from Amazon
Wednesday, February 02, 2011
The HR per FB skill
There is no such thing as randomness if it involves human beings making a decision. So, if someone were to ask me: “Does the clutch skill exist? Does the platoon handedness skill exist? Does the day/night split exist? Does the park skill exist? Does the hits-better-with-a-fast-runner-on-base skill exist?” The answer must be yes. All these things must exist because humans are involved, and they are asked to respond to some environment. Let’s forget about asking “Does it”. The question to ask is: “How much does it” and “How can I find it”.
Take the HR per FB skill. Of course it must exist. The real question is if we can find this skill. Dave Cameron starts us off with some great data, where he takes the pitchers with the lowest HR/FB rate over a period of time. Those guys averaged 8.6%. In the post-sample period, they averaged 9.9%. The league mean is 10.6%. This means the regression rate was 65%. Since these guys averaged 1135 innings in their in-sample period, I’ll take a guess that the number of flyballs was an average of around 1200 flyballs.
Our regression equation follows simply enough:
regression rate = 2200 / (2200 + FB)
With 1200 FB, regression rate is 65%. In order to regress only 50%, you need 2200 flyballs (or about 10 years worth of full-time pitching). Based on this limited amount of data, it’s easy for us to say that there’s “little” HR per FB skill.
Dave also gives us the other side: those with observed bad rates (average of 12.2%) in-sample had an out-of-sample observed of 11.4%. That sets the regression rate as 50%. Dave doesn’t note how many innings we are talking about, but let’s take a guess that it’s about 750 innings (just throwing a number out there for illustrative purposes), which might be 800 flyballs. The equation explains itself in this case:
regression rate = 800 / (800 + FB)
Quite a big difference. When I did my more elaborate view of the HR per FB skill last year, I think I came out with a number like:
regression rate = 600 / (600 + FB)
Something like that. This requires more study because we have to control for the pitcher’s parks (what may look like a HR per FB skill is actually a bias from the park instead).
Anyway, the skill is there, it exists, and now, it’s just a matter of answering the two questions:
How much skill is there?
How can I find it?
These are the two questions that you should be asking of ALL OBSERVED DATA.
Recent comments
Older comments
Page 1 of 344 pages 1 2 3 > Last »Complete Archive – By Category
Complete Archive – By Date