Post by ericmvan on Jun 17, 2015 13:53:05 GMT -5
(Yeah, a study like this should be at The Hardball Times or someplace ... but it's a half-day's work, quick and dirty, so what I've figured out so far goes here. Some winter I'll redo it in more depth. And this post is Sox 2015-specific.)
If you said wRC+ ... not quite. wRC+ accounts for just 26.8% of the offensive contribution to wins (as measured by Win Probability Added).
Here's the breakdown:
.268 wRC+
.221 Base/Out Situation Clutch hitting. A few years ago at the SABR conference I showed that hitting differently with men on versus the bases empty is something that players do (good or bad) consistently over the course of a career, so you would expect this factor to have a very small but significant year-to-year correlation for clubs. And it does: r = .15, p < .004 with n = 358 (2002-2013 plus the year after). That means that 2.3% of a team's situational hitting can be explained by their previous year's performance. Given the turnover in lineups, I'd guess that about 4-5% of this is real, and the rest is luck or psychological factors without long-range predictive power.
.175 Inning/Score Clutch Hitting. There are some factors that influence this very interestingly, which you'll see below and which I'm omitting here. The year-to-year correlation is .06, which has a 27% chance of being random. The jury is still out on the very interesting question as to whether there's any skill component here, but at the team level we can regard it as all luck.
I'll cut to the chase: so far, the Sox are -0.8 wins in Base/Out Clutch, and -1.9 wins in Inning/Score Clutch including the extra factors, which have actually been good for the Sox (separating them out is a pain). Add them together, and that's at least 2.8 Wins of bad clutch hitting (yeah, rounding!) with essentially no predictive value, on top of the 3 wins of BABIP luck I estimated earlier (which I might repeat and refine at some point soon).
And no, I'm not double-counting. Not only have we hit balls right at people with the game on the line, even when you factor in the number of balls we've hit at people, you still come up with a clutch deficit. Even the balls that have avoided people have been ill-timed.
Now, I agree with everyone that poor clutch performance in the short run sure seems to be psychologically driven, as when teams are pressing. There's probably a small real component there (less than it seems) that you can arguably pin partly on the manager for not getting the guys out of their bad frame of mind. But partly real or not, it has essentially no predictive value. Collective team psyches can change overnight. And you should never, ever make the mistake of thinking that a team with bad "karma" (to use my term for luck + possible real factors) is untalented.
This is 6 wins of stuff with no predictive value at all ("luck" in some sense of the word), and the double catcher injuries essentially add at least 2 (I think it's closer to 3, actually) That's enough to put us in first place, with the Hanley in LF disaster, with the Ortiz and Sandoval vs. LHP disaster, and with the pitchers being unclutch as well, and with an actually mediocre bullpen. Anyone think we need to tear it up and start over?
The bottom line is that there's an immense world of difference between "this team has played like horses***" and "this team is simply not good." If you think this roster (including the 60-day DL) is not good enough to contend next year with the usual set of improvements that any good GM does as a reaction to new information (e.g., Hanley can't play LF), then you must be grossly underestimating just how horses*** the team has played. Reading the game threads, who would ever think that? We all think the team has played like huge mounds of hores***. If you also think the team is perilously short of talent, you are double-counting. Playing like horses*** has no predictive value. If you watch a demonstrably good player play like hores***, you don't change your mind about him being good. (How exactly is Pedroia failing to catch two infield popups in the same game different from Sandoval, always an average defender, being -9 DRS in 65 team games?) If this team were actually short of talent and were playing this badly relative to that talent, they would be historically bad.
Here's a breakdown of offensive wins for the Sox, 2002-2014. ExpWins is based on wRC+ and other factors it excludes, mostly varieties of baserunning. "BaseOut" is base/out situation clutch hitting. "CluTen" is factors that contribute to inning/score clutch; it's probably too low for the Sox for reasons I'll get into in a bit, so some of that negativity probably belongs in the last column, which is Inning/Score Clutch Hitting
This is the second study I've done that found that successful post-season teams tend to have negative Inning/Score Clutch Differentials. That stuff is not predictive, but it's crucial to getting to the post-season. If you're good enough to get into the post-season despite it, you are especially talented.
In the next post, I'll continue with the breakdown of all the other factors, which are really quite interesting.
If you said wRC+ ... not quite. wRC+ accounts for just 26.8% of the offensive contribution to wins (as measured by Win Probability Added).
Here's the breakdown:
.268 wRC+
.221 Base/Out Situation Clutch hitting. A few years ago at the SABR conference I showed that hitting differently with men on versus the bases empty is something that players do (good or bad) consistently over the course of a career, so you would expect this factor to have a very small but significant year-to-year correlation for clubs. And it does: r = .15, p < .004 with n = 358 (2002-2013 plus the year after). That means that 2.3% of a team's situational hitting can be explained by their previous year's performance. Given the turnover in lineups, I'd guess that about 4-5% of this is real, and the rest is luck or psychological factors without long-range predictive power.
.175 Inning/Score Clutch Hitting. There are some factors that influence this very interestingly, which you'll see below and which I'm omitting here. The year-to-year correlation is .06, which has a 27% chance of being random. The jury is still out on the very interesting question as to whether there's any skill component here, but at the team level we can regard it as all luck.
I'll cut to the chase: so far, the Sox are -0.8 wins in Base/Out Clutch, and -1.9 wins in Inning/Score Clutch including the extra factors, which have actually been good for the Sox (separating them out is a pain). Add them together, and that's at least 2.8 Wins of bad clutch hitting (yeah, rounding!) with essentially no predictive value, on top of the 3 wins of BABIP luck I estimated earlier (which I might repeat and refine at some point soon).
And no, I'm not double-counting. Not only have we hit balls right at people with the game on the line, even when you factor in the number of balls we've hit at people, you still come up with a clutch deficit. Even the balls that have avoided people have been ill-timed.
Now, I agree with everyone that poor clutch performance in the short run sure seems to be psychologically driven, as when teams are pressing. There's probably a small real component there (less than it seems) that you can arguably pin partly on the manager for not getting the guys out of their bad frame of mind. But partly real or not, it has essentially no predictive value. Collective team psyches can change overnight. And you should never, ever make the mistake of thinking that a team with bad "karma" (to use my term for luck + possible real factors) is untalented.
This is 6 wins of stuff with no predictive value at all ("luck" in some sense of the word), and the double catcher injuries essentially add at least 2 (I think it's closer to 3, actually) That's enough to put us in first place, with the Hanley in LF disaster, with the Ortiz and Sandoval vs. LHP disaster, and with the pitchers being unclutch as well, and with an actually mediocre bullpen. Anyone think we need to tear it up and start over?
The bottom line is that there's an immense world of difference between "this team has played like horses***" and "this team is simply not good." If you think this roster (including the 60-day DL) is not good enough to contend next year with the usual set of improvements that any good GM does as a reaction to new information (e.g., Hanley can't play LF), then you must be grossly underestimating just how horses*** the team has played. Reading the game threads, who would ever think that? We all think the team has played like huge mounds of hores***. If you also think the team is perilously short of talent, you are double-counting. Playing like horses*** has no predictive value. If you watch a demonstrably good player play like hores***, you don't change your mind about him being good. (How exactly is Pedroia failing to catch two infield popups in the same game different from Sandoval, always an average defender, being -9 DRS in 65 team games?) If this team were actually short of talent and were playing this badly relative to that talent, they would be historically bad.
Here's a breakdown of offensive wins for the Sox, 2002-2014. ExpWins is based on wRC+ and other factors it excludes, mostly varieties of baserunning. "BaseOut" is base/out situation clutch hitting. "CluTen" is factors that contribute to inning/score clutch; it's probably too low for the Sox for reasons I'll get into in a bit, so some of that negativity probably belongs in the last column, which is Inning/Score Clutch Hitting
Year ExpWins BaseOut CluTen InningScore
2002 4.88 5.53 -1.16 -3.15
2003 13.28 -4.13 -.57 .22
2004 7.06 1.49 -1.19 -1.52
2005 7.68 1.82 -.98 3.69
2006 1.33 -.91 -.83 4.23
2007 7.21 -1.40 -1.11 -4.64
2008 7.03 -.18 -1.27 .40
2009 3.62 2.33 -1.43 -.51
2010 6.66 -.68 -.80 .27
2011 9.92 -2.36 -.20 -6.58
2012 -.96 -2.52 -.13 -1.72
2013 11.73 2.69 -1.90 -2.51
2014 -4.90 -.39 -.42 -1.86This is the second study I've done that found that successful post-season teams tend to have negative Inning/Score Clutch Differentials. That stuff is not predictive, but it's crucial to getting to the post-season. If you're good enough to get into the post-season despite it, you are especially talented.
In the next post, I'll continue with the breakdown of all the other factors, which are really quite interesting.
