tanda
06-20-2001, 09:53 PM
I am looking for input on oddsline algorithms. I understand that many people may be protective of any insights in this area. However, any thoughts are appreciated.
I have been using a modified version of the formula/algorithm described by Dick Mitchell in Winning Thoroughbred Strategies.
That process involves indexing the ratings to a power rating and processing the power rating through another formula to generate an ability rating. The above average ability ratings are given an 80% chance (contenders) and the below average are given a 20% chance (non-contenders). The 80% for the contenders are distributed in proportion to their respective ability ratings.
One of my modifications allots .40 fair chance to a non-contender (so 4 non-contenders in an eight horse field would get a combined 20% chance i.e. 4/8 = .5 *.4= .20). As with the Mitchell method, my method often gives 20% to NCs but usually a little more or less.
I have been experiementing with squaring the ability rating to generate more of a spread between the horses and further experimenting with raising the ability rating to an exponent determined by the size of the field (the larger the field, the larger the exponent) for the purposes of spreading the ratings a bit. With indexing as used by Mitchell, large fields tend to get horses bunched with smaller than accurate (I believe) spreads.
Also, I see the Del Mar Throughbred Club sells an oddsline generator that is inexpensive. Any insights on that? Any references to probability/mathematics texts/software would also be of interest.
Thanks, tanda313@aol.com
I have been using a modified version of the formula/algorithm described by Dick Mitchell in Winning Thoroughbred Strategies.
That process involves indexing the ratings to a power rating and processing the power rating through another formula to generate an ability rating. The above average ability ratings are given an 80% chance (contenders) and the below average are given a 20% chance (non-contenders). The 80% for the contenders are distributed in proportion to their respective ability ratings.
One of my modifications allots .40 fair chance to a non-contender (so 4 non-contenders in an eight horse field would get a combined 20% chance i.e. 4/8 = .5 *.4= .20). As with the Mitchell method, my method often gives 20% to NCs but usually a little more or less.
I have been experiementing with squaring the ability rating to generate more of a spread between the horses and further experimenting with raising the ability rating to an exponent determined by the size of the field (the larger the field, the larger the exponent) for the purposes of spreading the ratings a bit. With indexing as used by Mitchell, large fields tend to get horses bunched with smaller than accurate (I believe) spreads.
Also, I see the Del Mar Throughbred Club sells an oddsline generator that is inexpensive. Any insights on that? Any references to probability/mathematics texts/software would also be of interest.
Thanks, tanda313@aol.com