- Banned
- #1
Hi all, long time listener first time caller, I am interested using multi nominal logit regression for horse racing data. Anyone had experience in using these? I'm told there was a free program called R once upon a time but don't think it is supported on windows 10.
For those who have experience in this type of analysis I have the following:
One independent variable, whether the horse won or lost, a 1 or a 0. So this essentially acts as the final probability in the race, the winner had a probability of 1, all others in that race had a probability of 0.
One race identifier field, numeric, simply 1,2,3,4 up to several thousand.
Lots of fields of numeric quantative data.
I have had a look at XLStat and fiddled around with their program and they have a multinomial regression model in there, and I ran it over the data. However I couldn't find a way to use the race identifier, so in the end all the regression program is trying to do is fit the data to predict the winner as close to 1 as it can. This is not right. The race identifier needs to be used so that the score each horse is given in each race is then adjusted back so that each race totals 1.000 in estimated probabilities.
Know what I mean?
For those who have experience in this type of analysis I have the following:
One independent variable, whether the horse won or lost, a 1 or a 0. So this essentially acts as the final probability in the race, the winner had a probability of 1, all others in that race had a probability of 0.
One race identifier field, numeric, simply 1,2,3,4 up to several thousand.
Lots of fields of numeric quantative data.
I have had a look at XLStat and fiddled around with their program and they have a multinomial regression model in there, and I ran it over the data. However I couldn't find a way to use the race identifier, so in the end all the regression program is trying to do is fit the data to predict the winner as close to 1 as it can. This is not right. The race identifier needs to be used so that the score each horse is given in each race is then adjusted back so that each race totals 1.000 in estimated probabilities.
Know what I mean?