OZmium Sports Betting and Horse Racing Forums - View Single Post

Not wanting to rain on anyone's good ideas but IMO the idea of (Monte Carlo simulations) being applied to horse racing assumes normal distributions and correlation coefficients of zero. In my mind this is not typical of race betting markets. The problem with using this approach to calculate probable future mathematical outcomes is that risk assumes that you have an understanding of the future outcomes which are the underlying inputs of any Monte Carlo simulation. Probably waffling now in terms of jargon but the bottom line is that the distribution of betting market returns are often not within the so the called normal range.

IMO and in my own application, a better way top quantify the likely range of potential outcomes in horse racing is to build a model (which we have done) that relies on standard deviation as the basis to measure how spread out the numbers are. The formula is simply the Square Root of the Variance. Variance is the average of the squared differences from the mean (its important to square the differences as if you just added them the negatives would cancel the positives).

Once you do the above you can then simulate your results over 1 or 2 lines of deviation which will tell you just how far your future results can and will vary. 1 line will tell you your results 68% of the time whilst 2 will cover them 95% of the time. That way you can learn the full range of probable outcomes both in terms of POT and actual real dollars profit/loss.

If anyone would like me to run a simulation for them and send them the results them I'm more than happy to oblige. All I need is long term strike rate, expected number of bets over the period, assumed profit edge and risk threshold (the % of your bank that you are willing to lose and continue betting without change) both in terms of a percentage.