In last night’s loss to the Diamondbacks, Lucas Duda went 0-4 with a strikeout. It obviously wasn’t a particularly good day at the plate, but it was made worse by the fact that he left seven runners on base. That strikeout came in the first inning with the bases loaded. The Mets as a team are hitting .143 with the bases loaded this season. Does that mean they’re not “clutch”? Can we really measure “clutch”?
The word clutch is incredibly loaded when it comes to evaluating a player and their performance. It’s a hot topic because a player’s relative “clutchness” doesn’t appear to be a repeatable skill and varies from year to year, regardless of how much people think the opposite. There’s sometimes a fundamental flaw in the way we view performance statistics, though, and how we apply them to the players that generated them.
Most stats can (very generally) be grouped into two broader categories: predictors for the value a player is expected to add to his team (based on his skill) and the value a player has provided to his team (what he’s done).* The former category would include statistics such as weighted on-base percentages, strikeout and walk rates, and generally outcomes predicated by the player’s natural ability that can be used, with a large enough sample, to gauge what a player will add to the team moving forward. The latter category would include stats like RBI, where they identify how much value a player has provided their team, in very specific ways, but don’t necessarily indicate particular skill or player quality.
For instance, Player A and Player B could put up almost identical statistics in most of the skill categories, but Player B might have more RBIs than Player A by a decent margin. Some may use that as an indicator that Player B is the better player. However, it could just be that Player B has had more opportunities to drive in runs than Player A (maybe Player B is a number five hitter and Player A is a number two hitter, for example) and has taken advantage of those opportunities. Those RBIs are examples of what the player provided to the team based more on the situation rather than strictly based on his ability, at least when compared to Player A.
With these ideas in mind, we need to understand that most traditional statistics are context-neutral. A single in the top of the fourth in a blowout is much less valuable than a single with a runner on second in the bottom of the eighth in a tie game. The players’ batting average would simply reflect a single in either case, but that does nothing to reflect the vastly different situations in which that player got that single and the differing affects to the outcome of the game. Luckily there is a statistic we can look at to measure this, Win Probability Added (WPA).
WPA calculates the affect a player’s performance has on his team’s Win Expectancy, which takes into account the score, number of men on base, the inning, number of outs, etc., on a per-play basis. In other words, the context of the situation is taken into account. Basically, the team has x chance at winning a game before a player comes to bat and y after that at-bat, depending on what said player did at the plate. The outcome increases or decreases the team’s chance to win and, thus, either increases or decreases the player’s WPA.
WPA is a counting statistic, accumulating in a similar way to Wins Above Replacement (WAR), and one point of WPA is equivalent to one win. This means that a player’s contribution to his team can’t be fully evaluated until the end of the season, but we can still track it as the season progresses. A WPA of 1 over a given season is considered average on this scale. So far this season, Duda has a WPA of 0.20. That means that he’s actually helped the team win more than he’s hurt it. Juan Lagares leads the team with a 0.80.
Does this tell us that Lagares has been more “clutch” than Duda, though? Well, not really. Clutch is a pretty hard thing to measure, and even the Clutch stat available at sites like FanGraphs don’t really get to what most people think of as “clutch.”** This simply tells us that Lagares has helped the team at the plate more than he’s hurt them and more so than Duda.
The Mets as a team have a WPA of 0.17, which is actually good enough for sixth in the National League (NL) and ahead of teams like the Braves, Nationals, and Cardinals. What does this tell us? It doesn’t really tell us anything, yet, since again WPA is a counting stat. The Mets were 11th in the NL in WPA in 2013, while the Cardinals and Braves topped the list. From 2011 to 2013, the team accumulated 2.57 WPA, good enough for 9th in the NL. Over that same span the top three teams in the NL are the Cardinals (25.39), the Giants (15.23), and the Braves (13.92).
It’s not really entirely accurate to classify Duda or the Mets as clutch or not clutch, since again we’re looking at statistics that only show what a player did not predictors of what they will do. We can certainly say they haven’t performed very well in high-leverage situations over the last few seasons, though. This brings us to the obvious: good players will generally perform well while bad players will not, regardless of the situation. That translates to team wins. The Mets have not been a particularly good bunch recently, in case you hadn’t noticed. So it’s no surprise that they haven’t executed when it matters most.
*Of course this is an incredibly simplistic way to break them down, but a good way to differentiate performance predictors from “story telling” stats.
**The Clutch stat measures a hitter’s performance in high-leverage situations, but it does so using a baseline created from that own player’s performance. This means that a player who constantly performs at a high level, regardless of situation, may not appear to be clutch.