Will to powerby David Gassko
February 14, 2008
For two decades, it has been accepted sabermetric wisdom that high-strikeout pitchers last longer than do their low-strikeout brethren. As best as I can tell, that tenet was first established by (who else?) Bill James in the 1981 Baseball Abstract. In the '87 edition, James wrote: “If you study the issue... you can’t possibly miss seeing that the strikeout pitchers last a lot longer than the control-type pitchers.”
In the New Historical Baseball Abstract, James writes that “the influence of strikeouts on a pitcher’s future can be compared to the effect of height on a man’s chances of playing in the NBA.... Power pitchers last much longer than finesse pitchers.” James suggests that if you look at pitchers with a similar number of wins at a given age and divide them into high- and low-strikeout groups, the high-K pitchers win more games over the rest of their careers than will their low-strikeout counterparts.
In other words, pitchers with identical performances at or up to a certain age diverge over the rest of their careers depending on how many batters they strike out. And the fact is, if you study the issue as James suggests, that will be true. Using ERA instead of wins, Keith Woolner came to the same conclusion, finding that high-strikeout pitchers last longer and pitch better later in their career than do their low-strikeout cohorts.
The evidence is bulletproof, right? Of course not. The big problem with statistics like wins or ERA—as James himself has often preached—is that they involve a big dose of luck. Sometimes, it is appropriate to assume that the luck cancels out in large groups. In this case, it is not.
We know from DIPS theory that the number of hits a pitcher allows on balls in play is largely out of his control. Overall, a pitcher can suppress the number of hits that he allows by striking out more batters (and thus giving them fewer chances to put the ball in play); but on a seasonal basis, hits allowed numbers can jump around quite a bit. And with hit rate varies ERA, and with ERA, wins and losses.
When we use wins or ERA to establish similar pitchers, we allow luck to creep into our groups. High-strikeout pitchers with high ERAs tend to be unlucky, whereas low-strikeout pitchers with low ERAs tend to be lucky. Another way to put it is that if you took a bunch of pitchers with identical ERAs or win-loss records in one season and grouped them according to whether they struck out many batters or few, the high-strikeout pitchers would have better ERAs and win-loss records not only in the future (James’ hypothesis) but also in the past.
Clearly, while James' study is a good start, it is not faultless. Instead what we need is a measure that's relatively free of luck; one that roughly captures each pitcher’s true talent and allows us to find truly comparable pitchers, some with high strikeout totals, some with low. In this case, we can use FIP, which tells us what a pitcher’s ERA should have been based on his counts of home runs, walks, hit batters and strikeouts.
Let’s take all pitchers who, since 1946, threw at least 500 innings before the age of 26. There are 352 pitchers in all. So that we are comparing apples to apples, we’ll adjust all their numbers to the context of the 2007 American League. If we then sort the pitchers by strikeout rate, we find that everyone in the top third averaged at least 7.5 (adjusted) strikeouts per game, whereas everyone in the bottom third averaged no more than 6.4. The overall average is about 6.6 K/G.
So let’s take those two terciles—the high-strikeout pitchers and the low-strikeout pitchers—and see how they compare. Before the age of 26, the 117 high-strikeout pitchers had thrown an average of 818 innings with an (adjusted) 4.03 FIP. The 117 low-strikeout pitchers, on the other hand, averaged 710 innings before turning 26 and with a much uglier (adjusted) 4.67 FIP.
In other words, we have what James might call “quality leakage.” That is, we’re comparing two quite uneven groups: The high-strikeout pitchers are quite obviously better than the low-strikeout hurlers. We should expect the former group to have better careers the rest of the way, even in percentage terms, because they are better pitchers and because they’re much further away from dropping off a cliff.
What if we try to correct for that quality leakage? What happens if we remove the 50 best pitchers (as measured by FIP) from the high-strikeout side, and the 50 worst pitchers from the low-strikeout side? If we do this, the remaining high-strikeout pitchers average 763 innings before the age of 26 with a 4.36 FIP, while the low-strikeout pitchers average 742 innings with a 4.41 FIP. That’s much better, but there is still some difference between the groups.
But if we remove just two more pitchers from each group—the ones with the most innings pitched before turning 26 of those remaining on the high-strikeout side, and the ones with the fewest innings pitched before the age of 26 of those remaining on the low-strikeout side—the two groups look almost identical: 744 innings with a 4.37 FIP for the high-strikeout pitchers, 749 innings with a 4.41 FIP for the low-strikeout pitchers.
Now we can actually make some comparisons, and with 65 pitchers in each group, we still have a healthy sample to look at. As it happens, we really need to make only one comparison: How did the two groups pitch from age 26 on? If James' hypothesis holds, the high-strikeout pitchers should have significantly better statistics, especially in terms of longevity. Is that the case?
Well, sort of. After turning 26, the high-strikeout pitchers average 1,072 innings with a 4.32 FIP for the rest of their careers; those numbers represent a 44% increase in innings and a 1% decline in FIP. The low-strikeout pitchers average 933 innings through the rest of their careers with a 4.51 FIP—a 25% increase in innings and a 2% increase in FIP. Although the FIP difference looks insignificant, the inning difference certainly does not. Nevertheless, statistically, there is about a 35% shot that the latter difference could have occurred by random chance alone. That’s a pretty big number, and way over the usual statistical threshold of 5%.
So where does that leave us? I don’t know. We certainly haven’t disproven James’ theory; after all, we did find that high-strikeout pitchers last longer than do equally skilled low-strikeout pitchers, even if that finding is not significant in the statistical sense. On the other hand, we certainly could not prove James' theory, either; a 35% probability of something happening just due to pure luck is very high—more than double the odds of correctly guessing how a six-sided die will land (and no one would attribute one correct guess there to skill).
In the end, this is a question that deserves more research. It needs to be poked and prodded from different angles. But it certainly isn’t something that should be taken for granted. For now, it is fair to say only that we know nothing about how strikeout rate affects a pitcher’s aging process.
David Gassko is a former consultant to a major league team. He welcomes comments via e-mail.
<< Return to Article