A metric compresses a backtest. It does not replace the evidence.

Five details change what the metric means.

Metric names are not universal contracts. When a platform publishes a source-specific formula, that definition governs the value shown there; the common forms below are interpretation guides, not substitutes for that software's documentation.

• Sample: full history, in-sample, and out-of-sample are different evidence sets even when the formula is identical.
• Unit: money, percent, pips, or another result basis changes the scale of component values.
• Costs: commissions, spread, slippage, and other costs matter when they are included in the underlying trade P/L.
• Trade count: a statistic from a small sample can be dominated by a few observations.
• Selection: a metric chosen after screening many candidates inherits the selection process.

Net Profit

Net Profit is an aggregate outcome, not a rate of return. Its meaning depends on the P/L accounting behind each trade: costs, currency conversion, position sizing, open-position treatment, and other adjustments can change what is included.

What can mislead you?

• A few large trades can dominate the total.
• Two samples with the same Net Profit can have very different drawdowns, durations, and capital requirements.
• Do not compare totals across different capital, sizing, currency, or cost conventions without normalizing the context.

Profit Factor

Profit Factor compares the total magnitude of winning trade P/L with the total magnitude of losing trade P/L in the selected sample. A value above 1 means gross profit exceeded absolute gross loss in that sample; it is not, by itself, a validation verdict.

What can mislead you?

• A very small loss denominator can make the ratio unstable.
• The ratio does not describe trade order, drawdown depth, recovery time, or concentration.
• Comparisons are only meaningful when trade accounting, costs, sample scope, and units are comparable.

Expectancy

Expectancy summarizes the average outcome per trade under the trade definitions and P/L units used by the result. With consistent accounting, this is the same average-outcome idea as mean trade P/L; zero-result trades contribute zero to the probability-weighted form.

What can mislead you?

• A few unusually large trades can dominate the average.
• Positive expectancy does not describe the sequence of losses or capital path.
• Changing costs or trade classification changes the underlying inputs.

Return / Drawdown Ratio

This form compares cumulative net profit with the deepest observed decline from a prior peak. If both inputs use the same result basis, the ratio is dimensionless. Some platforms use different return conventions, so the exact definition should be checked before comparing ratios across systems.

What can mislead you?

• A near-zero drawdown denominator can make the ratio unstable or undefined.
• Maximum drawdown is sample- and path-dependent; the ratio inherits that dependence.
• A large ratio can result from an unusually favorable sample and does not establish future stability.

Maximum Drawdown

Maximum drawdown is the deepest observed decline from a prior running peak within the selected sample. It is useful context for Return/DD, but the maximum alone hides duration, recovery, clustering, and the possibility of larger declines outside the observed history.

Full, in-sample, and out-of-sample values are not interchangeable.

Applying Profit Factor, Expectancy, or Return/DD to a held-out sample answers a different evidence question from applying the same formula to data used for development or selection. Keep the sample label attached to the value whenever you compare or report it.

Use metrics together, not as a ranking recipe.

These statistics summarize different properties of the same observed result. They do not prescribe a threshold, prove a strategy is robust, or replace checks for selection effects, sample dependence, uncertainty, and data quality.