from Amazon
Wednesday, March 30, 2011
Misleading those who want to be misled
I thought this entire blog post was pretty good up until the item in bold below:
There is no such thing as “Expected Runs”
One of the most misleading of all the sabermetric devises is the use of probabilities that a run will score or a team will win as a measure of individual results.The basic idea, made popular by a guy with the nom-de-plume of Tango Tiger, is how many runs to teams score a run or games do they win after a change on the field.
As an example. how many runs do teams score with one out and runner on first, compared to one out and a runner on second. How does that compare to to two outs and no one on base. This is used to determine the value of a stolen base, compared to getting caught stealing. You can see a recent example of this methodology here: http://www.twinkietown.com/2011/3/29/2078456/baserunning-was-it-really-a-problem-last-year .
You can take this a step further and ask - what is the impact on wins. For instance, You start with how often teams win with a score of 3-2 in the bottom of the 8th with no one out and no one on base. After a player doubles, you compare that to how often a team wins in the new situation, with the score still 3-2 in the bottom of the 8th with no outs, but now a runner is on second.
At first glance this seems like a reasonable measure. But it is no more reasonable than it is to say “How often do players hit home runs?” and then apply the result to Justin Morneau and Matt Tolbert. While the number of outs and runners on base are certainly factors in how often a team will score, who those runners are, who is pitching, who is in the bullpen, who is batting next .... are all at least as important. So when you make a decision to bunt or steal, you have a lot more to consider than what will happen on average in the new situation.
See, this is the bullsh!t I have to deal with. We wrote a book in 2006. In it was a 50-page chapter on bunting by MGL that specifically talks about NOT using average, that there’s a large number of parameters to consider, that you CANNOT just use the basic expected runs table. You have to apply adjustments.
Just because some people practice bad quantitative analysis doesn’t mean that we all do. Don’t impugn the use of the expected runs table as a matter of fact.
Recent comments
Older comments
Page 1 of 344 pages 1 2 3 > Last »Complete Archive – By Category
Complete Archive – By Date