from Amazon
Thursday, August 03, 2006
Spread In Talent, Pitching
Following the same process I laid out for hitting, I did the same for pitching. This time, I focused on batting average on balls in play (BABIP).
Those who have read Solving DIPS will find no surprises here. In this first study, I kept my focus on AL pitchers, 1973-1996 (so as to not worry about the opposing pitcher hitting). That’s 24 seasons.
I selected the 100 pitchers who allowed the most number of balls in play for each season. This came out to an average of 500 BIP per pitcher. The observed standard deviation was .024, while the standard deviation we can attribute to random chance is .022. You see where I’m going here? Just by luck, we should expect to see some random scattering. That amount of scattering is represented by the .022. On the other hand, what we actually observe is a smidge higher at .024. So, this is the indication we have that what we see is not random. And, in fact, we can establish how much of it is not random. Such a distrubution was caused by pitchers where the true talent of their abilities (plus their fielders and parks) to get outs on balls in play has a standard deviation of .011. That spread in talent is one-third that of batters abilities to get on base.
Now, let’s do this for all of baseball history. This time, I’m including all the leagues. And, I’m going to select a number of pitchers equal to 6 times the number of teams in the league. In the 1973-1996 time period, which now includes both leagues, I end up with a true spread of .010, which is consistent with what I found with the AL-only pitchers.
Anyway, how does it look when going back in time? Starting from 1905 until today, the five-year average was between .008 and .012, for every single 5-year period! That is, the spread in talent in a pitchers ability to get outs on balls in play has not changed.
From 1895-1904, the spread in talent was .017.
Now, I should caution you that I have not yet distinguished between pitchers, fielders, and parks.
The equation we have is:
.010 ^ 2 = pitchers ^ 2 + fielders ^ 2 + parks ^ 2
Depending on how the true distibution of fielders talent and park effects were year-to-year, would affect our view as to the true distribution of pitchers talent.
Nonetheless, if we give pitchers a spread of .008, that gives us about 1 standard deviation is 4 hits per 500 BIP, or that 95% of pitchers give up hits +/- 8 from the league mean.
Alot of that spread is based on the GB and FB tendency of pitchers. More outs on FB, but more extrabase hits on FB. If we looked at total bases, as opposed to jsut hits, that spread would likely be even smaller.