Quantprove gives a strategy full statistical weight at 500 trades, where the confidence multiplier sqrt(n)/sqrt(500) reaches 1.0. Below 500, every score gets discounted. In Backtest, a sample near 20 trades hits the 0.2 floor, so a raw 80 reports about 16. Monitor mode needs at least 100 trades to run at all.
A trading strategy reaches full statistical weight in Quantprove at 500 trades. That is the point where the confidence multiplier sqrt(n)/sqrt(500) equals 1.0 and a score is reported without any sample-size discount. Below 500 trades, the multiplier shrinks the score because the evidence is thinner. No single universal minimum fits every strategy, and the 500-trade anchor is where Quantprove stops discounting for sample size.
A short backtest tends to flatter a strategy. With a small number of trades, an ordinary result can look exceptional because a handful of lucky outcomes pull the average up, and there are not enough trades for the unlucky ones to balance them out. The fewer the trades, the easier it is for a profitable curve to be luck rather than a repeatable pattern.
Bailey and López de Prado document the formal version of this problem in The Deflated Sharpe Ratio (2014): performance estimates drawn from small samples systematically overstate true edge, and the shorter the track record, the larger the overstatement. Quantprove builds that finding directly into how every score is reported.
Statistical significance is the concept underneath all of this. A result is statistically significant when it is unlikely to have happened by chance. A strong-looking strategy with 30 trades has not cleared that bar, because the result could be real or it could be luck from 30 trades.
In simpler words... a hot month on 30 trades proves the month, not the edge.
The confidence multiplier is the number Quantprove uses to scale a raw score down when the sample is thin. The formula is sqrt(n)/sqrt(500), where n is the trade count. The final score equals the raw score times this multiplier.
The multiplier grows as the trade count grows, but it grows on a square-root curve, not a straight line. The early trades buy a lot of confidence and later trades buy less. Going from 50 to 200 trades moves the multiplier more than going from 350 to 500 does. The curve reflects diminishing returns: each additional trade adds proportionally less new evidence than the one before it.
Publishing this formula is a deliberate transparency choice. The trader can see exactly how sample size changes the headline number rather than having a score handed down without explanation.
Final score = raw score x confidence multiplier. The multiplier is sqrt(n)/sqrt(500), which reaches 1.0 at 500 trades.
Two verified anchors frame the reading in Backtest mode. At 500 trades the multiplier is 1.0 and the raw score is reported in full. At roughly 20 trades the multiplier hits its 0.2 floor, the lowest it goes in Backtest. A trade log near that floor with a raw Edge Score of 80 reports about 16 after the discount.
The gap between those two numbers shows the effect of sample size. A reported score that looks weak on a 20-trade sample shows there is not enough evidence yet to give the idea full weight. The same raw quality on 500 trades would report far higher.
The practical reading: treat a low score on a thin sample as a request for more trades. Collect more trades, re-run, and watch whether the reported score climbs as the multiplier rises toward 1.0.
Quantprove has three modes, and each handles trade count in its own way.
Backtest produces the Edge Score and applies the multiplier with a 0.2 floor. The floor means a tiny sample is discounted heavily but never to zero, so a 20-trade log still reports a fifth of its raw score and the trader sees something rather than nothing.
Validation produces the Stability Score, which compares the backtest against the live record. It uses the same sqrt(n_live)/sqrt(500) multiplier keyed to the live trade count, with no floor. A validation built on very few live trades can therefore be discounted all the way down, because there is no live evidence yet to defend.
Monitor produces the Health Score and works differently again. It needs a minimum of 100 trades to run, then recomputes a rolling Edge Score across the live record in adaptive windows. Below 200 total trades it uses a window of 20 trades stepping 8 at a time; at 200 or more it widens to a window of 50 stepping 15. Monitor reports what the rolling score is doing across the record and leaves the decision to the trader.
Aim for 500 trades if the goal is a score reported at full weight with no sample-size discount. That is the cleanest reading Quantprove offers. Below 500, scores stay useful, and they are honest about how much evidence stands behind them.
If a strategy trades for a year and produces 120 trades, the confidence multiplier sits below 1.0 and the reported score carries a visible discount. That is the correct outcome: 120 trades is real evidence, and it is not yet the 500 trades that earns full weight. Reading a thin-sample score means reading the discount as part of the answer.
For live monitoring, 100 trades is the entry requirement and 200 trades is where the rolling analysis widens its windows for a steadier read. The getting-started guide covers the upload, and common mistakes that distort a backtest covers what else skews a thin-sample read.