Estimating a new break-even point for stolen bases based on league-wide OPS (a bit math-centric)

[Hallas]

I'd like to warn anyone that this is a lot of theorycrafting, so if that stuff scares you, look away now.

There was a little note in one of Britt's articles about how Buck wanted to promote a little more aggressiveness on the basepaths, since we're tied for last in the league in SBs.

I think the general consensus is that lower league-wide offense lowers the break-even point for stolen bases. Since I don't have aggregate stats, unfortunately I can't post a detailed yearly analysis of base/run opportunities. However, it's no secret that offense is down. Last season the league-wide OPS was .728. This season is .710. While some of that might be due to a rainy April, if it persists then I think we have to re-evaluate small ball.

Based on TangoTiger's run expectancy matrix, if you steal 2nd between 1999 and 2002, you should have been successful about 65% of the time to help your team. If you steal 3rd, the number is much higher than that, but stealing 2nd is more common, so we'll stick with that. Now, reducing all run expectancies by 1/3 would seem appropriate, except that it probably wouldn't be correct. When you have 0 out, you have 2 or 3 chances to generate a run-scoring event, so the probabilities are multiplicative. With 1 out, you have 1 or 2 chances. With 2 out, you have 1 chance only.

The objective here is to try and estimate the difference in run-expectancy without actually going through all the PBP data and generating a new matrix. My idea for approximating this would be to take the run expectancy for 2 out. Come up with a number that represents the percentage difference (this number should be close to the total difference, 0.25). We'll call this number estimated difference or ED. Calculate your new RE with 2 out such that:

REnew = REold * (1 - ED).

Then, take the value for 1 out, and calculate a new RE, such that:

REnew = REold * (1 - ED) * ((2 * % no double play) + (1 * % double play)).

Then take the value for 0 out, and do the same calculation... it should be something like (3 * chance of no double play) + (2 * chance of double play). I'll assume triple plays are rare enough to be safely discarded.

You can then weight these three values if any of the base/out states occurs more frequently than another.

Anyway, my quick calculation (my lack of double play data forced me to ignore them) says that the percentage goes down to around 62%. It's only about 3%, and probably not enough to make a huge difference in game planning, but it's still worth considering for teams that are looking to squeeze a few extra runs. I do think that, given the offensive environment, it does mean that players with the ability should be a bit more eager to take an extra base when an opportunity presents itself. I think that a pretty key validation to this conclusion is the fact that Mr. Moneyball Billy Beane has his team with 20 SBs against 6 CS. Obviously they are putting themselves in a much better position to score given their success rates, but it's telling that his team is actually running.

[sangar]

Very interesting, thank you.

[SteveA]

You know, I was just thinking the other day, with scoring down, that one-run strategies could see a revival in importance. Thanks for running some numbers so we have an idea of the magnitude of the change.

Now, if you get to the point it's a tie or 1-run game in the late innings, you probably even have a lower threshhold because the importance of the 1 run is magnified. Probably under 60%. Meanwhile, there are some teams out there (Boston, Chicago) that can't throw anybody out and are giving up ridiculously high percentages. I know Roberts' injury hampered us until recently, but we probalby need to take advantage of the guys we have that can steal a base a bit more than we have been.

[Hallas]

AJ can probably steal enough to be borderline/slight positive. Other than Roberts I don't really see another baserunning threat.

It's kind of an art. I don't think Roberts is the fastest guy on the team, yet he's clearly the best base stealer by a long shot. I don't think anyone exemplifies this any more than Barry Bonds, who stole 69 bases after his steroid-crazy age 34 season against only 11 CS. He was basically molasses at that point, and yet his instincts propelled him to be an above-average basestealer.

[Slappy]

I've never thought that one-run strategies vs. multi-run strategies should be examined in a vacuum. Clearly there are times when you need to play for one run, usually when you are down by 1 or up by 1 or 2 late in the game. Other times you should hold back and play for multiple runs. This stuff has been covered with discussions of leverage. I'm not sure science can add a whole lot to improve perceptions (and, frankly, existing measurements) of who can and who can't steal bases.