Settling the Cantu-Uggla clutch debate


OK, so “settling” may be a very strong word. Instead, perhaps I am quantitatively saying what I’ve been qualitatively claiming for months now about Jorge Cantu vs. Dan Uggla. With the induction of FanGraphs’ splits, we now have access to situational wOBA that we had to calculate before. They’ve made it much easier to deal with this question, and here I’ll attempt to do so. After the jump, you’ll see me show my work with regards to this issue, with the help of FG’s splits and a little excerpt from The Book.

The Methodology

According to The Book, there is indeed a clutch talent to be had, but it is so small as to be insignificant. The league average performance in the clutch is no different than when compared to normal situations, but some players can exhibit a (slight) talent. For a quick projection of this so-called “talent,” you can regress a career split between clutch and non-clutch performances by about 7600 clutch PA! I’m not sure anyone could garner that many clutch PA in a career, and that would only get you to 50% regression to the mean.

Anyway, how can we apply this to Cantu and Uggla? Well, armed with that regression to the mean component, we can check out the split difference for Cantu and Uggla for their respective careers and regress accordingly. For clutch and non-clutch PA, I’m going to consider the career splits for “high leverage” and the other PA. Let’s start with Uggla. Over the course of Uggla’s career, he has been a part of 251 clutch/high leverage PA that have not ended as intentional walks, and he has been awful in those situations. Confirming most Marlins fans’ beliefs, Uggla has performed to the tune of a miserable .301 wOBA in the clutch, with a batting line of .232/.339/.370. In all other situations, Uggla has done fairly well, with a career value of .357 in PA outside of this high leverage category.

To calculate the projected split, we start off with the ridiculous career split. The clutch-nonclutch split for Uggla in his career is 84.3% or 0.843 ratio. The average ratio is 1 (no difference). So using the 251 PA of observed clutch performance and regressing it to 7600 PA of the average and you get a projected ratio split of 0.995 clutch to non-clutch PA. Assuming an 2010 average distribution of clutch and non-clutch PA equal to Uggla’s career distribution, you get an expected 91% of PA of the non-clutch variety and 9% in the clutch variety. Using CHONE’s projected wOBA of .355 for the season, you would get a projected wOBA of .353 in the clutch for Uggla, a difference of 0.002 wOBA points.

Doing the same for Cantu gets you essentially an even split (he had a career clutch/non-clutch wOBA ration of 1.009, which regressed to 1.00032 with 7600 PA of the mean). In other words, according to CHONE’s projection of a .338 wOBA, we’d expect Cantu to perform at a .338 wOBA in the clutch as well, meaning that Uggla is projected at the higher clutch level.

Now, how many runs would that 0.002 wOBA difference cost the Marlins? Just using the convention of wOBA would give you the answer of -0.1 runs, or essentially nothing. But, to be fair, I think we should consider the fact that Uggla would be facing higher leverage situations in the clutch situations, and adjust the leverage accordingly. Assuming Uggla sees an average leverage of 1.0 for all of his PA and an average leverage of 2.5 in the high leverage PA, we would then have to multiply those runs by the ratio of the difference. That would be a value of -0.25 runs, still insignificant. Consider that if we had just taken the career split, that difference including those leverages would have amounted to about nine runs taken away in 60 PA. That’s a big difference.

Now of course, I don’t really know what’s going to happen. Maybe Uggla will struggle again, as he has in three of his four seasons. Or maybe Uggla will hit at a .370 wOBA clip in the clutch like he did in 2008. Who knows? That’s why we watch the games, I guess. Let’s hope both of these guys deliver when it counts the most.