EdgeWatchFXby ErgodicLabs

The RecordEvidence

How many trades before a track record means something

A profitable record of 100 trades can come from a strategy with no edge at all. How many trades it takes depends on how large the edge is against the noise around it, and the answer is usually more than people expect.

2 min read

A strategy has 100 closed trades and is up. Is that skill, or was it a good run? The honest answer is usually: too early to tell. It helps to see why, because the same arithmetic tells you roughly when it stops being too early.

A record with no edge

Imagine 100 strategies that have no edge at all: each trade wins or loses the same amount with a 50% chance. After 100 trades, how many would show a win rate of 60% or more?

About 3 in 100. And about 18 in 100 would show 55% or more. If you look at many strategies and keep the best ones, a few will always look good after 100 trades, and some of them will have nothing behind them.

A short positive record is good news that is still early. It is not proof, and it is not nothing either.

Edge against noise

What decides how fast a record becomes convincing is the average result per trade against the spread of results around it. Call the ratio of the two the edge-to-noise ratio. A strategy that makes 0.1R per trade on average, with results that scatter by about 1R, has a ratio of 0.1.

A rough rule from basic statistics: to see an average two standard errors above zero, you need about (2 / ratio) squared trades.

Edge-to-noise ratio Trades needed, roughly
0.20 100
0.10 400
0.05 1,600

Most real strategies sit at the lower end. A ratio of 0.1 is already a good strategy. So 400 trades is not a long record for it; it is the beginning of a convincing one.

Time, not only trades

The same idea works on daily or weekly returns, with the Sharpe ratio as the edge-to-noise measure. Bailey and Lopez de Prado gave a formula for the minimum track record length: how long a record with a given Sharpe ratio, skewness and kurtosis must be before its returns can be told apart from zero.

Measured weekly, with normally distributed returns, a strategy with an annual Sharpe ratio of 1 needed about 143 weeks, close to three years. One with a Sharpe ratio of 2 needed about 38 weeks. Fat tails and negative skew, which most trading returns have, make the record longer still.

Looking every day makes it worse

There is a second trap. If you check a record every day and stop the first time it looks significant, you will be fooled far more often than the usual 5%. Each look is another chance for luck to cross the line.

The fix is a bound that stays valid however often it is checked: a confidence sequence. It is more conservative than a single test, and it is what EdgeWatchFX uses before calling a record Established.

How EdgeWatchFX words it

The Evidence readout starts from a sceptical assumption, that the edge is zero, and lets the trades move it. It reports the chance that the edge was above zero, its likely size with an 80% range, and roughly how many more trades a firm claim would need. A record under 30 trades is called very early whatever it shows. A positive record is called early or developing until the evidence is strong. It is lenient with good records and honest about how much they show.

Figures in this note describe recorded results. None of it is a forecast or investment advice.

More from The Record

When deposits flatter a return

A record can show a strong percentage return and still have lost money. It happens when the good results came on a small balance and the bad ones…

Evidence2 min read