Posted by Andy on January 8, 2010
Raphy's recent post about starting pitcher performance in 2009 touched on something I've been thinking about for a while. I did some additional research and have a two-part post looking at the results.
First, here is a plot showing how many innings were pitched in each start in 2009.
I generated this data by using the PI, simply looking up total number of games by the starting pitcher for each number of innings. Keep in mind this covers 2009 only.
Tomorrow I am going to break out the data in the above graph by AL and NL starts. For today, though, let's just look at the above data.
Here are some of the interesting things you can conclude from the data:
- The average starting pitcher went a little under 6 innings, with exactly 6 innings being the single-most common number of innings pitched. You can figure the exact average number of innings per start by looking at the 2009 pitching splits by role, which shows that in 4860 starts, starters totaled 28257.1 innings, or 5.81 innings per start.
- Starters were much more likely to pitch an even number of innings, especially later in the game. This can be seen from the high spikes at exactly 4, 5, 6, 7, and 8 IP, as opposed to the much lower numbers in between. This makes a lot of sense for a few reasons. Firstly, it's quite customary to take a pitcher out between innings, especially if looks tired or has reached a high pitch count. Secondly, in the NL, starting pitchers often come out for a pinch-hitter when their turn at bat comes. Thirdly, it seems to be more and more common for managers to use their bullpen in fairly rigidly defined ways, particularly using one reliever as the 7th-inning setup guy, another as the 8th-inning setup guy, and a 9th-inning closer.
- There's an interesting trend in these later innings. A total of 233 starters were pulled during the 4th inning, meaning with 1 or 2 outs in the 4th. However, 685 starters went a full 5 innings. That means pitchers were 2.94 times as likely to pitch 5 innings than they were to get pulled during the 4th. Moving one inning later, pitchers were only 2.54 times as likely to pitch 6 innings than they were to get pulled during the 5th. Advancing another inning, the ratio is 2.36 times as likely to go 7 instead of 6.1 or 6.2. The ratio falls all the way to 1.74 for likelihood of going 8 full instead of 7.1 or 7.2 This all makes sense--the later the game gets, the more likely that the starter gets pulled in a critical situation. However, once a starter gets into the 9th inning--meaning 8.1 or 8.2 IP, he's 5.73 times as likely to pitch the full 9 innings than to get pulled at that point. This follows intuitively--if a manager has made the decision to let the starter pitch the 9th inning, most of the time he finishes it, often I would assume because his team has a comfortable lead. Keep in mind that if the starter gives up a baserunner or two to start the 9th but doesn't record an out, such an occurrence is invisible in this study because that counts as 8 IP.
- Also of note is that there isn't much difference in the odds that a starter gets taken out with 1 vs. 2 outs in the inning. You can tell this since the data points for 5.1 and 5.2 are pretty similar, or 6.1 and 6.2, etc. This makes sense since the starter is likely to get pulled for two reasons: either he's gotten in trouble, which is pretty much just as likely to happen with one out as with two outs, or a special matchup has arisen, such as a tough lefty, for whom the manager wants to put in a lefty specialist reliever. This, too, is equally likely to occur with any number of outs in the inning.
In Part 2, we'll look at the difference in this graph in the AL vs. the NL, which I found quite interesting.