Posts Tagged ‘Markov Chains’


Mike Scioscia is making me crazy

May 21, 2009

Sabermetrics has come around on the stolen base a bit. Back in the Moneyball era, the steal was a no-no. Since then, front offices such as the Athletics and Red Sox have come around on the stolen base, so long as it’s successful at a very high rate.

If your stolen base attempts are successful less than 70-75% of the time, then you’re penalizing your team in two ways. First, the stolen base removes a baserunner, something that the average at-bat only generates about a third of the time. Even more important than the loss of the baserunner, however, is the impact of the out. This topic has been beaten into the ground by the geniuses at Hardball Times, Fangraphs, and Tom Tango’s The Book Blog, so I’m going to be quick in my explanation so that I can get to the point. Baseball has no clock. Outs are akin to the “time” a team has left to score. They are your “most precious commodity” (hat tip to Earl Weaver), and should never be squandered.

Today’s LA Times article about the Angels’ reckless attempts to take third by theft is a terrible example of justifying a stupid tactic through veteran endorsement. The article’s thesis is that attempting to steal third frequently is OK because Mike Scioscia and Torii Hunter say that it is worth the risk and the consequences of getting caught.

Let’s break it down using an expected runs framework*:

Bases    0 outs    1 out    2 outs
xxx    0.54124    0.30065    0.10974
1xx    0.91891    0.56768    0.24182
x2x    1.11888    0.73508    0.33831
xx3    1.25203    0.98077    0.42975
12x    1.55291    0.95307    0.47846
1×3    1.77027    1.18257    0.61926
x23    2.10112    1.52899    0.6437
123    2.29851    1.71074    0.83677

For those of you who haven’t seen this before, here’s how the chart works. You take the bases from the left and the number of outs from the top of the starting state, and then you take the same combination for the conclusion of a play to see how many expected runs you gained or lost from your move. This chart is how you can demonstrate that in most situations a sacrifice bunt leaves you in a worse position than before the bunt.

Let’s say Torii Hunter is on second base with no outs. The Angels should score 1.12 runs in the inning on average. If Hunter steals third, they should score 1.25 runs on average. His gain of .13 runs is very small. The reason for the tiny gain is that most plays that score a runner from second also score a runner from third (especially a fast runner like Hunter), and that with no outs there are lots of chances for Hunter to advance, even on outs. Now let’s see what happens if Hunter is caught. The Angels go from an expected 1.12 runs to .30 runs, a loss of almost an entire run (.82 runs to be exact).

If Hunter were stealing with 1 out, the Angels expected runs rises from .74 to .98, about a quarter of a run, and the possible loss from being caught is the difference of .74 and .11, a loss of about two-thirds of a run. If he succeeded with two outs, the rise is from .34 to .43, about a tenth of a run, and the potential loss is .34 (you can’t score when you have three outs). The point is that stealing third successfully gets you a very marginal advantage, and failure comes at a great cost. The penalty goes down as the number of outs increases, because it becomes less likely that the inning will say alive long enough for Hunter to be driven home.

Stealing third tends to be less successful than stealing second for a number of reasons, including the much shorter throw from the catcher. With this in mind, a manager should only attempt stealing third if they think the defense is napping (such as when the third baseman is way off the bag or covering for a bunt), or when they have a combination of a very good base stealer and a very bad catcher. Kenji Johjima is a terrible hitter at this point, but he is a very effective defensive catcher. The Angels used the argument that because Felix Hernandez was pitching that they had to do everything possible to score against him. They are correct that Felix might change the situation, but they are dramatically overestimating the impact of the steal. They would be better off hoping that one of the next two hitters got a base hit (about a fifty-fifty shot) than having Hunter atempt a low-probability, low-benefit play.

What infuriates me is that attempting to steal third in a one-run game when there are two outs and a bad hitter up could be an acceptable tactic. Scioscia’s insistence on suiciding his runners in all situations just provides another example that he’s unwilling to adopt any sort of nuanced approach to his unabashed love of small-ball.

Of course, this is the same nonsense we should expect from a club that doesn’t advocate taking walks, and thinks that because they won a World Series with small-ball once that it should trump all other tactics. Scioscia needs to realize that 1) his team is not as young and fast as it once was and 2) that while small-ball can be effectively mixed with other tactics, the Angels’ current recklessness on the basepaths is only costing him further outs that his on-base percentage deprived team of Mickey Hatcher** acolytes can not spare.

*Expected runs courtesy of Baseball Prospectus, but they are available from The Book and several other sources. Expected runs are the basis of Hardball Times’ RE24 statistic.

**Hatcher’s philosophy is that swinging more and swinging earlier is the route to winning. Sounds more like a recipie for a guy batting .275 with an on-base percentage of .300 to me.