Finding wins

imageimage
Who’s a better bet for wins down the stretch? (Icon/SMI)

In the last month of the season, fantasy owners (especially in rotisserie leagues) often find themselves desperately looking for help in a particular category. Perhaps the most frustrating standard category to find yourself behind in is wins. Wins are notoriously hard to predict because the correlation between wins and pitcher ability is much less than that of, say, strikeouts or ERA. I know of some fantasy purists who prefer not to use it as a scoring category at all. That said, if wins is a scoring stat in your league and you’re behind in it, perhaps there is a small way to tilt the odds in your favor by using American League pitchers instead on their National League counterparts.

Why are AL pitchers more likely to record a win? Well, the longer into the game a pitcher pitches, the more likely he is to record a decision and, in particular, a win. Thus, AL pitchers are less likely to finish an inning that they’ve started but also more likely to start any given inning. Both are due to the fact that pitchers bat in the NL.

AL pitchers are taken out of the game for pitching reasons only. NL pitcher may be taken out or kept in the game for batting reasons. A manager may take out an NL pitcher early because his turn to bat came up and it was advantageous to pinch hit for him. Alternatively, though, a manager might keep an NL pitcher in longer if he was due to bat in the next half-inning, so as to avoid a double switch or wasting a relief pitcher. The former effect would tend to mean AL pitchers pitch more while the latter would tend to mean less. As it turns out, using data from 2008 on starting pitchers, AL pitchers pitch longer and are thus more likely to record a win.









































League AL NL
Win % 37.0 33.5
Loss % 34.6 33.9
Strikeouts 4.08 4.24
Inn Pitched 5.92 5.80
Batters Faced 25.52 25.13
Pitches 95.94 95.05
Proportion finished inning .666 .725

These are data from non-interleague games only. AL starting pitchers win 37 percent of the time they start whereas NL starting pitchers win only 33.5 percent: basically AL starters are 10 percent more likely to win than NL starters. (For those wondering, this difference is “statistically significantly different”—the same is true for all the other differences, except for the probability of losing a given start.) Unsurprisingly, AL starters strike out fewer batters (and of course have a higher WHIP and ERA), so the higher win percentage does not come for free.

AL starters are much less likely to finish an inning that they started, but they still pitch more innings and face more batters than their NL brethren. (Note: I was not able to tell if a pitcher started an inning but did not record any outs.) AL pitchers even pitched about one more pitch per start. For what it is worth, a simple (probit) regression of wins on whether or not the pitcher finished his last inning of work tells us that pitchers that finish their last inning are much more likely to record a win. I haven’t reported the regression’s results here because regressions imply causation and I would want to control for many more effects before I’d be comfortable with the numbers.

Apparently, NL pitchers are much more likely to come out of the game early for a pinch hitter than they are to stay in the game to avoid a double switch. Even though they face an easier lineup, they don’t pitch as far into games as AL pitchers. Oddly enough, in interleague games, NL starters pitch longer in AL parks than in NL parks whereas AL starters pitch longer in NL parks than in AL parks. Nevertheless, if you are desperate for wins in the final weeks of the season, everything else equal, pick up a Yankees or a Angels starter rather than one from the Dodgers or Phillies.


1 Comment
Oldest
Newest Most Voted
Inline Feedbacks
View all comments
Paul Singman
14 years ago

Nice research, Jonathan. If you are looking for a win, you should take this into account, but there are also I think more meaningful factors I discuss in this article.