Test EA parameters and timeframes fairly

Write the test plan before seeing winners

Save the original preset, specify the inputs to vary and define the range and scoring criteria. Change one parameter at a time initially, then examine selected interactions when warranted. Record every completed, failed and missing run. A search that retains only profitable settings conceals how much selection took place.

Inspect the neighborhood and sample

Look for a broad usable region across profit, drawdown, trade count and different periods. Check whether a result is dominated by a few trades or one year. A smooth neighborhood is encouraging evidence under tested conditions, but correlated inputs or repeated selection can still overfit. Do not call a period independent out-of-sample if it influenced development or candidate choice.

Distinguish chart and signal timeframe

Read the documented signal timeframe: an EA can use a fixed internal timeframe even when placed on another chart. A different chart label may therefore test nothing new. If signals really change, keep that as a separate setup and compare matched history and risk. OHLC screening is useful for a declared preliminary study, but tick-sensitive finalists require detailed execution verification.

Checks to make before deciding

• Declare development, selection and evaluation periods before examining the holdout. Store a dated manifest showing when the binary and preset were frozen.
• Keep the latest full year separate, but call it independent out-of-sample only when it demonstrably did not influence development or candidate selection.
• For walk-forward tests, specify training window, test window, step size and refit rule. Apply only information available at each refit and record all selected parameters.
• Distinguish historical forward, demo forward and real-money forward. Identify broker, account type, server, deposit changes, start date and evidence source without ambiguous live labels.
• Select a small justified neighborhood around the frozen preset. Name the parameters and ranges before viewing results; separate robustness checks from a new optimization campaign.
• Keep all other settings and data fixed. Change one dimension at a time initially, then inspect interactions where the strategy has a clear reason for coupled parameters.
• Show net return, equity drawdown and trade count across the grid. A smooth profit surface can still conceal unacceptable risk or too few observations.
• Include boundary values and small plausible perturbations. Inspect whether slight changes reverse the result, eliminate trades or trigger unexpected order behavior.