Giancarlo Stanton and the question of clutch

One of the things that has been brought up about Giancarlo Stanton's game that he struggles "in the clutch." Fish Stripes reader Sandman13 has brought this up several times. Here's one example of the discussion:

Unbelievable

0-14

That’s Stanton’s success rate with two outs and runners in scoring position. Those are the situations that define a hitter.

Stanton hits the ball harder than anyone else? No, he hits in less than anyone else when it matters.

Now, ignore the fact that 14 plate appearances, with no scouting information to back that up, you cannot differentiate any difference in talent versus random variation. I am happy to discuss the overwhelming point about the always-contentious question of "clutch." Because the truth is that Stanton has not hit as well when the game is most on the line as he does when the situation is less important!

But first, let's talk some vocabulary that is going to be used in this article. Definitions from FanGraphs' Library.

Win Probability Added: This is a statistic that we use on our Game Threads to discuss which players were present for the most high-impact plays in the game. Which hits were most important? How did the game change with each event? If we attribute the amount of win percentage a game changes from one event to another to the pitcher and hitter involved, we can add up Win Probability Added for individual players.

What does this tell us? It is entirely context-dependent, meaning that important situations will matter more than less-important chances. A homer in the ninth inning tied matters more in WPA than a homer in the ninth inning up six runs. It doesn't mean that that homer doesn't matter, but rather that it matters less. But it should be noted from the FanGraphs article:

It doesn’t tell you how well a player performed, it tells you how important their performance was.

Leverage Index: Leverage Index (LI) is a measure of how important a plate appearance is. If the average plate appearance has an LI of 1.0, a greater LI means more importance and vice versa. So that tied game in the ninth inning might have an LI of 2.0, meaning it's twice as important. With runners on base, it could be 3.0 or 4.0, three or four times as important. That six-run lead in the ninth? That could have an LI of 0.2, making it five times less important than an average plate appearance.

FanGraphs has this to say about their official designations for low and high leverage.

An LI of 1 is average. We officially classify anything below 0.85 as low leverage and anything above 2.0 as high leverage.

The above is shorthand, but it more or less an accurate representation of what LI means. We'll be using these two things to discuss Stanton.

FanGraphs also has a "Clutch" statistic, which measures how much better or worse a player has performed in high-leverage stituations versus his average performance.

Context:

The majority of players in the league end up with Clutch scores between 1 and -1, with zero being neutral, positive scores being "clutch", and negative scores being "choke". Only a few players each year are lucky enough (or unlucky enough) to have extreme Clutch scores.

Rating	Clutch
Excellent	2.0
Great	1.0
Above Average	0.5
Average	0.0
Below Average	-0.5
Poor	-1.0
Awful	-2.0

Basically, the stat shows how many "wins" a player has contributed more or less than his context-neutral performance. Note that in that above table, those clutch scores are rated for a single season.

Stanton: Low versus High Leverage

Let's take a look at Stanton and the nine other best hitters in baseball since 2013 and their performance in medium leverage and high leverage states. In other words, how do Stanton and the others play in an average situation versus the most important, game-impact situations, "in the clutch" so to speak?

It is clear that the difference between Stanton's batting lines in medium-leverage states and high-leverage states is one of the biggest downward differences among these elite hitters. That can't be good! And that's verifiable truth that happened over the last few seasons, encompassing 330 high-leverage plate appearances.

This difference in performance is the reason why Stanton has been considered "unclutch" in terms of FanGraphs' clutch statistic. Here are those same ten players presented in order of best batting lines overall since 2013.

Stanton's -2.3 clutch wins implies that his total WPA was about 2.3 wins better than his performance without leverage included. Since WPA is all about context, taking context into account made Stanton look like a worse hitter, thus the negative clutch score in the last three-plus seasons.

You will also note that almost every player on this list has a worse performance in "clutch" situations than in normal states. In other words, most of these great players performed worse overall in clutch situations than they normally do. Only a few did better. Does that make them bad players? If you have the confidence to tell me that Mike Trout is bad because of this three-year sample of data, you are really asking for it.

But maybe you take issue with the fact that Stanton's performance at the plate was so much worse in the clutch. After all, that difference in the first table is a nearly 20-point drop in performance from one type of state to another. But remember that this difference is being taken from a comparison to the same player. Stanton owns a three-year batting line in clutch situations of .248/.379.485. No Marlins player has had a better overall batting line in that same time period. Even if you thought Stanton's true skill level was at his clutch performance, that would still be better than any Marlins hitter on hand, as well as almost one-fifth of all of baseball with at least 200 high leverage plate appearances. The point here is that, when compared to himself, Stanton has done poorly, but when compared to the world, Stanton is still well above average. That batting line is nothing to scoff at.

Also, you will note that Stanton's WPA is still pretty comparable to these greats. Stanton owns the 10th-highest WPA in all of baseball. If WPA, a stat that is entirely context-dependent and rewards you for better play in crucial situations, points to Stanton being a top performer, should his clutch performance even factor in?

And this does not even get to the crux of the matter: clutch skill is nearly impossible to discern in a player. From FanGraphs:

Clutch does a good job of describing the past, but it does very little towards predicting the future. Simply because one player was clutch at one point does not mean they will continue to perform well in high-leverage situations (and vice versa). Very few players have the ability to be consistently clutch over the course of their careers, and choking in one season does not beget the same in the future.

This does not mean that clutch skill does not exist. Research done by the authors of The Book: Playing the Percentages in Baseball noted that there was a slight on-base skill increase.

Anyway, as for actually finding a clutch skill, Andy did in fact find it, and the results are published in The Book.  On p.103:

Batters perform slightly differently when under pressure. About one in six players increases his inherent "OBP skill" by eight points or more in high-pressure situations; a comparable number of players decreases it by eight points or more.

But as Andy concludes later on p.108:

For all practical purposes, a player can be expected to hit equally well in the clutch as he would be expected to do in an ordinary situation.

The problem is that it takes a lot of plate appearances to find statistically significant differences, a lot more than things like platoon advantage. Thus, it is hard to discern whether any individual player has a real skill in it or not, even if we know it exists.

The Stanton clutch question is not really much of a question at all. He plays extremely well, he has played less well in clutch situations in the past, and we have no idea whether or not any of that is real or simply a blip in the data. Without true knowledge of the workings of a player, something that is more garnered by scouts watching field-level on a daily basis or having advanced biometric data on hand, we really cannot discern what any of these numbers mean, if anything at all.