Evaluate mechanical systems as a three-way fit test

Start from the operator, not a formula

Editorial reading: treat system evaluation as a three-way fit test. Keep the mechanical trading system one procedure after costs and missed-signal constraints. Make the drawdown limit the binding gate rather than a leftover statistic. Score the trading psychology process by whether the operator can still take the next signal after the worst plausible losing streak.

A mechanical trading system is supposed to match one operator's profit objective and risk tolerance. Evaluation starts from those limits rather than from a single best formula.

Keep the mechanical rules one procedure

Historical profitability of a mechanical trading system is incomplete until commissions and slippage are charged. In one worked case, 100 trades and a 5,000-unit gross result become a 2,500-unit net loss at 75 units of cost per trade.

Agreement between a tested mechanical procedure and later results is described as rising with the number of trades in the sample and falling when the sample is small. High-activity rules pay more fixed costs, while lower-activity rules leave more room for path-to-path variation.

Scarce capital can omit a later signal while other positions are open. For mechanical trend procedures that win 35 percent of trades or fewer, missing one winner can change live results relative to the test.

Make the drawdown limit the binding gate

Ranking systems by total profit can hide equity declines some accounts will not accept. A 50-percent drawdown may sit inside one operator's limit and outside another's.

A drawdown-limit comparison must match the sizing method. Read peak-to-valley drawdown in cash for single-contract tests, and as percentage peak-to-valley equity change when a fixed fraction such as 2 percent of the account is risked per entry under equalized-risk sizing.

The profit-margin ratio can rank the same dollar result higher in a lower-margin market and therefore mislead, even though posted margin still blocks that capital from other signals.

Identical profit-drawdown ratios do not imply identical fitness. Procedures that share the ratio can still open with a 5,000-unit drawdown against a 10,000-unit account, a 50-percent decline treated as a common close-out threshold for pooled accounts.

When historical risk is 10 percent of equity on two contracts at 500 units of risk each, the smallest whole-contract scale is one contract, so the drawdown limit can be cut only in half unless a smaller contract exists.

Score the process by the next signal

A trading psychology process must tolerate probable losing streaks, not the shortest streak in a test. A 50-percent win rate that printed four consecutive losers can still produce about ten losers in a row in a sample of roughly 1,000 trades.

Operators abandon mechanical rules that do not match those limits. Editorial reading: the process fails the fit test when the operator cannot take the next signal after that plausible streak.