Trading glossary

The average profit or loss you expect from one trade, in currency or in R.

Expectancy is (win rate × average win) − (loss rate × average loss). It answers "what is one more trade worth?", which is the question position sizing and cost analysis both depend on. Quoted in R-multiples it becomes comparable across instruments — but like every average, it hides the distribution, so an expectancy carried entirely by two outlier trades is not the same as one earned evenly.

The largest peak-to-trough fall in equity over the test period.

Measured from the highest equity reached to the lowest point before a new high, usually as a percentage. It is the single most useful risk number because it approximates the worst stretch you would have had to sit through. Note that it is a sample minimum: the longer the backtest, the larger the maximum drawdown will tend to be, so comparing the figure across periods of different length is meaningless, and your future worst drawdown is very likely to exceed the historical one.

A strategy that bets an extended price moves back toward a recent average.

Mean-reversion systems fade moves rather than follow them: they buy weakness and sell strength, expecting price to return to a moving average, VWAP or session midpoint. The characteristic profile is a high win rate with occasional large losses, which flatters most summary statistics and makes the tail risk easy to miss. It is the structural opposite of breakout trading, and the two tend to work in opposite market regimes.

Entering when price breaks the high or low of a defined period after the open.

The opening range is the high and low of the first N minutes of a session — 5, 15 and 30 minutes are common. A breakout strategy buys above that high or sells below that low, on the theory that the range defines where the session's initial balance was found. Results are extremely sensitive to the range length, the session definition and the fill assumption, which is why two people can test "the same" ORB and report opposite conclusions.

Gross profit divided by gross loss.

A profit factor of 1.0 breaks even; 1.30 means you took in $1.30 for every $1.00 given back. It is a ratio of totals, so it says nothing about how many trades produced them — the same figure over 40 trades and over 4,000 trades carry completely different amounts of information. For an intraday futures strategy net of realistic costs, a value far above 1.5 is more often a modelling bug than an edge.

A trade’s result expressed in units of the risk taken on it.

If you risked 4 points and made 8, that trade was +2R; if you were stopped out as planned, it was −1R. Expressing results in R makes trades comparable across instruments and position sizes, and turns a P&L curve into a statement about the strategy rather than about the account. It relies on the initial risk being defined before entry, so it does not apply cleanly to strategies without a hard stop.

Return above the risk-free rate divided by the volatility of those returns.

Sharpe measures return per unit of variability, usually annualised. It is the standard cross-strategy comparison because it is scale-free, but it has two well-known weaknesses: it penalises upside volatility exactly as much as downside, and it assumes returns are roughly normal, which strategies with a high win rate and rare large losses are not. Always check the number of periods it was computed from — an annualised Sharpe from a few months of data is noise with a decimal point.

The difference between the price you expected and the price you got.

Slippage is not random noise that averages out — it is systematically against you, because it is largest exactly when the market is moving fastest and your stop is being hit. A backtest that fills stops at the exact stop price removes the worst part of every losing trade, which is one of the most common reasons results collapse in live trading. Model it as a cost per round turn and check at what cost assumption the strategy breaks even.

Repeatedly optimising on one window of history and testing on the window immediately after it.

Rather than fitting parameters once over all history, walk-forward fits on a window, tests on the next unseen window, then rolls both windows forward and repeats. Stitching the out-of-sample segments together gives an equity curve made only of data the parameters had never seen, which is a far more honest estimate than a single in-sample fit. It does not eliminate overfitting — reusing the same walk-forward run to choose between many strategies reintroduces it.

The percentage of trades that closed profitably.

Win rate on its own says nothing about profitability: a 70% win rate loses money when losses are large enough, and a 35% win rate is excellent if the wins are big enough. It is only meaningful alongside the payoff ratio (average win over average loss), since profit factor = payoff ratio × win rate / (1 − win rate). Treat any strategy marketed on win rate alone as marketed on the least informative statistic available.