forex-scalping.com

Comparing the backtest with a period of live trading on the same dates is a useful check if you have it. If you traded live for three months, run the backtest on those exact three months and compare trade by trade. Differences show where the backtest is too optimistic, like missing slippage, trades you'd have skipped in real life, or fills that wouldn't have happened. That comparison is often more revealing than any number of statistics, because it shows exactly where reality and the simulation part ways. Even a few weeks of overlap is enough to start seeing the gaps.

How much worse should live results be before you stop trading a strategy? I'd set a number in advance, based on the backtest's own variation. For example, if the backtest's worst three month stretch was a ten percent drawdown, a live drawdown of fifteen percent might be the point to stop and review. Without a number, it's tempting to keep going hoping things improve, or to quit too early after a normal bad patch. Deciding the threshold before going live takes the emotion out of that call when the time comes.

Setting the stop-review level in advance is the part I like most here. Before going live, I write down something like: if the live drawdown goes past 1.5 times the worst backtest drawdown, I stop and review. For a system that showed a 9 percent maximum drawdown, that's about 13.5 percent. With that written, a bad month doesn't turn into an argument with myself at 2am.

The bigger issue for scalpers is costs. A backtest at 0.8 pips of fixed spread and no slippage can look lovely. Say it takes 600 trades a year at 6 pips average target. Add 0.5 pips of real extra cost on each trade and you've lost 300 pips a year. If the strategy only made 450 pips in the backtest, a third of it is gone before anything else goes wrong.

I now run each test twice, once with clean costs and once with the spread doubled and a random slippage. If the second one looks bad, I don't go live.

How long do you forward test on demo before moving to small size? I've been using about 100 trades, but it's hard to say if that tells you much for a fast system.