Quote:
Originally Posted by JerryBoyle
Been lurking around the forum for a while without contributing, and this post got me thinking. I like the idea, but wanted something that penalized/rewarded runners relative to the rest of the race. Inspired by BIC, and using the public odds, I've come up with:
(ln(p) / sum(ln(pj))) - (ln(1-p) / sum(ln(1-pk))) where p is the runner we're scoring, pj are the runners we've beaten, and pk are the runners to which we've lost.
Looking Race 1 at GP yesterday (2-23-18), the scores would have been:
HTML Code:
Program # Finish Public Prob Score3 1 .047 10.738 2 .244 -1.882 3 .065 4.981 4 .057 3.674 5 .047 2.615 6 .244 -20.047 7 .011 2.836 8 .296 -60.29
I plan to use it in a regression to try and determine significance, but so far it seems to capture the idea that the OP is after. Feel free to refine/improve
|
Thanks for coming out of lurking to suggest this - seems to be onto the right path for what I was thinking.
Thanks also to others who followed up. Some of the other ideas which only used winners I don't think fully capture what I was after, I think all finish positions are useful in a review of this nature.