"I've achieved 80%, but those systems don't make enough trades to give a great annual return."
Actually, the system I was thinking of generated hundreds of trades per tradable per year - not that it matters much, because, as I said, you couldn't really trade it. The concept wasn't bad, however. In more tradable versions, it worked decently well for an intraday system.
<<What function do you seek to optimize when developing a system?>>
I also used to optimize for net profit first. I then turned to optimizing for profit factor (gross profit/gross loss). Lately, I've been testing first for return on maximum intraday drawdown (ROMID) - partly because I've found that it usually equates with a relatively smooth, rising equity curve, and partly because the highest ROMIDs usually occur in conjunction with higher NPs and PFs. I also want a good read on worst-case scenarios, though I'm aware that there's always a "worse-than-worst" case out there lurking. No trading system can protect you from getting hit by a bus either.
In general, I try to get a read on every factor that I can think of examining - avg. net gain, win/loss, number of trades, and much else - but I have to confess that I don't customarily apply chi-square tests to my results. I can imagine how one might do so, but I'm not absolutely clear on how you're using them. |
