Enter the results of your backtest or of your live trades. The calculator tells you how likely a strategy with no edge would be to look this good, how many trades you need, and the drawdown to expect.
The guide shows how to test an edge properly: out of sample, with costs and against random entries, with Python scripts. Validating a Strategy Without Fooling Yourself
A 46-page guide with Python scripts: out-of-sample tests, costs, random-entry benchmarks, multiple testing and a diary of strategies that failed. English PDF.
Validating a Strategy Without Fooling YourselfLa stessa guida in italiano, in PDF oppure in versione cartacea su Amazon.
PDF su Payhip · Cartaceo su AmazonEach trade is counted as a win of the average winner or a loss of the average loser, measured in R (one unit of risk). The expectancy is the average R per trade. Its standard error shrinks with the square root of the number of trades. The ratio between the two, the t-statistic, says how many standard errors the edge is away from zero.
The first number is the chance that a strategy with no edge at all would produce an expectancy this high or higher, using the normal approximation. Below 5% is the usual threshold. Below 1% is better.
If you try 20 versions of an idea and keep the best one, one of them will look good by chance alone. The page corrects for this with the Šidák formula: the chance that at least one of your variants looks this good by luck. This is the most common way backtests fool their authors.
The table replays your win rate and win/loss ratio 5,000 times in random order, with your risk on the current balance. It shows the drawdowns and losing streaks a strategy like yours produces, so they do not surprise you. If your real results are far from this, the backtest numbers are probably not the ones you are trading.