Strategy Review
The most common analytical error in retail trading is drawing a conclusion from twenty trades. Twenty trades of a 50 per cent system can produce anything from four wins to sixteen. The result you observe is dominated by noise, and acting on it is how traders end up changing systems monthly.
Rough thresholds
| Trades | What you can reasonably say |
|---|---|
| Under 30 | Nothing about profitability. You can check execution and rule adherence. |
| 30 to 100 | Whether the mechanics work: fills, costs, stop placement, whether the setup occurs as often as expected. |
| 100 to 300 | A weak signal on expectancy. Large edges become visible; small ones do not. |
| 300 plus | A usable estimate of expectancy, assuming the market regime did not change midway. |
That last caveat is the uncomfortable one. Collecting 300 trades on a swing strategy taking two trades a week takes three years, during which the market is not obliged to stay the same. This is a real limitation of discretionary trading, not a problem you can solve with better statistics.
Measure in R, not in currency
Expressing every result as a multiple of the risk taken removes account size, position size and instrument from the comparison. A plus two R trade on a micro lot and a plus two R trade on a full lot are the same evidence about the edge. Currency results mix the quality of the idea with the size of the bet, and then you cannot separate them again.
What to do while the sample is small
- Trade the smallest size that keeps you honest. The purpose of the first 50 trades is data, not income.
- Judge process metrics, which stabilise far faster than profit: adherence, average slippage, whether the setup appears as often as the plan assumed.
- Do not optimise parameters on 40 trades. You will fit noise and feel clever doing it.
Beware the survivorship trap in your own record
If you quietly stop recording the trades you took outside your rules, your sample describes a strategy nobody actually traded. Include everything, tag the violations, and analyse both the clean subset and the full record. The gap between them is the real cost of your discipline problem, expressed in R.
Small samples do not produce cautious conclusions. They produce confident wrong ones.
Written by Adrian Vestberg — former interbank FX dealer, now running a two-person discretionary desk.