JFC,

Welcome to the discussion (flawed as it may be)!

In THEORY Kelly is fantastic and, as you point out, gives the optimal staking to increase you bank at the fastest rate. In PRACTICE however there are several major problems with it that limit its usefulness for horse racing and sports gambling.

Firstly the Kelly formula requires you to know the “edge” you have on each bet. As far as horse racing goes this means you must be able to say with certainty that your selection has say a 57% chance of winning the race so that you can work out your edge based on the odds available. Now maybe you are such a good racing analyst that you can come up with specific numbers like that but I am ************ed if I can. The best I can do is come up with an approach that says if I back certain horses within certain odds ranges then in the long run I will end up with say a 25% advantage IN THE LONG RUN. Now whether I have a 25% advantage on a specified runner, or a 50% advantage or only a 5% advantage I really don’t know with any certainty so what value do I plug into Kelly? At one time I was trying to come up with a specific price for a runner and then only bet when it was over that price but do you know what I found – when runners started well over that price (so in theory I had a big advantage) then my strike rate was poor but when they started below that price (so no bet as no advantage at all) the strike rate was very good. Had I been using Kelly I would have gone broke real fast!

The second problem (in a way linked with the first) is that the Kelly formula can give huge bets when your selections are at short odds (so place betting, sports betting, etc). For example lets take a sports bet which you assess as a sure thing but the bookies are offering even money on. Lets say you assess your edge as 50% then your bet is .5/1 = 50% of your bank Do you really want to be risking 50% of your bank on a single bet? Are you that sure of your advantage? Using full Kelly on bets where the odds are frequently in the odds-on range results in your bank jumping around like a hyperactive kid with a sugar rush – you can easily double you bank one day only to see it down to half of the starting bank a couple of days later. (This where the half Kelly and quarter Kelly come in).

So I would be interested to hear how you accurately assess the edge for every single bet and if you have done any past analysis on how good these assessments have been in practice?