The Art of Clutch: Part II

Earlier this week I posted the method to figuring out what I think is a fairly accurate way of calculating how clutch a player is. Today I am going to post some of the results I obtained from the formula and player statistics for the last three years (2007-2009). I will be analyzing this year's statistics once the season ends.

The requirements for inclusion in the study is that the hitter must qualify for batting title (502 PAs). Each year around 150 players qualified, for an overall sample of 464 hitters (that is, 464 seasons worth of hitting, some hitters qualified twice and some even three times). As I mentioned in the last post, out of the 464 results, there was only one negative number and four numbers greater than 2 (none greater than 2.2). This means that but for one outlier, every player drives in more runs in the clutch situation (2 outs and RISP) than every other situation.

So enough with the explanations, here are the results. The graph for each year follows the best and worst for each year. (Note: the graphs don't show every player, just every third or fourth player. The blue points represent the players' clutch factors and the red points represent the players' player clutch).

In 2007, there were 162 players who qualified for the batting title (and thus this study), and the league average "player clutch" was 0.210 (the average player drove in approximately .210 RBIs more in clutch situations per PA than all situations). As a reminder, each player's individual player clutch was then divided by the league clutch (.210) to figure out the clutch factor for each player. The clutch factor represents how a player performed in the clutch compared to the league. A ClF of 1.0 means that the player was just average in his increase in performance. A number of 2 means that the player increased his performance twice as much as the league average. Finally, a number of .250 means that a player was 4 times worse than the league average in increasing his performance (but still did increase his performance, only numbers less than 0 mean the players performed worse in the clutch).

The most clutch player (MCP) in 2007 was Jim Thome of the White Sox, by just a hair over Brad Hawpe of the Rockies. Thome had a clutch factor of 2.185, due to his crazy clutch RBI/PA of .61 (22 RBIs in 42 PAs with 6 IBBs). He is actually the "King of Clutch", with the highest ClF out of any player in the three years. Hawpe finished with a clutch factor of 2.184, with a clutch RBI/PA of .59 (49 RBIs, 89 PAs, 6 IBBs). Other top finishers were Ichiro (2.016 ClF), Ryan Howard (1.983 ClF), and Pudge Rodriguez (1.879 ClF). Interestingly enough, both in 2007 and 2008 Howard had exactly 111 PAs in clutch situations, which was the most PA by any player in the three year period (and his 52 clutch RBIs in 2007 rank as the most by any player in the three years).  

On the other side of the coin, in 2007 the least clutch player (LCP) was Dan Uggla of the Marlins. He had 88 RBIs in 728 PAs (with no IBBs) in the season, and only 12 RBIs in 93 PAs in the clutch. This led to a player clutch of only 0.009 (he had a RBI/PA of 0.12088 for the season, and only 0.12903 for clutch situations), and a clutch factor of 0.044. The second worst player was Rickie Weeks of the Brewers, with a ClF of 0.062, as he only had 5 RBIs in 63 PAs in the clutch. Other LCPs were Jason Varitek (0.233 ClF), Jeff Kent (0.275 ClF), Michael Cuddyer (0.280 ClF), and Vernon Wells of the Jays with a ClF of just 0.299.

In 2008, there were 147 players that qualified for the batting title, and the league average player clutch was 0.201. The MCP was Josh Hamilton of the Rangers, with a clutch factor of 1.916. He had 130 RBIs in 704 PAs (with 9 IBBs) during the season, but really stepped up his game in the clutch, driving in 44 runs in 90 PAs (with 6 IBBs). Other MCPs were Alexei Ramirez (1.901 ClF), Johnny Damon (1.884 ClF), and Grady Sizemore (1.853 ClF).

The LCP during 2008 was Milton Bradley of the Rangers, with an astonishing clutch factor of -0.112. Yes, Milton Bradley was the only player in the three years with a negative ClF, meaning that he was actually worse when the pressure increased. He drove in 77 runs in 509 PAs (13 IBBs) during the season, but only had 8 RBIs in 65 PAs (6 IBBs) in clutch situations. Thus, Milton Bradley has been crowned the "King of Choke", with a performance that no other player even came close to. Other LCPs included Paul Konerko (0.043 ClF) and Adrian Beltre (0.087 ClF).

Finally, last year there were 155 players who qualified for the batting title, with an average player clutch of .213 (the highest league average of the three years). The MCP of 2009 was Hanley Ramirez of the Marlins, with a clutch factor of 2.120 (the third highest total). He had 106 RBIs in 652 PAs (13 IBBs), but had 34 RBIs in 65 PAs (6 IBBs) in the clutch. Other MCPs include Joey Votto (1.796 ClF), J.D. Drew (1.721 ClF), and Edgar Renteria (1.661 ClF).

Lastly, we find the Bengie Molina was the LCP of 2009. He had a clutch factor of 0.137, with only 14 clutch RBIs in 78 clutch PAs. Other LCPs include Troy Tulowitzki (.260 ClF), Mark Teahen (.303 ClF), and Jason Kendall (.306 ClF).

There are the results I have obtained for the past three years using my clutch statistic. It is very interesting to see which players stepped up their game in the clutch, while other players (Milton Bradley!) actually perform only marginally better, sometimes even worse. I am going to do another post later on with some interesting patterns and phenomena that we can see in the data, most importantly, whether or not the statistic shows that players can perform consistently in the clutch.

The Art of Clutch: Part I

In today's post I am going to discuss something that has been debated by baseball fans and statisticians for many years: whether or not a certain player is "clutch". Certainly, there are some players out there (Derek Jeter anyone?) with the label of being a "clutch" player, but so far there have not been any studies done that prove that some players are more "clutch" than others. Today I am going to attempt to figure out an appropriate statistic that will measure how clutch a player is, and then later this week I will present my results and any conclusions that can be drawn from the findings.

*Note: I am only measuring batting statistics, a study on pitchers being clutch might come later

The first thing I wanted to determine was an appropriate statistical equation to measure how "clutch" a player is. First of all, what exactly is clutch? It is the ability to score runs in clutch situations, those with high leverage. The clutch situation I am going to be measuring is with runners in scoring position and two outs. There are some, shall I say, more statistically accurate measures of clutch, but the situation of 2 outs and RISP happens multiple times per game, giving us a large enough sample size to be able to draw reasonable conclusions. So the first thing I want to measure is the RBIs per plate appearance for hitters with 2 outs and RISP. One adjustment to make is to eliminate intentional walks from the equation, as when hitters are intentionally walked they are not given the opportunity to drive in runs, so they should not be penalized. I am also going to put a C in front of each statistic that is measured from clutch situations to denote that it is a "clutch" statistic. So the first part of the equation is going to be C RBI divided by (C PA minus C IBB).

Now, we need to expand upon our definition of clutch a little bit. Being clutch not only means driving in runs, which varies widely with the hitter, it means that the hitter raises his game as the pressure of the game raises. This means that a hitter who is clutch should have a higher RBI/PA when the game is on the line opposed to his overall RBI/PA. We can measure this by taking the RBI/PA for clutch situations and subtract the RBI/PA for all other situations. This number will represent the increase in RBI/PA that a player will have in clutch situations over regular situations. The formula is a little convoluted, but all pieces are necessary. We can see that the first part is just the piece from above, and the second part is figuring out the RBI/PA for all non-clutch statistics.

The final step for our equation is to normalize the statistic onto a scale of one. To do this, we are going to divide the individual player's clutch statistic by the league average statistic, which is just the same calculations with the league totals for all eligible players used as the numbers in the formula.

What the resulting number will express is the average increase in "clutch" for each player, relative to league average. A number of one means that that player is average in clutch situations, that is, the player drives in an average extra of runs per plate appearance in clutch situations. A number less than one means the player is a below average clutch player, and a number greater than one means the player is an above average player. The resulting numbers will almost always be between 0 and 2, a number less than 0 means the player is downright miserable in clutch situations, and a number greater than 2 means the player is absolutely amazing in clutch situations.

So this is the process to determine how clutch a player is. It will be interesting to view the results, and determine whether or not a player can consistently maintain how clutch they are from season to season.