Classifying relievers, part 1

I am fascinated by relievers. I can’t exactly say why, but there’s something that I find, well, marvelous about the bullpen. I thought of trying to reason or explain my love of the bullpen, but I don’t think I can effectively do that. So let’s just leave it at that.
In this generation of the game, relievers are crucial in deciding games that are close and in the later innings. Into the late 1980s and early 1990s, as pitch counts became commonly tracked and complete games declined, relievers began to take on a much more important role. In the American League in 2009, approximately 3% of all games were
complete games, down from 16% in 1988.
With an increased focus on relievers, means of evaluation are important. There’s certainly plenty of debate as to what the best way of evaluating reliever performance is, but I actually wouldn’t like to look at that right now. Personally, I’d like to take a look at a reliever’s “role”: what the situation was when he entered, when he left, for how long did he pitch, etc. In this post, I’ll use some metrics that I think are useful for classifying members of the bullpen.


AIPGR

The simplest statistic I use when evaluating relievers is Average Innings Pitched per Game Relieved (AIPGR). It can be found by dividing the total innings pitched by the number of games relievers (IP/GR). It's not going to say a whole lot about the pitcher, but it will say something about how long a reliever pitched on average in a particular season. These days, it’s uncommon to see a reliever average over 2 innings per appearance, and the average over the past few seasons has been between 1 and 1.1 (it was 1.05 in 2009). Something that struck me was just how “specialized” relievers are now compared to in previous decades. Here is a chart that shows the AIPGR of the average big league reliever over the past 35 seasons.




That’s a pretty steep incline, especially around 1990. It’s not that surprising, considering pitch counts were first counted by STATS LLC in 1988. Similarly, the closer role has become more defined with the increase in specialized relievers. This graph shows the percentage of 2 inning+ saves since 1974.




Again, not that surprising, considering overall drop in AIP. But it’s still cool to see on a graph. At least I think it is.

Leverage Index

But really, how much else can you get from AIPGR? It would make sense that “specialty” relievers (“one-out-guys”) would have AIPs well under 1, closers would have AIPs around 1, and long-relievers would have AIPs well over 1. But AIPGR doesn’t say anything about the situation that the pitcher faced. That’s where Leverage Index comes in. LI is a statistic that shows the “pressure” of a particular situation, with 1 being the “average” situation. It would take me all day to figure out how to accurately and coherently explain it, so I’ll just refer you to this article by the metric’s creator, Tom Tango. Anyway, LI can be viewed in a number of different ways. Game LI (gmLI) shows the average leverage that a pitcher faced when entering the game; Play LI (pLI) shows the overall leverage that a pitcher faced; Inning LI (inLI) shows the leverage when the pitcher entered an inning; Exit LI (exLI) shows the leverage when the pitcher left the game. For the purpose of showing a pitcher’s role, I think it is most effective to use gmLI, as that shows the situation that the pitcher’s manager wanted him to face. Before I go on, here is a chart to show the difference in pLI/gmLI since 1974.




There’s not a whole lot of difference between pLI, but the gmLI numbers have dropped precipitously. This would lead me to believe that in previous decades, relievers were used primarily to get out of jams created by starters, while they are more commonly used to start clean innings these days.

Using both of these things

This graph shows all qualified* relievers since 2005 in terms of AIPGR and gmLI.
*either 50+ IP, 50+ GP, or 20+ saves (there are only two of these pitchers that don’t fit the other requirements. I included them because I will be looking at closers in a future post.)





This pretty much confirms what I said earlier. The highest concentration of pitchers with gmLI over 1 are those that have AIPs around 1; I’d guess those primarily closers. To left are the lower AIP pitchers, the specialist, and they seem to have slightly lower LI marks than the closers. And way over to the right have AIPs well under 1, for the most part.

So, we’ve looked at the average length of reliever appearances through AIP and the average “pressure” faced by the pitcher through LI. In my next post, I’ll scratch the surface a little bit more than I did here, as this was just an introduction. I’ll look at some MLB averages and specific examples.

Data is from Fangraphs and Baseball-Reference.