from Amazon
Sunday, June 11, 2006
The Value of the Stolen Base
Going over the Dave Roberts steal off Mariano Rivera in 2004.
In The Book, we get very detailed with respect to the value of the stolen base. While the standard breakeven point is around 72%, the true breakeven point is 69%. Timing is everything. In certain situations, the breakeven point is 90% and in others it’s 60%. Clearly, there are times when you must run.
http://www.retrosheet.org/boxesetc/B10170BOS2004.htm
On Oct 17, 2004, the Redsox were trailing the series 3-0, it’s the bottom of the ninth, and down by 1 run. With no outs, Roberts came in to pinch run. At the plate was Mueller, and pitching was Mo. How certain did Roberts need to be to attempt the steal? With two even teams, the chance of winning for the Sox at that point was .353. Getting to 2B would increase it to .462, and an out brings it down to .124. These numbers should be modified since Mariano was the pitcher. A cynic would say that the numbers should be bumped up. As well, a steal of 2B brings with it the chance of an error. Sticking with these numbers, a steal adds .109 and a caught stealing removes .229, making the breakeven a lowly 68%. Given that Mo was the pitcher, and it’s the bottom of the order at bat, the breakeven point is probably a bit lower.
The leverage index (LI) of this situation was 5.4, meaning that everything that happens here is maginfied more than 5 times. It’s as if you have 5 at bats all rolled into one. It wa a very high leverage situation, and the breakeven was very favorable for even an average runner. It’s situations like these that speed was made to leverage.
Here is something I’ve wondered about for a long time, dealing with the way Lwt run values are calculated for on-base events.
Let’s say MGL goes thru the PBP database for 1955 and finds 1b=.47, sb=.19, and cs=.45. So, the sb break even point is 70%, but maybe the actual sb% then was only 60%. So, the .47 for the single includes the avg runs lost on sb attempts following a 1b. So, it seems like there is a double penalty for the sb attempt. IOW, if your Lwts events inclusde sb/cs, you may want to use a ‘sb attempt’ excluded value for 1b, such as perhaps .475 instead of .470 (given that low 60% sb%).
This would also apply to gdp lwts. Baserunning lwts, since it is calculated using the actual advance rates, should not have this problem.
Or, maybe I’m unclear on how the PBP calculation process actually works…