Fully analyzing a strategy or an EA to the point of being confident about letting it to touch my money is time consuming. There are thousands of free and commercial strategies available, and analyzing all of them would be unavoidable.
Fortunately, it is easy to discard more than 95% of the strategies almost without committing any time to them. I have some essential rules that all long term profitable strategies needs to comply with. Failing to comply to any of the rules will make me discard the strategy and avoid wasting time with further analysis. Obviously complying with all requirements don't make the strategy long term profitable, it just allows me to go to the next step.
My list of long term profitability requirements
My list of the most important long term profitability prerequisites helps me discarding the majority of the EAs and strategies and focus my time in more important things. It is important to note that this requirements are previous to start any analysis. Whenever I didn't follow all this rules I lost money.
Here I list the most important profitability requirements
- Surviving a 12 year backtest: I want strategies that survives as much different market conditions as possible.
- Trading all signals: It is essential to avoid hiding drawdown periods behind opened trades and to avoid trade chain dependancy (different results depending on when you started trading)
- Not requiring tick data to backtest accurately: If a backtest with tick data the strategy shows much better backtest, it is definitely curve fitted to the data. Tick data usually don't have separate Bid and Ask prices. And tick data don't take into account the swap, which was different historically.
- Trading on M15 or above: M15 is the limit, as long as it uses various bars to generate a signal. Below M15 there is simply too much difference between brokers to trust the strategy.
- Symmetric strategy for Long and Short: If the strategy has different strategies for longs and shorts, it is curve fitted to specific market conditions. For example, giving more weight to shorts on USDJPY would have worked for the last decade, but nothing guarantees that this will keep being the same. The same happens going only long on GOLD. It is suicidal from a long term profitability point of view.
- Less than two or three indicators / freedom degrees: With enough number of indicators and rules, you can make any strategy to fit perfectly the past. And this guarantees a very quick fall (weeks or even days) once you start tradeing the strategy.
- Trading at the close of the bar, at least for entries: There are too many backtesting tricks/errors that can be done trading at the open of a bar. And if a strategy can trade at any moment, it will fail to comply on rule #3 too.
- Not having a straight up incredibly nice equity curve: Drawdown periods are essential and healthy for a long term profitable strategy.
- Not using position sizing tricks such as martingales or grids: They cannot be statistically analyzed through a backtest and, obviously, from a long term profitability point of view it's absurd to risk your whole capital on every trade.
- Not using fixed targets in pips: 100 pips when price is at 0.7000 is not the same as 10 pips when price is at 1.4500. Targets have to be adapted to market conditions.
- No scalping, targeting a decent amount of pips: When a strategy scalps the market for a few pips, it cannot be accurately backtested. I'm not saying that a scalper cannot make money, I say that I just cannot be sure. So I cannot trust it to manage my money on the long term.
- Trading enough number of trades: The more trades the strategy does, the more robust will be the system. Very few trades means that it is easy for the system to be curve-fitted to the data. Anything below 100 trades/year for 15 minute charts or 50 trades/year for hourly charts should be discarded.
Failing to comply with any of the above will make me discard the strategy.
Update 1 (03/2012): Added rule 12. It was an obvious rule that I was applying but I didn't realize it was not explicitly in this list.