Well, I'd like to do what you're doing -- tick-by-tick stuff -- just to increase the overall trade frequency. I've got a set of working 5-minute bar strategies that consistently make money and backtest well; however, what I want to do is apply the working model that I have and somehow increase the trade frequency to higher time frames, maybe on the order of seconds. I've got a few dilemmas I'm working on though: 1) Don't know how to generate "lists" of PnL results for trades done on backtests, waiting for a forum response. Want to do this using EL, not by manually having to export the data. 2) Haven't figured out how to get around the fact that I can't have separate automated strategies for the same stock. 3) Haven't figured out how I can successfully backtest strategies at the per-tick level; mostly because I can't "combine" charts together at the tick level, and it looks like I have to write some kind of DLL to do this. 4) Slippage is very high; I can come up with model after model that is tolerant of high slippage, but the more tolerant the model becomes of commissions and slippage, the less opportunities I have. 5) Can't script or control back-testing after automatically creating windows. This takes a really long time to do manually. Since my current trade-decision engine sees opportunity only once every few days per basket, this is becoming a bottleneck on short-term growth. 6) Can't programmatically request all the market data with a script. Someone posted to the forums showing me how, but I've yet to implement this. My biggest problem is that I want to take working concepts and generate a small stream of revenue quickly (through EL automation) so I can finance taking time out to write real, rigorously tested C++ code that can be run independently of any platform and connects with execution engines separately. So, I'm hitting all kinds of barriers and the turn-around time on the TS forums is like a day or two for response from the vendor. Portfolio-level testing -- I'm confronted with the task of rewriting stuff in EL for other platforms, given what I'm reading in the forums. I'm wondering whether it is worth the time to commit to writing something that can do the work in python. (Even if you don't have solutions for the above, I wouldn't mind others chiming in with solutions.)
Have you looked into Multicharts? I know they offer portfolio level testing, not sure about the rest of your needs. I will offer this for encouragement: If it's tough for you to implement something, then there's probably a pretty decent edge there. Good luck to you. D