One of the tried-and-true adages of the sabermetric (baseball research) community has always been that clutch hitting simply doesn’t exist. For years, statheads have tried to figure out if any player has a significant clutch impact vs. other players, and pretty much no one managed to prove it.
Times have changed, and so have the stats. Today, we have win probability-based stats, which can serve to reduce the entire contribution of a player to one single statistic: win probability added (or lost). We call this WPA. While many of the relied-upon statistics show great predictive value (OPS, strikeout and walk rates, among others), they don’t often tell the whole story. While WPA still has some bugs to work out, it comes the closest to doing just that.
WPA has been around for a couple of years now in various forms, but only now are people figuring out good ways to separate the “clutch” component from this performance. The latest is through a WPA number weighted by Leverage Index (a measure of situational importance). In Monday’s post, I was calling that LWP, but Fangraphs has chosen to call it just WPA/LI, so I will do the same.
By adjusting WPA by Leverage, we can accomplish the necessary step toward measuring clutch hitting, which is removing the context of a plate appearance from the calculation. WPA/LI measures a player’s win probability performance without regard to the score, inning, or base-out situation. The difference, then, between WPA/LI and WPA is what we call Clutch.
All of the categories preceded by a lower-case ‘p’ are per-plate-appearance numbers for Leverage Index, WPA, WPA/LI, and Clutch. The latter three are multilplied by 100 to make the number more readable.
Like I said, there are some bugs to work out of the WPA numbers, so this is not a perfect measure, but it’s already brought up some interesting results in the early season. I couldn’t wait for a weekly update to post the results.
Here are the Braves’ WPA stats, along with a few other relevant stats, through April 11:
So, are the Braves really 4.5 wins better than they would be with normal clutch play, through only 8 games? My best answer right now is yes. (Remember that .500 in WPA is one win, and the total of hitting and pitching “clutch” is 2.26.)
How is this even possible? The Braves are 7-1, so would they really be 3-5 without their clutch play? I’m not totally convinced, but my instinct says that is right. Pythagorean-wise, the Braves should have won 5 games, so they’re obviously playing at least a little over their heads, but they’ve been unstoppable in crucial situations.
Brian McCann and Bob Wickman have been the kings of Clutch so far, but for two very different reasons. First, let’s tackle Wickman. As the Braves’ ace reliever, Wickman’s appearances have come in situations over two times the league average in Leverage Index (2.13). In a sense, Bobby Cox is inflating his clutch numbers (as long as he’s doing well) because he uses him only in tight situations. On a per PA basis, Wickman’s leverage adjusted WPA (or WPA/LI) is 0.43 (or really .0043), which is still very good, but behind the performances of Hudson, James, and Soriano at this point.
McCann, on the other hand, is not used exclusively in clutch situations like Wickman is. His pLI of 1.07 shows that he’s only seeing about 7% more important situations than the league average. The Braves’ pLI as a team, though, is 1.11, so that says even more about what McCann has done through eight games. His WPA/LI is just .219, which puts him well below Renteria (.480) and in a group with Diaz (.268) and Francoeur (.205) as the Braves’ best performers regardless of game situation. (McCann is, however, the team leader in OPS.) His Clutch number, on the other hand, is .566, which mean’s he’s had well over a full win of clutch production thus far. No one else is even close on the Braves, and he actually leads the major leagues.
Looking at McCann’s plate appearances, it’s easy to see why this is so. His biggest homers and a couple of doubles have come in very tight situations, while he has made a majority of his outs in less important situations. So far, the kid really has been Superman. I don’t know if it will continue, but it sure will be fun to watch.
As for everyone else, Andruw Jones (-.255) and Francoeur (-.218) have been anti-clutch, while Chipper Jones (+.224) and Scott Thorman (+.197) have been somewhat clutch. Below Wickman (+.510), a mess of pitchers could be considered clutch so far: McBride (+.291), Davies (+.268), Soriano (+.268), Smoltz (+.244), and James (+.162). Pitching in a lot of close games will tend to cause this, but Soriano and Smoltz are the only ones in this group with a pLI over 1.00.
Of course, there’s still nothing out there that suggests some players are significantly more clutch than others. McCann was average in Clutch for the past two years, so his WPA improvement may not be for real. For now, let’s hope the Braves will continue being extremely “clutch” throughout the year. It’s so much fun to say they’re 7-1 and leading the division once again.
2 thoughts on “Are the Braves really this Clutch?”
could it be that the past couple of years McCann’s average clutch numbers were just the result of not near as many “close games” as there have been since the start of this season? I don’t totally agree with the concept of the Pyth. Record, but I don’t totally disagree with it either.
6-1 (this was commented before anybody has scored in today’s game) seems a bit too good for our team. Either way, I don’t expect to see extremely close games continue throughout the rest of the season. They’ll let up, hopefully in our favor.
Yeah, I’m hoping the offense will be a little more consistent and be just as good as last year (if not better). I looked at McCann’s pLI for the past two seasons, which would measure whether or not he’s appeared in more clutch situations than the average player (which would be 1.00).
In 2005, he had a pLI of 0.99 and Clutch of -.15, which would be pretty reasonable. In 2006, it was 1.06 – a little high, but not unreasonable – and his Clutch was +.18, so he may have benefited very very slightly from that above average pLI. Through Wednesday’s games, his pLI was a fairly normal 1.07 (especially for the small sample sizes this time of year), and his Clutch was +.57, which for 8 games is simply ridiculous.
I think your general point there is very true, though, and it makes me wonder how well a player’s Clutch rating usually correlates with his pLI. It seems true that the higher the pLI, the more a player’s Clutch is affected, both for good and for bad.