This is prompted by a post by Tom Tango (aka tangotiger) titled Best and Worst Clutch Hitters of the Retrosheet era .
Tom has a clutch stat based on WPA or "win probability added." The idea there is that every plate appearance by a hitter either increases or decreases his team's probability of winning. A HR with the score tied in the bottom of the 9th has more impact than one in the first inning with the score 10-0.
But Tom adjusts this by how often a hitter gets to hit in "high leverage" situations. Then that it is compared to what his WPA would be if he always hit in average leverage situations. I hope I got that right. But, of course, Tom explains it much better. That stat ends up telling us how many more games a player's team wins (or loses) because he hits better or worse in high leverage situations than he does overall.
Nellie Fox is #1 with +13.4 wins since 1950. That is, by hitting better than he normally did in high leverage situations, he added 13.4 wins to his teams over his whole career. Sammy Sosa was last with -16.8 wins. That is, he hit worse in high leverage situations than he normally did and this cost his teams 16.8 wins over the course of his career. These two hitters maybe could not be more different and they may be good illustrations of what is going on with this clutch stat.
So let's call Tom's stat Clutch. That's what it is called at Baseball Reference. I took all the right-handed batters and left-handed batters since 1950 who had 4000+ PAs (653 players). Then I divided their Clutch stat by their PAs. I did the same thing for HRs and strikeouts. The I ran a regression with Clutch/PA being the dependent variable and HR/PA and SO/PA being the independent variables. I also added a dummy variable for being a righty (1 for righties and 0 for lefties).
Here is the regression equation
Clutch/PA = 0.0007 - .00025*Righty - .0169*HR/PA - .00157*SO/PA
All three variables seem to be significant. Here are the t-values:
Righty -6.31
HR/PA -10.49
SO/PA -3.06
R-squared is .314 (meaning that 31.4% of the variation in Clutch/PA across players is explained by the equation) and the standard error per 700 PAs is .33.
Mutltiplying -.00025*700 gives us -.172 (assuming 700 PAs is a full season). So simply being a righty means you will have a negative Clutch rating of -.172, meaning you will cost your team .172 wins. This could be because righties can't use the hole at first base with a runner on as well as lefties. When a runner is on first, it makes for a slightly higher leverage situation. Also, righties might have to face right-handed pitchers more often in high leverage situations than lefties face left-handed pitchers.
To see the impact of HRs and SOs, I found the standard deviation of HR/PA and SO/PA and then checked to see how much Clutch/PA would change with a one standard deviation increase in both stats. Here they are
HR/PA: .014
SO/PA: .0449
The coefficient on HR/PA was -.0169. That times .014 = -0.00024. But that times 700 PAs is about -.166. So being one standard deviation above average in HR/PA costs your team .166 wins per season. Maybe HR hitters cannot adapt well in high leverage situations since they generally just swing for the fences. But that is just a guess.
Something similar could be going on for guys who strikeout alot. The coefficient on SO/PA was -.00157. That times .0449 = -0.00007. That times 700 = -.049. So increasing your strikeout rate by one standard deviation costs your team .049 wins per season. Maybe guys who don't strikeout alot have better bat control and they can hit the ball the hole at first base better than average or they can adapt to the situation better.
Let's look at how all this affects Nellie Fox. He was a lefty, so he does not get the righty penalty. His career HR/PA = .003488. The average for all the players in the sample was .0268. So he was .0233 below that. To see the effect for the whole season, we multiply that first by -.0169, the coefficient on HR/PA from the regression equation and then times 700. This gives us -.0233*-.0169*700 = .276. So his lack of power added .276 wins to his teams each year.
What about for his entire career. He had 10,035 career PAs or 14.33 seasons. With 14.33*.276 = 3.96, Fox gets 3.96 clutch wins for his whole career just due to his lack of power.
For SO, Fox had a career rate of .0206. The average was .133. So he was .112 below that. Let's multiply that by -.00157 and then 700 to get .122 (-.00157 was the coefficient on SO/PA). It amounts to -.112*-.00157*700 = .122. So his ability to not strike out gave his teams .122 clutch wins per season. For his career that would be 1.76 Clutch wins. Then 3.96 + 1.76 = 5.72. Just by being a low HR, low SO guy added 5.72 clutch wins. That is nearly half his total.
For Sosa, we have a HR/PA rate of .06154 and a SO/PA rate of .233. Doing the same exercise as I did above for Fox has him with the following "clutch losses" per season due to his high HR rate and high SO rate:
HR/PA = .41
SO/PA = .11
Sosa had 9,986 career PAs or 14.14 seasons. His HR hitting cost him 5.81 clutch wins and his striking out cost him 1.54. And being a righty cost him 2.43 wins (14.14*.172 = 2.43). The .172 was how many wins a righty lost per year, as explained above. Then 5.81 + 1.54 + 2.43 = 9.78. That is more than half of his clutch losses.
All of this, is, of course, an approximation. The regression is not perfect, since the r-squared was only .314. But the variables all were significant and the F-stat was 98 (that is significant and it means that the 3 variables together probably explain some part of the dependent variable).
So Tom Tango's clutch stat is great in terms of what clutch stats should do but it may have some biases. But those biases might be ones teams should care about since HR hitting ability and SO avoidance ability are identifiable traits.
******************************************
I did a very different kind of study several years ago called Do Power Hitters Choke in the Clutch?. I have a link to a similar study by Andrew Dolphin. In this other study it did not look like they did choke. Also, here are some other comments I made at the tangotiger link:
I happened to have a list of players with 6000+ PAs from 1987-2001 with their OPS in close and late situations (CL) and their OPS in non-CL situations. I took the ratio of CL/nonCL. Tino Martinez did the best, with 1.095, meaning that his OPS in CL situations was 9.5% higher than nonCL. The correlation between CL OPS/nonCL OPS and SO/PA is -.364. So it looks like guys who strikeout alot have a little harder time doing well in the clutch
Also, if you go to the rankings, you can see that 10 of the 12 best players in maintaining their OPS in the CL were lefties or switch hitters
http://cyrilmorong.com/clutch.htm
And it looks like 8 of the bottom twelve are righties
Subscribe to:
Post Comments (Atom)
No comments:
Post a Comment