In my first post on clutch hitting I made a simple point--clutch situations exist and sometimes hitters perform in those situations, sometimes they don't. While there is a statistically significant difference in batting average in clutch situations vs. the bases empty, it's a big leap from statistical significance to meaningful difference. Yesterday I focused on the differences in batting average based on base situation, but today I'll do something completely different.

Which I'm pretty sure will show completely different results.

**THE LEVERAGE INDEX (LI)**

I'll quote directly from Baseball-Reference.com (B-R):

Within a game, there are plays that are more pivotal than others. We attempt to quantify these plays with a stat called leverage index (LI). LI looks at the possible changes in win probability in a give situation and situations where dramatic swings in win probability are possible (runner on second late in a tie game) have higher LI's than situations where there can be no large change in win probability (late innings of a 12-run blowout).

The stat is normalized so that on average the leverage is 1.00. In tense situations, the leverage is higher than 1.00 (up to about 10) and in low-tension situations the leverage is between 0 and 1.0.

What this doesn't state is that it was developed by Tom Tango, the

*nom de baseball*for person/persons who create extremely advanced baseball metrics that I can't calculate but can interpret. I've seen the LI for years but it never meant much to me since everyone was grouped right around 1.0--if everyone is around the same number, what deep meaning can it have?
That was until I started playing around with individual plate appearances, possible if you shell out the unheard-of sum of $36 to subscribe to the B-R Play Index feature--once I did that, well, I felt like Saul at Damascus when the scales fell from his eyes. Everything I had done was based on aggregate yearly data, which has value, but individual plate appearances, well, now I could separate hits from HITS.

B-R uses the following definitions:

LI<.7 Low Leverage (i.e., low pressure)

.7<LI<1.5 Medium Leverage

LI>1.5 High Leverage

Here are some recent examples using

**Albert Pujols**plate appearances:Contrary to the definition above, there ARE LI situations below zero and as would be expected, describe situations in which Pujol's performance was generally irrelevant. The last example is Pujol's highest LI plate appearance in his career (through 2012) and is an excellent illustration of what the LI can explain--the Cardinals were down one run with two outs in the bottom of the 9th with the bases loaded--if this isn't a clutch situation, nothing is. The LI does not take into account (at least to my knowledge--I could be wrong) the fact this plate appearance was very early in the season, giving the Cardinals plenty of time to overcome the fact that he didn't win the game, but it's clear--using the LI can help us realize that not all situations are equal, which is the premise that I essentially set up yesterday and certainly espoused in my FanGraph piece.

Using B-R's LI figures, this is how many plate appearances Pujols has had (through June 27th, 2013) at the different leverage levels:

Low 3,937 (46.6%)

Medium 2,926 (34.6%)

High 1,589 (18.8%)

Total 8,452

I could have sliced the data using these guidelines, but I chose to divide the plate appearances into three roughly equal groups of plate appearances, and after some work made my own definitions for the LI:

<.57 Low Leverage

.57<LI<1.12 Medium Leverage

LI>1.12 High Leverage

I doubt that changing these definitions greatly affects the results and was more interested in creating equivalent sample sizes. I used only one data set (I had to create five different Excel spreadsheets for all this data) to determine the values, but I'll assume that using over 900,000 data points gave representative results:

I used a couple of tables to help break down the relative success near the end of yesterday's post, and I'll use the same tables today using LI. The first table breaks down the 837 players by career plate appearances:

Something very interesting has happened--the difference in batting average are much smaller than they were using my previous criteria. In fact, the difference between the medium and low leverage plate appearances are NOT statistically significant, .271 in the low leverage situations and .272 in the medium. High leverage situations have a batting average of .278, which does maintain statistical significance.

This table breaks down the players by All-Star appearances:

Much as in yesterday's post, the better the player the better the results, totally intuitive and what we would expect to see. This last table breaks down the players by career batting average:

This explains clutch hitting about as well as anything I've written over the past two posts. If we use batting average increase as a measure of clutch performance, it's obvious the largest improvement occurs at the lower end of the scale where there's the greatest potential for an increase. Likewise, players at the high end aren't necessarily excellent clutch hitters as much as excellent hitters period. Players with an aggregate batting average of .310 are hard-pressed to improve on that.

In the end, looking for improvements in batting average as a measure of clutch performance probably falls short--there are natural upward limits in batting average. For example, the person with the highest low leverage batting average is

**Wade Boggs**at .334, the lowest**Johnnie LeMaster**at .200 (the lowest by 16 points--ouch!), a range of 134 points. I won't even show the table it's so useless--the players with the greatest percent increase in batting average from low to medium index are all those with low batting averages and certainly not the players managers keenly desire to bat when the game is on the line.
Two last charts--I started off my FG piece with a scatter graph of player's batting average in low and medium leverage situations:

Players want to be on the ABOVE the line of this chart, meaning they hit better in medium leverage situations than low. It's pretty obvious that the players are all over the chart, implying that sometimes they performed, sometimes they didn't. It's also clear that hitters are what they are--.220 hitters may be able to bat .240 in clutch situations, but not .300. It's an improvement, but does it really make a difference? Remember from yesterday that even at the extreme, a 20% increase in batting average from a low-hitting player translates into around 5-6 hits.

This last chart compares low and high leverage situations:

It's natural to break data down and search for deeper meaning, and unfortunately life is rarely so simple that we can look at one set of circumstances and make concrete inferences. This doesn't prevent people from doing this on a daily basis, but in the case of clutch hitting, I really wish I would stop hearing:

"Player X sure knows how to come through in the clutch"

And begin hearing:

"Player X is our best hitter, and we really need a hit here--he's our best hope"

Sometimes Player X will come through, sometimes not. Albert Pujols is one of the best hitters in recent memory and destined for Hall of Fame induction, but even HE fails 70% of the time in clutch situations. However, I can state for certain that he'll fail LESS often than most, not because he's a great clutch hitter, but a great hitter. You'll find this last table interesting--it shows how Hall of Famers have performed by leverage situation:

**Rod Carew**had a 15% lower batting average in medium leverage situations, all the way "down" to .282--hard to call this failure. There ARE clutch situations, there ARE NOT clutch hitters, just hitters that occasionally perform in those clutch situations. The Hall of Famers above didn't have outstanding clutch numbers but were simply the best of their generation at baseball. The idea of clutch hitting will never go away, I just wish it would. In my next post on this subject, I'll do some case studies and show some individual players. And don't forget--I'm willing to share the data on these 800+ players so you can make your own comparisons--just hit the email link at the right and ask.

## No comments :

## Post a Comment