Custom Optimisation Criterion

Optimising for raw net profit routinely selects parameter sets that owe most of their result to one enormous trade, which is exactly the kind of setting that fails on the next few months of unseen data. The Strategy Tester supports optimising against a custom value instead - select "Custom max" as the optimisation criterion and it calls the EA's OnTester() function once at the end of every pass and ranks results by whatever number that function returns. This EA implements two such metrics, selectable with an input. The System Quality Number is the mean profit per closed trade divided by its standard deviation, scaled by the square root of the trade count - it rewards a strategy that wins consistently over one that wins big but erratically. Equity-curve linearity is the R-squared of the cumulative closed-trade equity curve against a straight line fitted to it - a smoothly rising curve scores close to 1.0, a lumpy one scores lower even with identical final profit. Neither metric is available from the tester's built-in statistics, so the EA builds its own trade-by-trade ledger during the run and both figures are computed from that ledger inside OnTester(). The underlying strategy is a plain fast/slow EMA crossover with an ATR-scaled stop and target, deliberately unremarkable - the point of this EA is the custom criterion, not the signal. Swap EvaluateSignal() for any other entry logic and the criterion still works, because it is computed from closed-order history, not from the signal itself. Notes and limitations - The System Quality Number here is computed from raw trade profit in account currency, not from R-multiples (profit divided by initial risk) as in Van Tharp's original formulation - a simplification that still rewards consistency over one outsized trade, but is not directly comparable to SQN figures computed the strict way. - Both metrics need a meaningful number of closed trades to mean anything; with fewer than two trades SQN returns 0, and linearity needs at least three. - The placeholder signal has no claimed edge; it exists only to generate a trade ledger for the criteria to be computed from. - OnTester() only runs inside the Strategy Tester - attaching this EA to a live or demo chart trades normally but the custom criterion is never evaluated outside a test pass. - The MQL5 ledger assumes each position closes in one exit deal (no partial closes); see MQL4 VERSUS MQL5 DIFFERENCES above.

Expert advisor · MT5 · v1.0 · Free

Tags: Moving Average · ATR · MT5 Expert Advisor · MQL5 Source Code

Expert advisors (EA) for MT4 & MT5 · PolyMaria