Trading research & validation
Read backtest numbers correctly and avoid fooling yourself with past results.
Win rate vs Avg R: which number matters more?
Win rate only means something alongside the average win and loss; average R per trade (expectancy) combines both and decides whether a system has an edge.
What is Profit Factor?
Profit Factor is gross profit divided by gross loss over the tested period; above 1 means profitable on that data, but it must be read with sample size, drawdown and costs.
What is drawdown?
Drawdown is the drop from an equity peak to the following trough; it shows how deep a losing stretch you must sit through, and it depends directly on your risk per trade.
How many trades are enough to conclude anything?
Sample size is the number of trades used to judge a system; the smaller the sample, the more the measured average R swings, and the harder it is to tell a real edge from luck.
In-sample and out-of-sample
In-sample is the data used to choose settings; out-of-sample is data set aside and used only to check whether the chosen settings still work on unseen data.
Overfitting: when a system is only good at the past
Overfitting is when settings are tuned so closely to past data that they capture random noise, so results look great historically but disappoint on new data.
Commission and slippage: how a small edge disappears
Commission and slippage are deducted from every trade; in R, cost equals the round-trip cost divided by the stop distance, so the tighter the stop, the more costs eat the edge.
Filtered and ghost trades: measuring what a filter is really worth
A ghost trade is a setup blocked by a filter but still tracked in the background as if taken, to see whether the filter removes bad trades or blocks good ones.
Data provenance: knowing where your chart's numbers come from
Data provenance means knowing how each number on a chart is produced: real exchange volume, a broker's tick volume, reference volume from another market, or an estimate.