With the same 80% win rate, are 20 trades as trustworthy as 1,000?
Suppose there are two trading records: one has 16 wins out of 20 trades, and the other has 800 wins out of 1,000. Both show 80% on the page, but those figures don’t provide the same level of precision.
In the 20-trade record, changing the outcome of just one trade shifts the win rate by 5 percentage points. In the 1,000-trade record, the same change affects it by only 0.1 percentage points. Small samples are more easily swayed by a few outcomes.
The methods for calculating proportion confidence intervals in the NIST/SEMATECH e-Handbook of Statistical Methods also take sample size into account. Given the same observed proportion, and assuming conditions such as independent observations and a stable win rate hold, a larger sample usually means a narrower estimation range—not that the rate will “definitely stay at 80%” in the future.
But the number of trades isn’t everything. If many positions in the same direction are opened on BTC, ETH, and SOL during the same market move, their outcomes may be highly correlated. Splitting one shared risk into dozens of orders doesn’t mean you’ve gained dozens of independent confirmations.
When I review trading records, I look at sample size, time span, market conditions covered, and whether signals are repeated. Many trades concentrated in a single one-way market may still tell us little about what will happen in a ranging market.
This is about how reliable a win-rate estimate is, not how profitable the strategy is. Profit and loss amounts and costs need to be assessed separately. A flattering percentage alone is no reason to decide how much to trust it.
$BTC $ETH $SOL
Tap my profile picture to view live trades
Suppose there are two trading records: one has 16 wins out of 20 trades, and the other has 800 wins out of 1,000. Both show 80% on the page, but those figures don’t provide the same level of precision.
In the 20-trade record, changing the outcome of just one trade shifts the win rate by 5 percentage points. In the 1,000-trade record, the same change affects it by only 0.1 percentage points. Small samples are more easily swayed by a few outcomes.
The methods for calculating proportion confidence intervals in the NIST/SEMATECH e-Handbook of Statistical Methods also take sample size into account. Given the same observed proportion, and assuming conditions such as independent observations and a stable win rate hold, a larger sample usually means a narrower estimation range—not that the rate will “definitely stay at 80%” in the future.
But the number of trades isn’t everything. If many positions in the same direction are opened on BTC, ETH, and SOL during the same market move, their outcomes may be highly correlated. Splitting one shared risk into dozens of orders doesn’t mean you’ve gained dozens of independent confirmations.
When I review trading records, I look at sample size, time span, market conditions covered, and whether signals are repeated. Many trades concentrated in a single one-way market may still tell us little about what will happen in a ranging market.
This is about how reliable a win-rate estimate is, not how profitable the strategy is. Profit and loss amounts and costs need to be assessed separately. A flattering percentage alone is no reason to decide how much to trust it.
$BTC $ETH $SOL
Tap my profile picture to view live trades