Question to practitioners of high-frequency trading strategies on futures; how do you model fill probability in your backtest? For example, in equities i can simply assume fills only when price trades through my limit order. In futures, especially indices, a lot of the volume is done on the top/bottom tick. How do you adjust your tests to account for realistic fills?
If you want the conservative approach, assume the same thing as you do in your stock trading - price must trade through. You'll find that especially true on the ES.
Many of the strats are high-frequency which exhibit small avg winner, large losses, and high probability. I also utilize limit orders so slippage of a tick would likely kill all the strats (not to say thats wrong). Similarly with completely removing bottom/top ticks because a significant number of winners and frequency occurs at these prices (more so then equities due to tick size) An idea: Break down into smaller time-scale and analyze volume at price and apply some rule of thumb on probability of fill at that price. For example, only fill if volume @ lmt price > x.
Taking off a % of profit is still not a clear test, the resolution for drawdowns, etc. would not be available.
Another Question: Interactive Brokers simulator replicates your order and place in line in the order book so it makes an effort to simulate accurate executions, does TradeStations Simulator do this as well??
Questions is : 1) in which markets do you want to apply "high-frequency" trading ? Commodities / financial futures ? 2) Average volume / lot size = AUM ? 3) Margin / equity ratio ? 4) intial lot size ? 5) Scaling in / scaling out ? Yes / No ? ( suggesting we are not talking about single diggit lot sizes )...