if one has several strategies mechanical and discretionary, what is the easiest way to combine them? each strategy has its trades recorded (stock name, date, entry/exit price, profit, etc.). the goal is to get a superior risk/reward ratio by varying strategy combinations and the proportion of the account allocated to each strategy. is there a program out there that will allow one to mix and match strategies using the above outputs? or does everyone write their own software to do this?

you can overlay the profit curves on backtests for several strategies on CQG. there is a London based consultancy company called Oasis who can come and give you a lecture on how to do this. in my/our case what the guy did was took 3 strategies, one trend following, one counter trend and one range based, and showed us them individually - as you would expect they each had very volatile profit curves. he then overlaid them as if you were applying all 3 at once and they had the effect of smoothing eachother out. one caveat is that this required you to potentially run all 3 strategies at once, requiring 3 accounts.

but the idea is to work strategies when your discretionary intuitive mind (skillset) says "these mkt conditions" REQUIRE you to revert to range trading if desired and discard the trend strat for the time being. You adjust..........

With a little math background, this is not that hard to do. The idea is to think of the systems, each as a coordinate axis that is linearly independent of the others. In other words, you want the systems to span some vector space. The space here is probably the space of profitable trading decisons. Linear Algebra takes care of the rest. In theory what you want are systems that completely uncorrelated (pairwise correlation zero to each other) most of which should make money (there is a strange case where adding a losing system to several winning system actually smooths equity curves). Obviously, if most of the systems lose money, they will simply be uncorrelated to each other in how they lose money and having them be uncorrelated won't help. So you have to have "systems" or "setups" that have a positive edge, with each edge making desicions that uncorrleated to the other edges/setups. Since the systems form a vector space, you can be confident that the decision each makes has no correlation to the others (again, assuming no correlation at the time of the decision making, probably a big if since correlation is a rear view mirror statistic), your equity curve should follow some averaging process of the total profitability of the total systems. Notice that because we are dealing with abstract spaces, you can have ten systems or a million systems and still be able to deal with 10-dimensional space, or million-dimensional space. The algebra doesn't care. If the systems have non-linear interaction, which they almost certainly do, most of this has to be understood on a much deeper level. nitro

some sort of "heat mapping" product would be a great start for you... lets us know if you choose this path and find something...