April 28, 2005

Crooked Numbers

The Ivy is Always Greener...

by James Click

Printer-friendly

LaTroy Hawkins has no stomach. At least that's how things sound in Chicago right now as Cubs fans and media lament the failure of Hawkins to succeed when donned with the "closer" label. Taking things at face value, it sounds like Hawkins' esophagus runs directly to his small intestine, a genetic trick leaving him without the ability to finish games he enters when his team is up by three or fewer runs, when the tying run is on deck, or if those three innings happen to be the last three innings of the game. Otherwise, he's great.

The idea, as always, is that it takes something special to pitch in save situations. So far in 2005, Hawkins has three saves and two blown saves. The first blown save was April 8 against Milwaukee where he allowed a single to Junior Spivey and a double to Lyle Overbay after getting the first two outs in the ninth. The Cubs eventually lost the game in 12 innings.

The second was against Pittsburgh this past Saturday where, after Greg Maddux pitched eight solid innings, the Cubs took a one-run lead in the bottom of the eighth only to watch Hawkins surrender a one-out home run to Jason Bay. Pulled after giving up a single to Craig Wilson, Glendon Rusch allowed a triple to Freddy Sanchez, scoring Wilson. Since Hawkins had allowed Wilson to reach base, the run counted against him and he was credited with the loss when the Cubs failed to score in the bottom of the ninth.

In his three games previous to last Saturday, Hawkins pitched very well and recorded three straight saves, maintaining two two-run leads and one one-run lead. But one misplaced pitch to Bay and the screamers came out in force, buoyed by Hawkins' reputedly shaky performance in 2004. Though he notched 25 saves last year, doubters pointed to Hawkins' nine blown saves, not his 82 innings of 2.63 ERA. Of his nine blown saves, eight came with the tying run at the plate when Hawkins entered the game.

Looking at those totals, it's easy to speculate that Hawkins is shaky in tight situations, a closer who's unstoppable with a three-run lead, but unreliable with less. But when Hawkins entered the game in 2004 with the Cubs ahead or behind by a run or with the game tied, he pitched 37.2 innings at a 3.34 ERA. While that figure is higher than his season total of 2.63, it's significantly better than league average and it's in the top half of Cub relievers last year. But if you're going to be a "closer," that's just not good enough.

Instead, let's look at things in a slightly different way. In Baseball Prospectus 2005, Keith Woolner unveiled the new pitcher reports in his essay on Win Expectancy. One of the more interesting metrics that arose from that gaggle of symbols and numbers that may only pass for communication if you went to MIT was a stat called "Leverage." Leverage is defined as "the impact on the probability of winning the game from scoring (or allowing) one additional run relative to a run scoring at the start of the game." Leverage is thus centered at one with values higher than one indicating the situation has a greater effect on the outcome of the game whereas values below one indicate situations less deterministic on the outcome of the game than before the game began. Keith further points out that "leverage is not a measure of a pitcher's actual performance, but rather a measure of how he was used."

This last point gets to the heart of the argument about "closers": is leverage independent of pitching performance? Three weeks ago, I mentioned the idea of context independence: the theory that a batter's performance is independent of the context in which it takes place. Essentially, a batter is just as likely to hit a home run with no one on base and down by 10 runs as he is with the bases loaded and down by a single run.

We can quickly check Hawkins' usage and performance for the last few seasons to see if perhaps his performance declines as his leverage increases. To do so, we'll break up leverage into five different strata, setting the ranges as to yield an equal number of innings pitched by all relievers into each stratum. Then we can look at Hawkins' performance in the last few seasons in each of those five strata. Here's how he fares in each level in RA:

Pitcher        Year   Low  MedLow Medium  MedHigh High
-------        ----   ---  ------ ------  ------- ----
Latroy Hawkins 2004  4.95   1.35   1.89     1.26  4.14
Latroy Hawkins 2003  0.00   2.88   2.88     2.79  1.89
Latroy Hawkins 2002  0.45   4.05   6.30     1.53  0.00
Latroy Hawkins 2001  5.94   6.03   5.40     6.21  6.12
Latroy Hawkins 2000  4.41   2.25   5.04     3.87  2.52

While Hawkins had a rough time in 2004 in the high leverage situations, he also did in the low leverage situations, instead excelling in the three middle tiers. But look at his 2003 and 2002: in both Medium High and High leverage situations, Hawkins had RAs that would make Bob Gibson jealous, going so far as to put up a 0.00 RA in High leverage situations in 2002. There's a small sample size issue in 2002 (Hawkins only pitched 17.2 innings in Medium High and 11.0 in High), but in 2003 he pitched 19.3 in Medium High and 28.3 in High leverage situations.

Now let's see how Hawkins compares to all relievers who threw at least 50 innings in a season over the last few seasons:

Year  Low   MedLow  Medium   MedHigh  High
----  ---   ------  ------   -------  ----
2004  4.29   4.55    4.01     4.03    3.93
2003  4.37   4.00    4.05     4.21    4.02
2002  3.87   3.95    3.86     3.93    3.88
2001  4.00   3.98    4.05     4.32    4.23
2000  4.61   4.53    4.56     4.67    4.19

Before we discuss those numbers and their bearing on Hawkins, remember that there is an inherent selection bias in this group. Unlike a batting lineup, managers can select when to use pitchers later in the game and thus, in situations perceived to be high leverage, pitchers who are perceived to be better are likely going to see more action. Thus, as opposed to the context independence espoused when looking at Hawkins' numbers, when looking at league totals, we should instead expect the numbers in higher leverage situations to be higher because teams will pitch their best pitchers in those situations.

On a very broad scale over the last five seasons, the higher the leverage of the situation, the lower RA becomes (though 2001 was an interesting exception to that trend). But this is most likely a result of the fact that teams are more properly deploying their available talent than those monkeys and their stock picks. Among the top 30 relievers in save totals last year, only 11 have an RA in High leverage situations lower than their RA in Low Leverage situations. On average, the top 30 relievers in save totals in 2004 had an RA of 3.51 in High Leverage situations and a 3.33 RA in Low leverage situations.

Getting back to Hawkins, his 4.14 RA in High Leverage situations in 2004 is enough to make even Jose Mesa fans cringe, but looking at his numbers back through 2000, that 4.14 looks like an outlier. Hawkins proved in 2002 and 2003 that he has the stomach to pitch with a higher degree of success in High Leverage situations than the best closers in the game. That he didn't in 2004 isn't any measure of his lack of makeup, but rather an unlucky year combined with the newly focused media spotlight that comes with that "closer" label.

Based on this information, can we definitively say that there isn't some mysterious psychological component needed to be a closer? No, but we can say that Hawkins--for all the hand wringing over his troubles last year and this month--has proven that he can succeed to the same degree as the bullpen royalty when the pressure is on. What might be most hilarious in this entire escapade is that all this is coming from a team that hasn't had a player lead the team in saves two seasons in a row since Randy Myers in 1994-95. The biggest lesson isn't that Hawkins is a good reliever and could close as well as most, but that the faith of Cub fans takes years to build and minutes to destroy.