For benchmarking our system against others (because it is statistically non significant to compare to just buy and hold ), we are implementing an engine which can simultaneously test multiple trading systems at once (so that it will avoid to read the same datas again and again above all when it concerns intraday datas). My idea is also to draw a picture of the "ecology" of trading systems : at each instant how many systems (moving average systems, pivot systems etc.) herd together for example ? This could become by itself an indicator.
You can do that with some of the existing trading packages. I have tested multiple systems on one issue before, I could get the trades to show up on the chart with an identifier as to which system took the trade, very easy to then see which systems are running together. Max
Yes I know that some packages could do it nevertheless none can really make calculation for my model. And it's just an idea that I give for fun - as I don't need to trade any indicator at all but my own model. Also I should be able to do hundreds or thousands of trading systems calculation simultaneously if I make the parameters vary. This would be not easy with a traditional package at least simultaneously.
In some neural book I saw they do that with plenty of agents nevertheless it's very basic as for strategies employed : it doesn't correspond to real world.