The Kazmir conundrumby David Gassko
July 05, 2007
It may well be the most lopsided trade ever made: On July 30, 2004, trailing the Atlanta Braves by seven games in their division with a 49-53 record, the New York Mets decided to pull the trigger to send top prospect Scott Kazmir to Tampa Bay for veteran pitcher Victor Zambrano. The Mets finished the season fourth in their division, 25 games behind the Braves.
It got worse. In two-and-a-half seasons with the Mets, Zambrano accumulated just over 200 innings with a mediocre 4.42 ERA. Kazmir has been all that was expected, posting a 3.73 ERA in 364 innings over that same time period, all while pitching in the much tougher American League. Today, Zambrano has a double-digit ERA pitching for the Toronto Blue Jays while Kazmir, still just 23, is averaging more than a strikeout per inning for the Rays.
Overall, the move was a clear steal for Tampa Bay. But the one criticism that followed Kazmir as he moved up the ranks of the Mets system has followed him to his new home: He doesn’t give you innings. For his mid-90s fastball and hard, un-hittable slider, the spindly Kazmir has never thrown 200 innings in a season.
For his career, he has averaged just 5.8 innings per start; Pedro Martinez, who also packs great stuff into a small frame, has averaged an inning more, 6.8, over his illustrious career. Though Kazmir is not as good as Pedro was in his prime—no one is—the fact remains that in light of Pedro’s endurance, Kazmir’s size is not an excuse for his lack of innings. What is?
Many have posited that the problem is Kazmir’s strikeout rate. Yes, it’s great for a pitcher to strike out more than a batter per inning, but at what cost? It takes at least three pitches to strike out a batter, whereas an out on a ball in-play might take just one. Some have posited that it might be better for Kazmir to hold back a little bit, that all his strikeouts result in high pitch counts and fewer innings.
But to every argument there is a counter-argument. On the one hand, yes, the average strikeout requires more pitches than the average ball in-play. But on the other, a strikeout is a guaranteed out; a ball in-play is not. So first we need to ask, do all those balls in-play actually lower pitch counts?
Let’s first try to answer this question theoretically. Thanks to Tom Tango, we know that the average walk takes 5.5 pitches, the average strikeout is done in 4.8 pitches, and all other plate appearances average around 3.3. From that, we can estimate pitch counts.
For example, a pitcher who averages six strikeouts, three walks, one home run, and 9.6 hits (that corresponds to a .290 batting average on balls in play) per nine innings will throw about 146.2 pitches per game. So what if we keep his rates constant, and vary only the number of strikeouts he gets?
As his strikeout totals go up, the pitcher will give up fewer hits and home runs because batters aren’t putting their bats on as many balls. But will the increase in strikeouts still mean an increase in pitch counts despite the number of hits allowed going down? In graph form, let’s take a look at what happens as we vary the pitcher’s strikeout rate from 0 to 27 strikeouts per game:
Would you look at that? The numbers don’t change at all! If you look very closely, there is actually a slight downward trend, but the fact is that there is only a .15 pitch difference at the extremes, hardly anything to worry about.
Basically, what we’re seeing is that striking out fewer batters has absolutely no effect on pitch counts, at least in theory. But perhaps the facts are different? One possibility is that our assumption that a pitcher’s other rates remain constant no matter his strikeout rate is erroneous. Perhaps if Kazmir let his strikeout rate fall, he would also see a drop in his walk or home run rate or his batting average on balls in play. We can’t know for sure without looking at some hard evidence.
So let’s have a look at the data. I selected every pitcher since 1946 who threw at least 150 innings in consecutive seasons. Some of you might immediately be thinking that introduces selection bias to this study, as pitchers don’t throw 150 innings unless they’ve had a healthy and non-disastrous year, but that’s actually desirable in this case: Kazmir likely would be okay if his strikeout rate fell, so we don’t want to include pitchers who struck out fewer batters due to injuries or loss of effectiveness in our study.
Next, I found every pitcher in that sample whose strikeout rate (adjusted for league average and placed in the context of the 2006 American League, as are all the other statistics) dropped at least 25% from the previous year, which would represent a one standard deviation change for the average pitcher. There were 217 such pitchers in all. Let’s take a look at how their numbers changed when their strikeout rate plummeted:
Year W L G GS ERA H HR BB SO HBP IP One 14 11 35 30 3.96 213 24 64 175 7 218 Two 13 12 34 30 4.42 222 26 63 119 7 209
You can see the peripheral numbers are all pretty similar except for one: Strikeouts. A look at these pitchers’ rate stats will demonstrate that even better:
Year W L G GS ERA BABIP HR/9 BB/9 SO/9 HBP/9 IP One 14 11 35 30 3.96 0.296 0.98 2.62 7.21 0.29 218 Two 13 12 34 30 4.42 0.293 1.12 2.70 5.12 0.30 209
Despite the relatively stable peripherals, the pitchers add almost half-a-run to their ERAs in the second year. Why? Simple: They’re allowing more balls in (and out of) play. Though their batting average on balls in play remains pretty stable, their total number of balls in play rises due to the decreased strikeout rate. And while their rate of home runs per batted ball also does not change much, the pitchers allow .14 more home runs per game because they’re allowing more total batted balls.
In other words, the data matches the theory. Just as we would expect, these pitchers don’t see any change in their rate statistics when they start striking out fewer batters, but they do see a rise in ERA because the drop in strikeout rate means that they’re simply allowing more of everything.
Kazmir’s problem is not his strikeout rate, but his propensity to issue a lot of walks. We may think that the two are intertwined, but the above study says otherwise. Pitchers who drop their strikeout rate do not see a corresponding fall in the number of walks they issue; in fact, their walk rate actually goes up a tick. Kazmir is walking 4.5 batters a game this season, and for an ace, that just isn’t acceptable.
If he could lower his walk rate to, say, the league average, which is where he was last year, Kazmir could probably last almost an inning more per game. For a Devil Rays team that is starved for pitching, that would be oh-so-sweet.
David Gassko is a former consultant to a major league team. He welcomes comments via e-mail.