Thursday, February 4, 2010

Predicting HR/FB Rates

A big part about knowing how a pitcher should do the following year, is knowing what his HR/FB rate will look like. It's understood in the sabermetrics community, that a pitchers HR/FB rate is mostly out of the pitchers control. This is to say, that it's mostly a factor of luck, and park based factors.

There are equations out there that try to normalize ERA based on a league average home run rate. This is not an accurate way to predict a future HR/FB rate. Someone pitching in a homerun friendly ballpark is obviously going to allow more home runs then someone pitching in a non-homerun friendly ballpark.

Likewise, some equations take into account park based factors as well. This is getting better, but it's still not perfect, because there are player based factors to factor in as well. What I mean by this is the following: Consider that ryan howard switches from the NL to the AL. Now a pitcher in the AL has to face ryan howard a few games a year, and his likelyhood to launch the ball over the fence is much higher then your average player. Now consider a change such is made to one of your division opponents, or better yet, consider that their roster is likely to change quite a bit. In reality, this is the case, and probably accounts for a lot of the variance in pitchers HR/FB rate differences from year to year.

I've attempted to determine a HR/FB rate for each ballclub. Theoretically, plugging this estimate in for each pitcher on a given club, should give you a good idea as to what their HR/FB rate should look like next year.

A achieved this by putting together a sample of data, and running some statistics against it. First I determined that using weights of 100, 66, and 33 for the previous 3 years respectively yielded the best results (the relevancy of HR/FB rate seems to fall off the further back you go). Then I took a group of players who had a significant amount of innings pitched, and played for the same club for the previous 3 years. Using this data, I attempted to determine what the most accurate way to predict the 2009 HR/FB would have been, using 2008 and older data.

My conclusion was that using a pitchers 2008 HR/FB as a predictor was poor. Using his previous 3 year average was equally poor. Using my "club factor" proved to be significantly more accurate at predicting the 2009 HR/FB rate.

Now without further a due here's the 2010 predicted HR/FB rates by ballclub:

Reds 11.82
brewers 11.54
Yankees 11.46
Astros 11.34
Orioles 11.29
Nationals 11.13
Phillies 11.04
Blue Jays 10.95
Tigers 10.81
Rays 10.78
Rangers 10.51
diamondbacks 10.3
Indians 10.26
Marlins 10.24
Rockies 10.21
White Sox 10.2
Twins 9.96
mariners 9.91
Padres 9.88
Cubs 9.86
Royals 9.8
Cardinals 9.79
pirates 9.71
Angels 9.62
Red Sox 9.36
braves 9.1
Mets 9.06
A's 8.81
giants 8.77
Dodgers 8.62