Profit factor is the sum of profitable trades' profit and loss divided by the magnitude of losing trades' P&L, using the same trade definition, currency and cost basis throughout. It measures how much positive trade P&L the strategy generated for each unit of realised losses. A value above one means the included trades made money overall. It does not establish that the strategy earned an attractive return on capital or will remain profitable.
Add the winners, then add the losses
For eligible completed trades with P&Ls , define
Then, when ,
The denominator is a positive loss magnitude, not a signed negative number. Profit factor is dimensionless: both sums use the same currency. With a positive denominator, a factor above one means positive total P&L, one means break-even and below one means negative total P&L.
Trading reports often call “gross profit” and “gross loss”. Here those labels refer to separate sums of winning and losing net trade P&Ls; “gross profit” does not mean that fees were excluded. TradeStation's definition, for example, explicitly accounts for commissions and specified strategy slippage. Always check the report's cost settings.
Consider four synthetic closed round trips, measured in USD:
| Trade | Net P&L (USD) |
|---|---|
| A | +10 |
| B | +20 |
| C | −5 |
| D | −15 |
Commission and execution-price effects are already included. This example has no financing charges, borrow charges, distributions or taxes. Positive P&Ls sum to USD, while losses total USD in magnitude. Therefore profit factor is , and total net P&L is USD.
Original synthetic closed-trade ledger in USD. The bars compare sums of net trade P&L: 1.50 USD of positive P&L for each 1 USD of realised loss magnitude. Bar lengths share a common scale.
The figure does not say the investor earned a 50% return. No capital base or holding period has been specified. Multiplying every trade P&L by the same positive constant preserves the ratio while changing the money gained and lost.
The ledger convention is part of the statistic
For the example, a trade is a complete round trip from no position back to no position; partial fills and exits belong to that round trip. Include trades whose final exit occurs in the declared reporting window, even if their entry preceded it. State the strategy, instruments, window, time zone, currency-conversion rule and treatment of multi-leg positions. Real trading engines can offer different fill-grouping and matching methods.
Grouping can change the answer even when total P&L is unchanged. If A and C were legs of one strategy trade, their combined P&L would be USD. The grouped ledger would be : positive P&L of USD divided by loss magnitude of USD, or , rather than . Total net P&L would still be USD. Compare reports only after aligning their grouping rules.
Use P&L after the specified transaction costs and carrying charges. Allocate costs consistently before classifying trades as winners or losers; a small pre-cost gain can become a net loss. Execution costs already reflected in fill prices must not be deducted again. The Financing, Borrow and Transaction Costs lesson develops that ledger distinction and cost stress tests; the full lesson requires membership, with a public preview.
Exactly zero net-P&L trades contribute to neither sum. They leave profit factor unchanged, although including them in a trade count changes mean P&L per trade. Open positions are excluded from this closed-trade calculation; disclose their marked P&L separately. Keeping a losing position open can make a closed-trade report look better without improving the account's economic value.
| Observed closed-trade record | Profit-factor interpretation |
|---|---|
| Losses exist, no positive P&L | Zero: and |
| Positive P&L exists, no losses | No finite ratio: and ; some reports display infinity |
| No eligible trades, or only zero-P&L trades | Undefined: |
A display of infinity means no losses were observed under that reporting convention. It is not evidence of zero future downside. Keep these states distinct rather than replacing every missing denominator with zero.
Different from payoff ratio and expectancy
Profit factor uses totals. The payoff ratio uses average winning P&L divided by average loss magnitude. If trades win, trades lose, and their respective means are and , then
In the four-trade example, the average win is USD and the average loss magnitude is USD. The payoff ratio is too, because there are two winners and two losers. Equal values here do not make the two statistics interchangeable. Profit factor combines winning frequency and payoff size; the hit rate counts winners without measuring their sizes.
Historical mean net P&L is another statistic:
This sample mean is often used to estimate trade expectancy. A forward-looking expectation requires assumptions about future outcomes, costs and sizing. Profit factor is neither a 1.5 USD expected profit per trade nor a forecast of that expectation.
Read it alongside the path and the search history
Profit factor ignores the order of the trades. Reordering the same P&Ls preserves the ratio but can change drawdown, funding needs and the time spent recovering losses. It also hides concentration: excluding the largest winner from the example would leave positive P&L of USD against losses of USD, giving a factor of and net P&L of USD. That sensitivity check does not justify deleting the trade; it shows how much the result depends on it.
Report the trade counts, money totals, open exposure, cost assumptions and drawdown alongside the ratio. Examine performance across periods and instruments, and distinguish results used to select the strategy from genuinely out-of-sample evidence. Choosing the best profit factor from many backtests can select historical noise; Bailey and co-authors' research on backtest overfitting explains why selection history matters across performance measures.
The Backtest Evaluation lesson extends this report to capital-adjusted returns, drawdowns, concentration and robustness. Full access requires membership, with a public preview.
Sources
- TradeStation, Profit Factor and Gross Profit: the ratio of winning and losing totals and the report's commission/slippage treatment.
- QuantConnect, Trade Statistics: configurable trade grouping, matching and closed-trade records.
- David H. Bailey, Jonathan M. Borwein, Marcos López de Prado and Qiji Jim Zhu, The Probability of Backtest Overfitting, revised February 2015, pp. 3–4: selection across many historical trials can produce misleading performance.
- Ernest P. Chan, Quantitative Trading: How to Build Your Own Algorithmic Trading Business, Wiley, first edition, copyright 2009, Chapter 2, pp. 22–23: transaction costs and the distinction between pre-cost and implementable strategy performance. The ledgers and figure here are original.