Bet Sizing Alternatives to Kelly

[TravisVOX]

While the Kelly criterion is a sound mathematical approach to wealth generation, in horse racing the true final odds are never known until after wagering has closed. Due to the antiquated system that processes horse racing bets, "odds changes" are a regular occurrence. A horse can be 8-1 while they load into the gate, only to end the race at 7-1.

Furthermore, as you step into exotic wagers - Exacta, Trifecta - the odds are even harder to determine.

Thus, I'm wondering if anyone can share or knows of other systems that are not purely arbitrary but have some logic to their approach when it comes to bet sizing that doesn't totally rely on the payoffs or odds?

[Dr.Gonzo]

Bet according to your ratings, to win a fixed percentage of your bankroll. You will need to determine how aggressive or conservative you wish to wager, obviously the higher the percentage the more aggressive. I will assume you will only bet betting probably overlays.

Example: You mark a horse at 5-1 and it is showing 8-1, you back it as if it were a 5-1 shot to win 2% of your bankroll. If it were to start 5/1 you would expect to break even in the long run, but anything over 5-1 is +EV.

You can tweak it a little and create categories. Small expected overlay (1%), medium expected overlay (2%), large expected overlay (3%). If you go down this road keep separate results for all categories, you might find that whilst you are able to identify an edge your large expected overlays only have the same edge as your small expected overlays, alternatively your small expected overlays may be losing propositions but your medium expected overlays are profitable and large expected overlays even more so. Thus you can adjust based on your results.

[Waz]

Betting Kelly is difficult because quantifying your exact edge is extremely difficult IMO. I like to use the half unit to 2 units approach, based more on feel and intuition rather than an exact calculation.

[TravisVOX]

Exactly, Waz. That's why I was wondering if there were some other approaches out there.

Appreciate the thoughts Gonzo... definitely a way to tackle this. I already started coding it into my software as one approach. I even extended it for the exotic wagers as well. I'll let everyone know how it turns out.

[trytrytry]

Bet according to your ratings, to win a fixed percentage of your bankroll. You will need to determine how aggressive or conservative you wish to wager, obviously the higher the percentage the more aggressive. I will assume you will only bet betting probably overlays.

Example: You mark a horse at 5-1 and it is showing 8-1, you back it as if it were a 5-1 shot to win 2% of your bankroll. If it were to start 5/1 you would expect to break even in the long run, but anything over 5-1 is +EV.

You can tweak it a little and create categories. Small expected overlay (1%), medium expected overlay (2%), large expected overlay (3%). If you go down this road keep separate results for all categories, you might find that whilst you are able to identify an edge your large expected overlays only have the same edge as your small expected overlays, alternatively your small expected overlays may be losing propositions but your medium expected overlays are profitable and large expected overlays even more so. Thus you can adjust based on your results.

adding Dr Gonzo to the must read poster category..well described here.