Strategy 1 Execute Simulation with traders
Input the values of your EA so that
• this values should be as accurat as possible
• this values should come next close to the original ones
• this values may deviate from the original ones
• when using both strategies simultaneously, then use for each one half of its calculated lotsize
• this values should be the most imortant ones when comparing strategies
1http://en.wikipedia.org/wiki/Kelly_criterion
2Instead of using Lotsize ~ Balance (B), with Account protection (AP) one of the folowing CRRA utility functions is used
pow: Lotsize ~ ((B/B0+1)(1-AP)-1)/(1-AP)
log: Lotsize ~ ln(B/B0*ln(1/(1-AP))+1)/ln(1/(1-AP))
3 valid for account balances with money equivalent of 1 lot x 1.000 pips, i.e. balance=10.000 \$ and lotsize 1 stands for 1 pip = 10 \$
4 The calculated values are correct, when there are sufficient many trades (otherwise the binomial distribution must be used) and when there is no account protection, so the distribution is log normal and the median is equal to the geometric mean. For log-normal there is mean>median, for Kelly fraction->0 the distribution becomes normal and mean->median. With certain combinations of Kelly fraction/Account protection it is possible to transform the log normal distribution into a more favourable one with a higher median by abandoning the high mean caused by the "fat tail" which is "mainly meaningless".

Disclaimer: