Bot Betting Edge

Bot Betting Edge

The Blog

This blog will contain my reflections and progress within the field of modelling and predicting sports probabilities, especially for soccer matches inplay. Although placing bets on different matches I don't consider myself as a bettor, I see myself as an investor placing money where I have edge and finds value. Sports betting is fantastic!

Optimizing betsize using Kelly criteria

Money managementPosted by Tobias Bramhed Mon, January 14, 2013 19:59:03
I will take a few minutes discussing which bet size to use. Within my current betting strategy I use a flat bet stake, meaning I put down the same amount of money on each bet without considering factors such as the implied probability of winning/loosing the bet or the value of the bet (my calculated overvalue).

So far I have only done a few simulations of portfolio developments and from those chosen a bet size which seems to avoid ruin. My current bet size is about 0.5% of the portfolio, but as I mentioned I keep it flat. My original plan was to update it once every quarter to get a slow and stable increasing risk take.

Lately I've been looking into optimizing the bet size, something that is done within both the betting and investment communities. The formula to determine the fraction of portfolio for each bet is given by:


p = my estimated probability of the bet
b = the odds for the event given by the market
f = fraction of portfolio to use as bet

Of course curious to see if I would get any effect on my historical portfolio if I had used Kelly bet sizes instead. Simulation time! And this is what I've found out:


Wow! It seems mad and of course I then tend to be skeptic.... Using smarter bet size would have taken me from ROI -0.4% to +6.0%. To good to be true? Probably. I wont implement this right away due to the fact that I have several new strategies in place this month, but in the beginning of next month when the new strategies have been evaluated I will investigate the theory deeper, make some more simulations and THEN implement if it still shows good result. I will get back to this topic for sure :)

  • Comments(0)//blog.bramhed.se/#post4