Premium and Super Premium Subscribers Get a 20% Discount at MLB.tv!
September 26, 2011
Resident Fantasy Genius
Earlier this season, I had a discussion with one of my leaguemates and friends in the CardRunners Fantasy Baseball Experts League, Chris Hill (a poker player who partners with Hollywood director Nick Cassavetes for the league), after he read my article about how long it takes for hitting stats to stabilize. In that article, I proposed that a hitter’s stolen-base success rate (SB%) is extremely volatile, but Chris said that I might want to reexamine that line of thinking. He said that he hadn’t run a study on it, but he had always attributed a caught-stealing spike to “either an injury indicator or a precursor to significant decline of all skills [and] almost never random,” rattling off a series of such examples. I thought that this was interesting and at least worth a look, so I asked him to propose a way to study his theory, and I would run the study.
Based on his proposal, I’ll be examining all hitters who played in three consecutive seasons dating back to 1993 (a “baseline, a decline year, and a rebound year”) and had at least 15 stolen-base attempts in the first season. From here, I’ll use this set of batters to answer the four questions he proposed:
Of course SB% drops significantly, but wOBA barely drops at all. We see a 1.2 percent drop, good for just 0.004 points of wOBA for the hitters in question.
When SB% drops, is a wOBA drop soon to follow?
Nope. Here, the batter’s wOBA actually raises a point. While it may make sense that a hitter who loses a step will lose it not only in his stolen bases but in his ability to beat out hits, his actual stolen-base success rate doesn’t appear to be a good enough proxy for losing that step. It appears that there’s just far too much random variation in it, backing up my initial findings in the article that spurred this study.
How much does SB% rebound after a large drop?
When a hitter experiences a large decrease in stolen-base success rate, we actually find that it bounces back—to an extent. It doesn’t return to its previous level, but it doesn’t remain low, either. While it looks as though it basically averages the two, I think more than anything, what we’re seeing here is regression to the mean. While we’ve required that the Year 1 SB% be high and the Year 2 SB% be low, the Year 3 SB% has no requirements. And because SB% is so full of random variation, while the previous two years’ performance matters a little, a true Year 3 projection would place the players very near the league average, which is what we see.
Chris was also interested in the percentage of players that bounce back to their Year 1 levels in Year 3. If I check that, I find that roughly 32 percent of the players in my study return to or exceed their Year 1 SB% in Year 3. So a steep dropoff isn’t damning, but a bounceback isn’t the most likely scenario.
Do batters attempt fewer steals if they’ve been less successful?
We see that in the year of the SB% dropoff, SBA% drops off at an almost identical rate (28-29 percent). In Year 3, when SB% regresses to the mean, SBA% loses one percentage point—nothing dramatic. Because SBA% doesn’t take very long to stabilize, we wouldn’t expect it to move much based on regression, and we don’t see it move too terribly much based upon the increase in getting caught—at least above and beyond the initial drop.
But what happens to the 41 percent of players that don’t bounce back at all? The ones that are successful at an equal or worse rate in Year 3 than they were in Year 2?
If a player continues to get caught at a significantly higher rate than he did in Year 1, his attempts will experience a further decline. Of course, this can be difficult to predict because of the unpredictable nature of SB% and because the two things happen simultaneously. We’d need to be able to predict that the player’s SB% won’t regress to the mean in order to predict the accompanying dropoff in SBA%, and that’s no easy thing to do. Still, if we have some information on the player—perhaps scouting information that tells us that he is indeed slower than he was in Year 1—this is a useful thing to know. Not only could the player’s success rate remain low, but he could also start attempting steals at a lower rate—a dreadful combination. After all, if the player has actually lost a step, his manager is going to recognize it far quicker than our stats will, and he’ll be giving him the red light more often.
Finally, in case you haven't seen, be sure to catch me tomorrow at 1 PM EST for a live chat here at BP!