THE BOOK--Playing The Percentages In Baseball

Yes.  This is a perfect example of why WPA is a great stat.  One of the best things to come out of sabermetrics.  It doesn’t answer any of the questions that we are usually looking for with our metrics, but it does tell us a lot.  Not who should win MVP, or who the best player is. It measures pure baseball awesomeness in a way that no other statistic could.

Denard Span led the league last year in PPA, which is playoff probability added.  He seemed like a perfect example of why that stat was even more ridiculous than WPA and meant less about how good a player is.  I won’t disagree with either of those points, but Denard Span contributed to a couple of my high points of last summer, and the absolute best point of my baseball life.  WPA, and PPA, which, I agree, is kind of ridiculous, but still a fun stat, measure something that no other stat has ever come close to measuring.  Except RBI’s.  That came pretty close.

Exactly!  I still don’t understand why anyone would call WPA a “junk stat.” It doesn’t do what the standard analytic crowd wants it to do, doesn’t fit into their sabermetric bag of tricks, I guess.  To me, that’s the point.  It does something else entirely, something that is kind of cool.

Is the WPA affected by Mariano Rivera being the pitcher?

Cyril - No, it doesn’t.  WPA is based off just the average win expectancy of each situation.  Anyone with Retrosheet database wanna check out the win expectancy of opposing teams against Rivera in the bottom of the 9th with a runner on 2nd and 2 outs?

Nick

Thanks. Here is a Rivera’s Fangraphs page

http://www.fangraphs.com/statss.aspx?playerid=844&position=P#winprobability

It shows that he has a clutch of -2.01 for his career. Maybe negative is good for a pitcher. But I don’t think so. As I understand it, from reading this blog, clutch at Fangraphs is WPA -WPA/LI. If I look at Fangraphs that should be 47.92 - 28.09 (which would = 19.83).

So I asked because I assume that with Rivera on the mound, an opposing team’s chances of winning might be less than average. Now I am not sure. I realize it might be hard to take the opposing batter/pitcher into account when doing WPA.

Anyway, Rivera does just about as well in close and late situations as he normally does (I know that does not answer your question). Here is a link to his Retrosheet splits

http://www.retrosheet.org/boxesetc/R/Lrivem0020.htm

Cy

The best Retrosheet splits to use, particularly in regard to WPA, are the ones for high, medium and low leverage.

Get this: in high leverage situations, Rivera has struck out 35 batters and walked NONE.

WPA can take any parameters you want.  It’s limited only to the data and your imagination.

The WPA as implemented by Fangraphs does NOT include the identity of the players.

***

Mo has performed much better in low-leverage than hi-leverage situations in his career, hence his below-average clutch value:
http://www.baseball-reference.com/players/split.cgi?n1=riverma01&year=Career&t=p#lever

But remember, he is being compared to himself. He is GREAT in high-leverage situations and SUPER-GREAT in low-leverage situations.

WPA is great at measuring the impact of what a player has done.  It just isnt as good at predicting what the player will do in the future.
As long as we are using it for what its good at, its a great stat.  Its just not the best tool for a GM to use to make personell decisions.

Has anyone done historical WPA for the postseason?  I’d guess Mo is probably #1 all-time.

Great idea Trev.  I would really like to see that.

WPA is great at measuring the impact of what a player has done. It just isnt as good at predicting what the player will do in the future. As long as we are using it for what its good at, its a great stat.

I would disagree with this, marginally. WPA measures the impact of what a player has done, if we don’t know the outcome of the game in question.

In other words, WPA treats the value of the ninth-inning homer as different from the second inning homer because anything could happen after the second inning home run (with varying probabilities attatched, of course).

But if we want to talk about the results of yesterday’s games, or of June’s games, etc., we don’t have to guess at what will happen after the second inning. And from that perspective, a home run in a one-run game is equally as valuable if it occurs in the second or the ninth inning.

Tom, thanks for reminding us that stats can be used to evoke the fan’s experience of baseball, the simple rooting interest, as well as various kinds of analysis.  The idea of reducing the feelings of a game into a graph anyone can understand (I showed it to my wife and she got it immediately, though of course she does have the Ph.D. from MIT and all) is wonderful even if it serves no other purpose. 

For those of us whose teams spend most of the summer far away from the pennant race and SportsCenter, we learn a lot about why we began to love baseball in the first place.  A well-executed play, a well-timed strikeout, a favorite player hitting a home run in the twilight of his career.

a home run in a one-run game is equally as valuable if it occurs in the second or the ninth inning

Well, I don’t know. If a team hits a lot of 2nd-inning home runs, their w% will be lower than if they hit a lot of 9th-inning HRs. Likewise, if the Red Sox let Papelbon pitch the 1st inning (instead of the 9th) every other day, their w% would fall.

Also, if Ichiro had homered in the second inning instead of the ninth, I don’t think you can assume that the game would’ve unfolded the same way with the Mariners still winning 3-2 and Rivera pitching the 9th down a run. So I don’t see how the 2nd-inning HR has the same value as the 9th-inning HR.

Well, I don’t know. If a team hits a lot of 2nd-inning home runs, their w% will be lower than if they hit a lot of 9th-inning HRs

I don’t see any reason to believe that this is true.  Runs count the same regardless of when they are scored.

Likewise, if the Red Sox let Papelbon pitch the 1st inning (instead of the 9th) every other day, their w% would fall.

This is different, since in this case Papelbon would be deployed w/o knowledge of whether the game would be close or not, and sometimes would be “wasted” in a blowout. 

If you could hit home runs at will, but were limited in when you could use them, then I would agree that you’d want to save them for the ninth inning of games where you really need some runs.

The value or worth of what you just did, be it baseball or life, is independent of the exact actual future events, and is dependent on the expected future events.

The value of what you have already done is dependent on the actual future events.

When you buy a lottery ticket for $1 (and for the sake of this argument, let’s presuppose that the government pays out based on tickets bought, not a 50/50 situation), the value of that ticket is $1.

If after the numbers are called out, and you don’t win, the value of that ticket is $0.  If you did win, it’s worth $1 million.

***

Hitting a HR in the 1st inning has a certain amount of value (say +.10 wins).  Hitting a HR in the seventh inning to give you the lead has a certain amount of value (say +.50 wins).  That’s based on expected future events.

Once the game is over, a run is a run is a run.  The value of the HR is based only on the base/out state, and not on the inning, score, base, out (or more accurately, by WPA/LI).

The answer to the value question is completely dependent on what your assumptions are to begin with.

The flipside of being able to hit home runs whenever you want is to prevent runs from scoring whenever you want, which in theory is what the closer is supposed to do. Obviously it would hurt the Red Sox if they didn’t use Papelbon in the late innings of close games, just as it would help them if they could hit home runs at will in the same situations.

Even after the game is over and we know the outcome, a run isn’t always a run. A decent closer will be 20 runs above average (based on innings pitched and runs allowed vs league average) but might have a WPA of 5.0, and in some cases will also win an MVP/Cy Young. Obviously the closer has some value that isn’t being measured by his innings/runs allowed alone.

I think people are willing to condede this point, and so maybe the issue is that while Ichiro can’t really control when he homers, Papelbon does have a degree of influence (via Terry Francona) as to when he pitches scoreless innings. And so it’s OK to give extra credit to Papelbon because of the “intentional timeliness” of his performance, whereas people aren’t quite as comfortable crediting Ichiro because of the “random timeliness” of his home run.

Still, if you accept the premise that value is generated when it occurs, and not after the fact, then “a run is a run” doesn’t hold water.

Even after the game is over and we know the outcome, a run isn’t always a run. A decent closer will be 20 runs above average (based on innings pitched and runs allowed vs league average) but might have a WPA of 5.0, and in some cases will also win an MVP/Cy Young. Obviously the closer has some value that isn’t being measured by his innings/runs allowed alone.

It’s obvious only because you have accepted the WPA conception of value.  Which is fine, mind you--my objection was to your statement “If a team hits a lot of 2nd-inning home runs, their w% will be lower than if they hit a lot of 9th-inning HRs”, not to the concept of real-time value with which I have no problem (but also do not accept as the one true value, which few do).

Still, if you accept the premise that value is generated when it occurs, and not after the fact, then “a run is a run” doesn’t hold water.

Sure.  But this is a far cry from the statement that a team will have a higher W% if they hit homers in the ninth rather than the second.  The home run in the ninth has more real-time value (assuming it occurs in a high-leverage situation and not in garbage time), but if you simply took away an equal number of runs in the second inning and scored them in the ninth, you would not be any better off.  And you would have to be demonstrably better off in order for the statement in question to be true.

As long as someone can answer the question: “What is the value of a lottery ticket”, and his answer goes one of two ways “Whatever you paid for it” or “Depends on if they called the numbers out”, then that decides for that person on the meaning of “value”.

Ichiro had fantastic in-the-moment value, and limited after-the-fact value, for his HR against Mo.

A person simply has to choose his viewpoint, neither of which is right or wrong.  It’s just his opinion.

I think Dackle is saying that a team with more ninth inning HRs will have a higher win% than those with more 2nd HRs because better pitchers are used in high-leverage situations. Therefore, a HR in a high-leverage situation, when the opposition is presumably better, will have more value.

At least, that’s how I think of it.

Well I was thinking of it as the flipside of the closer—your w% will be higher if you use your closer in the 9th rather than the 2nd, so presumably, if you could hit HRs on demand, you’re better off doing so in the 9th than the 2nd.

On the continuum of baseball statistics, with “descriptive” on one end and “predictive” on the other, I’d put projections on the predictive end, actual season stats somewhere in the middle, and WPA almost purely on the descriptive side.

No, 9th inning HRs will not produce any more wins than 2nd inning HRs.  You’re mixing up inning of game and closeness of score.  Using your closer in the 9th inning of randomly selected games wouldn’t be any more valuable that using your closer in the 2nd inning of those games.  The key element is using the closer in games with close scores.

Similarly, HRs in close games are more valuable than HRs in blowouts.  But inning is irrelevant.

The leaders in WPA from 2006-2008:

+21 Pujols
+16 Berkman
+14 Howard, Ryan
+13 Ortiz
+13 Wright
+12 Beltran
+12 Holliday
+12 Vlad
+12 Cabrera, Mig
+12 Manny
+10 Gonzalez, Adr
+10 Ramirez, Ar
+10 Morneau
+10 Tex
+10 Lee, Carlos

I agree that it’s less than the midpoint along the continuum, but in no way is it all the way on the descriptive endpoint.

It is a smidge less seasonal stats (and the more seasons, the less smidge).

If you are talking about only 1 season’s worth, it’s still pretty close.

A year-to-year correlation of WPA is slightly inferior to wOBA.

Similarly, HRs in close games are more valuable than HRs in blowouts.  But inning is irrelevant.

Said a little differently, home runs in high-leverage situations are worth more wins than home runs in low-leverage situations.  High leverage situations tend to take place in the late innings of close games. You’re saying that it’s the “close score” aspect of leverage that makes the home run valuable, not the inning part, even after the fact.

Don’t disagree, though I still think there is something to the quality of the opposition perspective.

Guy, I suppose what I was getting at was—if you could hit 9th inning HRs on demand (or pitch a perfect inning), obviously you’d rather save that for the 9th inning if possible. A team that used the perfect closer in the top of the 1st would have a lower w% than a team that used him in the 9th inning with a 3-2 lead. So, the closer must have value beyond 10 runs = 1 win, which is obviously true given that closers win MVP/CY Young Awards and are awarded large contracts despite being only 20 runs above average. So it’s hard to see why a run is a run, even after the fact.

I understand (I think) the concept that in a 3-2 win, all three runs, in a sense, contributed equally to the win. I just don’t agree that we can move individual runs around to different innings and say “see, the score is still the same, therefore a run is a run”. If Ichiro’s 2-run HR occurred in the 4th inning, the game would’ve been played differently. Rivera wouldn’t have pitched the 9th because the Mariners wouldn’t have batted!

Tango, what about this—

Projections - season stats = predictive
WPA - season stats = descriptive

For Ichiro, .858 WPA for his HR - .140 wins for a context-neutral HR = .718 “descriptive” content of his HR.

dackle: perfect.

Mmmm… These are great!  This is great stuff, the discussion of WPA vs WAR/WOBA/etc.

Very interesting that it correlates so well season to season, nearl as well as wOBA.  I say that only because the lists of WAR leaders and WPA leaders on FanGraphs are decidedly different.  (El Hombre being a notable exception.)

WAR includes position, fielding and WPA doesn’t.  And WAR has extra bonus for playing time.

D’oh.  Should’ve realized that myself.  That makes a whole ton more sense.  Thanks, Tango.