Guides · Due diligence

How long before a track record means anything? Sample size, longevity and Monte Carlo

ConfirmedTrades team · 27 April 2026 · 7 min read

Three months and forty trades is an anecdote, not evidence. How many trades and months you need, what the Score does about it, and what Monte Carlo adds.

The uncomfortable arithmetic

Suppose a strategy has no edge at all, a coin flip with costs. Run it for thirty trades and there is a meaningful chance it finishes up 10% or more by luck. Run a thousand such strategies, as the EA market effectively does, and dozens of them will show beautiful short curves. The lucky ones are the forex robots that get marketed. This is survivorship bias, and the only defence against it is sample size.

The question is never "did this make money" but "how likely is this result if the strategy actually has no edge". That probability falls as the number of trades rises and as the size of the edge per trade rises. Neither a big return nor a high win rate substitutes for trades.

How many trades?

There is no magic number, but there are useful thresholds. Under thirty trades, nothing can be said; the AI review attaches a short-record flag below that. Around a hundred trades you can start to believe the sign of the expectancy if the edge is large. A few hundred trades across different market conditions is where a strategy's statistics begin to settle, and the Score's longevity component reflects that: its trade-count part is fully earned at two hundred trades.

Higher-frequency Expert Advisors reach these counts quickly and so need more calendar time to be believable, not less; see the next section. Lower-frequency strategies may take years to reach them, and a two-year swing record with eighty trades is genuinely hard to judge.

The range of win rates consistent with a record, narrowing as trades accumulate90%72.5%55%37.5%20%50% — a coin tossthe real 55% edgefrom ~100 trades the edge is visible1025501002004008001600Closed trades in the recordWin rate
The blue band is the range of win rates still consistent with the record so far. For the first hundred trades or so it still contains 50% — meaning the numbers cannot yet tell a real edge from a coin toss, however good the total looks.

How much time?

Trades are not independent draws if they all happened in one kind of market. Two hundred trades in a three-month trending market tell you how the strategy does in a trending market. You want to see it trade through a range, a trend, a volatility spike and a quiet patch. In practice that means at least six months, and a year is much better.

The Score's longevity component earns its age part fully at about six months and adds points for recent activity, so a fresh account is capped below an established one regardless of performance. That is deliberate. Longevity is also why the monthly calendar on the account page is so useful: count the months, count the reds, and look at whether the good months are clustered or spread out.

A simple test: mentally delete the single best month. If the record is no longer interesting, it was one month, not a strategy.

Monte Carlo: turning one history into a thousand

A track record is one path out of many the strategy could have taken. Monte Carlo simulation reshuffles the actual trade results thousands of times and looks at the spread of outcomes: how bad was the worst drawdown across the shuffles, how often did the account hit a ruin threshold, what did the middle of the distribution look like.

The point is that the drawdown you observed is a single draw. If the same trades had arrived in a different order, the drawdown could have been twice as deep. Monte Carlo tells you how deep it plausibly could have been, which is a far better guide to sizing than the one historical figure.

On ConfirmedTrades the Risk Lab runs Monte Carlo, chance of ruin, value at risk and Kelly sizing for accounts on the deeper plans. The free risk of ruin calculator lets you do a simpler version yourself with win rate, payoff ratio and risk per trade.

Many reshuffled equity paths from a single trade history$26k$22k$18k$13k$9kstarting balancemedian outcomeunluckiest 5% still finish near $17kThe same trades, dealt in a different order
One trade history, dealt in many different orders. A record worth trusting is one where even the unlucky orderings finish somewhere you could live with.

What Monte Carlo cannot do

Reshuffling trades assumes the future trades will look like the past ones. It does not add the losing regime that has not happened yet, and it is blind to strategies whose trades are dependent on each other, which is exactly what martingale and grid systems are. For those, a shuffle of closed results understates the tail badly, because the danger lives in the open basket, not the closed trades. Check Max DD (equity) in Glance and the AI review's flags first, and only then trust a simulation.

Using Custom Analysis to test stability

If the owner has enabled Custom Analysis, you can test stability directly. Filter to the first half of the record and then the second half; a strategy with a real edge looks similar in both. Filter out the best symbol; a genuine edge is still positive. Filter to a single weekday or hour; a strategy that only works on Monday mornings is a curiosity, not an edge.

A waiting rule

If a record is too short to judge, the answer is not to guess. Follow the account (a free account lets you follow) and come back in three months. The cost of waiting is missing a few months of a good strategy. The cost of not waiting is funding a lucky one.

Do it in one place

On ConfirmedTrades, every published account shows its verification badges, an AI strategy analysis that flags martingale, grid and missing stop-losses, the full drawdown and risk stats, and execution costs per symbol — so you can vet a strategy before you trust it, or prove your own.

All guides