Wednesday, October 28, 2009

What Does The Past 3 Years Tell Us About The World Series? (updated)

I tried to use a tool of Tangotiger's called Marcel. There is a good chance I did not apply it correctly (I think I did, that is why it is updated) but I attempted to measure the skill level of the players using the last 3 years of performance with more recent years being weighted more and using a regression to the mean. Maybe in a day or two I will go through the numbers.

First I tried to generate an OPS relative to the league average for the 8 position players on each team. Then I took the simple average of that. For the Yankees, it was 10.13% above the league average. For the Phillies, it was 7.54% better. If we assume a league average of .750, then the Yankees would be at .826 while the Phillies would be at .807. I did not make any park adjustments and this might hurt Ibanez for his two years in Seattle.

For pitchers, I did the same thing using the FIP ERA from Fangraphs. Here are their ratios to the league average for the top 3 starters

Sabathia 0.751
Lee 0.829
Burnett 0.945
Martinez 1.052
Pettitte 0.907
Hamels 0.881

The Yankees have a big edge in the first and second matchups while the Phillies have the edge in the 3rd one. But Fangraphs has the same league average each year for both leagues. This may not be right, and if not, it would probably mean that the Yankees have an edge in all 3 slots. Then, as I mentioned yesterday, the Yankees were much better against lefties this year than the Phillies.

As I understand the Marcel method, the last 3 years have a weight of 5, 4, and 3. So the total is 12. Then the weight is

5/12 = .417
4/12 = .333
3/12 = .25

Now I think that assumes that the player has an equal number of PAs in each year. But I think those weights should be changed if PAs are not equal in each year. Let's take Jeter. Here are his PAs from 2007-9


The total is 2049. In each year, here are the %'s of the total for each year:


Now how does that change the weight of 5, 4, 3? In 2009, he had a larger than expected pct (which is .333). The pct was .345/.333 = 1.036 times the expected value. So instead of using .416 for 2009, I used .416*1.036 = .357. Something similar was done for all the other players. Here are Jeter's OPS divided by the league average from 2007-9


So what is his ratio for the 3 years? 1.11*.339 + 1.02*.316+ 1.14*.345 = 1.097. Now the regression to the mean. First multiply the PAs from the 3 recent years (going backwards) by 5, 4 and 3. That gives 8207. But the regression to the mean involves two seasons worth of league average hitting. Each year is 600 PAs or 1200. So we have a denominator of 9407 (8207 + 1200). So Jeter's OPS relative to the league average is

(8207/9407)*1.097 + (1200/9407)*1 = 1.0846

Jeter's skill level means his OPS is 8.46% better than average. So I did this for each player on each team. I added their relative to the league average and then divided by 8. I did something similar for the starting pitchers.

No comments: