Profit factor: the ratio of money made to money lost, and how easily it is faked

The profit factor is the sum of all of a strategy's winning trades divided by the absolute sum of all its losing trades. It answers one question: how much gross profit the system generates for every rouble of gross loss. The threshold it is measured against is the point at which what is earned exactly offsets what is lost: below it, the strategy gives more back to the market than it takes; above it, the strategy stays in the black. Everything else that is usually attributed to this metric — robustness, the quality of entries, readiness for real money — it does not measure.

What exactly is divided by what

The numerator is gross profit: the sum of the results of all closed trades with a positive outcome. The denominator is gross loss: the sum of the results of the losing trades, without the sign. The key word here is "gross": there is no averaging, no netting within the numerator, and the order of trades is not taken into account. The metric is indifferent to when the losses occurred — at the start of the period or in a row at the end. That is what makes it convenient for quick comparison and useless for judging survivability.

A direct consequence follows: the profit factor does not depend on position size if the position is the same in every trade, and it becomes almost impossible to compare between strategies in which volume changes according to money-management rules. Recalculate the same series of trades with pyramiding and the ratio will shift, even though the entry logic has not changed one iota.

Why the break-even threshold is deceptive

Measuring from the break-even point looks fair, but that point is calculated from trades, not from the money in the account. Commissions, the spread and slippage are either understated in a backtest or missing altogether, and it is precisely these that eat up the edge of systems with short holding periods. A strategy that holds just above break-even on historical data almost always ends up below it in live trading: costs hit the numerator and feed the denominator at the same time.

The second source of deception is rare large wins. The metric adds up profits without asking where they came from. If the lion's share of the numerator was built up by a couple of outstanding trades on a gap, the profit factor no longer describes the strategy but luck on a specific stretch of history. The check here is simple: throw out the best trades and recalculate. If the ratio collapses towards break-even, what you have is not a system but a record of a few lucky breaks.

What the metric does not see

It does not see drawdown. Two strategies with an identical ratio of profit to loss can differ in that one loses a little and often, while the other loses rarely, but so badly that the account does not survive until recovery. For this side of the question there is a separate metric — the recovery factor, whose denominator is the worst drawdown rather than the total of all losses.

It does not see the length of streaks. Consecutive losses change the psychology and the capital requirements, but in the gross total they look the same as losses scattered across the period.

It does not see frequency. A ratio built on a handful of trades is statistically empty: a small change in the outcome of an individual trade pushes the metric across the threshold. To understand how far the result owes itself to the particular sequence and sample, the trades are run through random permutations — this is covered in detail in the piece on Monte Carlo for a strategy.

It does not see the market regime. A ratio built up during a trending stretch changes beyond recognition in a sideways market. That is why the metric is calculated separately for sub-periods and the spread of values is examined, rather than the single figure for the whole history.

How to calculate it so the figure can be trusted

Step 1: include the full costs in the calculation — broker and exchange commissions, the spread, and a reasonable estimate of slippage for your order size and the instrument's liquidity. For securities with a thin order book this is the decisive adjustment; liquidity is worth checking against the equity market.

Step 2: calculate the profit factor on a stretch of data that was not used to fit the parameters. Keep the value from the training period separately — the gap between the two is a diagnosis in itself.

Step 3: recalculate without the best trades and without the worst. A robust system loses part of its edge but does not slump to break-even.

Step 4: split the history into sub-periods and look at the spread. A consistently modest ratio is more useful than a single brilliant one earned thanks to one lucky year.

Step 5: put alongside it metrics that look at drawdown and distribution, not just the total. The profit factor works as a first-pass filter, not as a verdict.

It is also worth separating similar-sounding concepts: the profit factor is a characteristic of a trading system, whereas factor investing and factor risk concern explaining returns through risk factors, and factoring relates to something else entirely — financing deferred payments. All they have in common is a word; the definitions are collected in the glossary.