money mgmt strategy utilising MCMC and confidence intervals?

[brettd]

I just want to run this past someone before I get around to doing this. But would this work as an effective money management strategy?

• So firstly, utilise Monte Carlo to formulate simulated distributions of betting runs, with various edge parameters, kelly parameters, etc. Accurately created from past data.

• Plot all the simulated distributions along a horinzontal or 'x' axis.

• Consider the actual betting run as an ever growing 'sample' and apply desired confidence intervals to this sample.

• Plot the range of the confidence intervals along the axis of simulated distributions, until the sample is large enough (and the confidence intervals are narrow enough) to be within the bounds of a single distribution.

We can then assume the betting parameters of that distribution is optimal for the actual betting run.

Hopefully someone understands what I'm on about here.

Wrecktangle? Or any other stat guru know whether I'm on the right track?

[tomcowley]

data mine to find variance, assume it will still outperform in the futue., go broke when it doesnt. book it

[Flying Dutchman]

Unless you have play-by-play information, or are simulating a sport like MLB, Monte Carlo will likely be more trouble than it is worth. Getting a historically accurate handle on the distributions involved with ** without a pbp db will lead you to guess at some distributions and can likely lead you to pick a Gaussian (Normal) as a default in many cases which can easily be wrong in sports.

Not sure where you are going with the rest of your question. Looks interesting though.

[brettd]

I'm not modelling the matches. Simply the results of betting them. I've got enough of a betting history to correctly structure a simulated betting run of x bets on most sports I bet on. I'm just wondering whether applying confidence intervals to the actual betting run is the best way of comparing betting reality to the ** simulation.