Start with when you are at the screen
This is the one setting you cannot find by search. If you look at the market in the evening, use the hourly and four-hour timeframes; if you have half a day, 15 minutes works. Five-minute signals demand presence and a quick reaction: a signal noticed twenty minutes later is four bars stale.
Sessions are set in the same place: "London only" cuts off the Asian session, where the majors usually move sluggishly.
Then the instruments
Take the ones you actually follow. A strategy with eight instruments produces eight streams of signals, and the decision on each is yours, not the service's. Start with one or two.
Instruments differ in how much history exists — that is shown when you pick them. A one-year run on an instrument with twenty-nine days of data is impossible, and it is better to learn that before starting.
Risk next, not indicators
Risk limits matter more than the set of factors, because they decide how much a bad streak costs you. Maximum concurrent positions, the daily cap on signals, the minimum reward-to-risk ratio — that is your frame. Indicators decide which signals arrive; risk decides whether you survive a bad month.
And only at the end — the factors
Change the threshold and the minimum number of cores one at a time and watch what happens to the number of signals. The classic mistake is turning everything on at once: you get either far too many signals or none, and you cannot tell which setting did it.
Three signs of a strategy that will not survive the market
- A perfect curve on history. Almost always overfitting — see the
article on it.
- Many settings chosen for the result of a single run.
- Testing on one period only. One stretch is one set of circumstances.