Just a question for you guys: how much statistical edge per trade are you looking for per basket/pair in your backtest? I've done cross-validation with a strategy to find out it works far back into the future, added slippage of 3 cents on both sides, and hefty commissions ($1 per side), and have an edge per trade of around $10 per trip. The strategy is based on intraday data. Is this something you would trade, if the equity curve is relatively smooth?
The book says that if you have a statistical edge, enough capital to cover the drawdowns, and your rate of return is high enough for you (10% a month? 1% a day? whatever your goal is), then you should go for it. Me, I'm working on a strategy that has a very high win ratio, even at the sacrifice of big profits. I would rather have 4 of 5 be winners at 1 dollar per win and 1 dollar per loss, than 1 of 5 be winners at 10 dollars per win and 1 dollar per loss. Even though the net is greater with the latter. I'm new to automated trading, and need to prove to myself (and the wife) that I can make a profit.
Hmm, seems like you are in a similar state of affairs as me. Perhaps we should PM and discuss what our experiences are regarding these systems, and whether the edge we seek is an illusion or obtainable.