Dan Uggla and the “clutch” question


This has been an overall interesting season for Dan Uggla. He is having his best season offensively and overall, with a career-high .380 wOBA and 4.4 WAR with a month left in the season. While he is unlikely to eclipse the 5.0 WAR mark, a 4.5+ WAR year is not out of the question and would be quite commendable for the Marlins slugger. It would also fuel the desire of the team to sign Uggla for a couple more seasons after this one, despite concerns about his age and skillset typically considered ripe for decline.

One of the interesting aspects of this season for Uggla is his unusual “clutch” performance in high leverage situations. Before this season, Uggla had 253 PA of what was considered “high-leverage” situations in his career (likely situations with a Leverage Index greater than at least 2.0). Before the season, he had a wOBA of .301 in those high leverage PA. But this season he has exploded with everything on the line, to the tune of a .456 wOBA in 50 high-leverage PA.

Have all the questions of clutch performance fallen by the wayside with this performance? I sure hope not.

Updating Uggla’s clutch

Before the season, I projected Uggla’s clutch/non-clutch split based on regression values found in The Book. I got that Uggla was projected to perform at 0.995 times the level of performance you would expect in any given situation. Adding in the meager .456 performance from this season brings his career high-leverage wOBA to .324. Compared to his career wOBA of .359, that’s a ratio of .903. However, after regression with 7600 (!) PA of the mean, you get a projected ratio of 0.996, barely any different from his projected rate before the season began.

That .996 ratio, compared to his projected future wOBA of .363, yields an expected clutch performance of .362 wOBA, or essentially no difference. In other words, just because Uggla has performed really well in the clutch this season does not mean he will perform significantly better in the future.

What about the “Clutch” score?

Given the current performance of Uggla in high-leverage situations, I was surprised to see that his clutch score as measured by FanGraphs is actually pretty low. Right now, Uggla’s Clutch is at 0.92 wins below his context-neutral performance, an oddity given his great performance. I was wondering why this was the case, so I dug into FanGraphs’ explanation about the stat and found this article by friend of the Maniac Eric Seidman. Here’s the official way to calculate Clutch:

"How it’s calculated: WPA / pLI – WPA/LI"

This stat considers all plate appearances, not just PA of a certain leverage index. This means that, despite being called clutch, it also considers low leverage PA as well. For high leverage PA, the relationship between the terms in Clutch (essentially WPA and WPA/LI, or context-neutral WPA) makes sense (higher if you perform better, lower if you perform worse), but if I understand correctly, it works in the opposite fashion on low leverage situations. I will have to dig further to find out, but it does not intuitively make sense.