03-20-2018, 03:26 PM
I'll admit I haven't played around with those parameters too much...
Population amount is the number of individual tests to do per backtest.
Each epoch, a number of totally new, random parameter sets are added - this is defined by variation * population amount.
The rest of the parameter sets are generated by mutating a few of the parameters from the best set yet. Mutate Elements set the maximum number to be changed at once.
Parallel Queries is simply how many backtests it runs in parallel.
For efficiency it is probably best to set parallel queries equal to the number of cores in your computer, and potentially to make population size a multiple of parallel queries so that you can always run as many queries as you have cores - however as backtests take different amounts of time depending on the candle size, there will always be a case when the CUP utilisation drops off towards the end of each epoch.
Population amount is the number of individual tests to do per backtest.
Each epoch, a number of totally new, random parameter sets are added - this is defined by variation * population amount.
The rest of the parameter sets are generated by mutating a few of the parameters from the best set yet. Mutate Elements set the maximum number to be changed at once.
Parallel Queries is simply how many backtests it runs in parallel.
For efficiency it is probably best to set parallel queries equal to the number of cores in your computer, and potentially to make population size a multiple of parallel queries so that you can always run as many queries as you have cores - however as backtests take different amounts of time depending on the candle size, there will always be a case when the CUP utilisation drops off towards the end of each epoch.