Starting/Sitting Pitchers

If you're lucky, you have a pitching staff so good that you never have to sit a starter ever. In reality, this never happens. Unless your leaguemates are all enamored of the likes of Edwin Jackson, Scott Baker et al., you're not getting a staff full of top-20 pitchers. Depending on how the draft unfolds, you may not even be able to rely solely on top-40 or top-50 pitchers. And once you dip outside the range of Matt Cain/David Price/Yovanni Gallardo, some pitchers just don't fare well against teams.

When faced such a conundrum, I like to head over to Baseball Reference. You can pull up the hitting statistics that any pitcher gives up to a team of your choosing (e.g. Wandy Rodriguez versus current members of the Pirates). My eyes go straight to one number in this case: OPS.

Checking a pitcher's OPS against encompasses just about everything relevant to fantasy leagues. Hits and walks issued, extra base hits and home runs are all directly accounted for in OPS. Earned runs and strikeouts are indirectly accounted for (though I will look at strikeouts per PA as well). In the case of the former, earned runs require earned runners, which are reflected in OPS. The more batters get on and the more bases they take, the higher OPS. In the case of the latter, batters who strike out lower OPS (though certainly no more than batters who ground out/fly out, but still). Looking at OPS allows you to assess the probability that a pitcher will actually damage your fantasy stats.

The question is then what should you use as a guideline when looking at OPS? At what OPS does a pitcher tip into the danger zone?

First off, you have to decide how to measure the quality of a pitcher's starts. OPS surprisingly does not correlate strongly with wins and losses, so those shouldn't factor in. OPS does, however, correlate with innings pitched, strikeouts, WHIP, and ERA, so those should be accounted for. OPS also correlates with other statistics (such as, not surprisingly, home runs allowed), but strikeouts/WHIP/ERA/Innings Pitched make up the bulk of a pitcher's value, directly or indirectly, in both roto and points leagues. So we'll stick with those.

Now that we've narrowed our focus down to K/WHIP/ERA, we have to figure out how to assign value to them. The easiest way to do this is to take standard fantasy points scoring and create a point value- one point per K, three per inning pitched, minus one per hit or walk and minus two per earned run.

(Incidentally, this gives you a pitcher's record-independant value in points leagues.)

The best next step is to go pitcher by pitcher and count innings pitched, Ks, hits, walks, earned runs and OPS for every team that pitcher faces. But, as you can imagine, this is incredibly time consuming. Even limiting the analysis to the top-40 pitchers, that's approximately 1,200 starts per year of data.

Instead, what we can do is get season data for each pitcher going back five years and calculate their record-independent points and OPS against. This gives a general idea of the relationship of OPS to fantasy performance, and we can interpolate the results to individual matchups. This isn't ideal, of course- I don't think anyone would want to roster a pitcher who has a season OPS against of .900, while any given pitcher might sport such a stat against a particular team- but it's a reasonable approximation.

The last step is to define what an acceptable fantasy performance is. In this case, I would use a mean split. By grouping pitchers on either side of average performance, we can look for an OPS cutoff that denotes a (roughly) 50/50 likelihood of getting a performance worth the active spot.

For this analysis, I gathered data on every starting pitcher who threw at least 100 innings from 2006 to 2010. I then calculated isolated pitcher points with the formula (3*IP)+K-H-BB-(2*ER). This gave us a mean of 246.49 points; for reference, this is roughly equivalent to Tim Hudson's 2008 season or Brett Cecil's season last year.

After defining that as the cutoff for above/below average, we can then see how OPS differs between the two groups. For the bottom half of pitchers, the average OPS against was .795 (ranging from .615 to 1.006). For the top half, the average OPS against was .698 (ranging from .557 to .838). The difference in mean OPS between above and below average performance is also statistically significant.

What this means, essentially, is that any pitcher who has an OPS against below .700 for a matchup is a good bet to start, and any pitcher who has an OPS against above .800 is a good bet to sit. In between .700 and .800, it's something of a gut call.

Here's how the .700-start/.800-sit rule of thumb has fared through the three days of games this season:

Liriano vs. Blue Jays (.956 OPS against): 4.1 IP, 4 H, 5BB, 4 ER, 3K (8.31 ERA, 2.08 WHIP)

Volquez vs. Brewers (.926 OPS against): 6 IP, 7H, 2BB, 5 ER, 5K (7.50 ERA, 1.50 WHIP)

Dempster vs. Pirates (.910 OPS against): 6.2 IP, 6H, 4BB, 6ER, 7K (8.10 ERA, 1.50 WHIP)

Lackey vs. Rangers (.877 OPS against): 3.2 IP, 10H, 2BB, 9ER, 3K (22.09 ERA, 3.27 WHIP)

Hudson vs. Rockies (.541 OPS against): 6.0 IP, 6H, 2BB, 3ER, 5K (4.50 ERA, 1.33 WHIP)

Weaver vs. Royals (.508 OPS against): 6.1 IP, 2H, 2BB, 0 ER, 6K (0.00 ERA, 0.632 WHIP)

Like any good analyst, I won't claim this guideline has perfect predictive power; Ubaldo Jimenez and Matt Cain have an OPS against in the .720-.750 gray area against the D'backs and Dodgers, respectively, but got very different results to the tune of 4 hits, 5 earned runs, and 2 home runs more for Jimenez in the same number of innings. Shields owns a .855 OPS against the Orioles but allowed only six runners and two earned runs in 7.1 innings. But as a general guideline, .700-start/.800-sit is a good rule of thumb.