JOwl
sum guy
Posts: 2,694
Joined: Jun 2005
I Root For: Rice
Location: Hell's Kitchen
|
RE: Baseball - The first half season in numbers
(03-30-2014 12:58 PM)waltgreenberg Wrote: (03-30-2014 11:16 AM)temchugh Wrote: (03-30-2014 10:51 AM)MemOwl Wrote: (03-30-2014 09:53 AM)temchugh Wrote: The r-squared is low, but this just means that there are a lot of factors in addition to SoS that affect OBP. The low r-squared does not mean that SoS is not important.
what is the t statistic for SOS?
Sure, make me go to the "data analysis" package in Excel.
For SoS as a predictor of OBP, the p-value is almost exactly 0.05 (p = 0.05004 according to Excel). In other words, there is a 5% chance that the apparent relationship between SoS and OBP is due to random chance. If you account for the fact that I predicted a positive relationship (OBP increases with increasing SoS) and observed a positive relationship, the probability drops to 2.5%.
This is taking statistics to the absurd level. You do realize that SoS is a sequential ranking and is NOT reflective of an absolute differential between teams, as OBP, AVG, SLG, ERA and other stats are. In other words, there might be very, very little difference in SoS between the #20 ranked team and the #35 ranked team...and the difference between the #50th ranked team in SoS might not be that much better than the 95th ranked team. To even attempt to calculate a correlation here is a bit ridiculous.
It's not ideal, but it's far from absurd. Having Boyd's SoS valuations rather than his ordinal SoS ranking would be better for sure, but it's not useless. Also, you raise an interesting point about the pattern of team quality; it might not be linar. #1 might be as much better than #10 as #10 is better than #100, or something like that. In which case temchugh could just perform regression against log of SoS.
But I think he's already done enough.
|
|
03-31-2014 05:10 PM |
|