikipedia.org/wiki/Truthiness

After you’ve finished laughing, come back here.

Clutchiness is the perfect word that we can all agree with.  It’s in our heart, we want to believe it, it exists, we feel it, and we couldn’t care less if books tell us otherwise.

“The Oxford English Dictionary has a definition for ‘truthy’ dating back to the 1800s....’The fact that they looked it up in a book just shows that they don’t get the idea of truthiness at all,’ Stephen Colbert said Thursday. ‘You don’t look up truthiness in a book, you look it up in your gut.’”

I’m sorry I didn’t finish all the reading,but is there any opportunity that Jeter makes this incredible WPA with his OPS as the leadoff baserunner&scoring guy?If so,that’s not a so called performance of “clutch”,is it?I don’t really understand the WPA formula,so it’s just a guess.If this is a rookie mistake and bothers you,then I’m sorry.