Common errors are using future information, ignoring costs and selecting only successful periods. Evaluation results must remain separate from rule-development periods.
Method, example and limits
The FXNXT lab uses closed candles to generate a signal and enters at the next candle open. A candle touching both stop and target without finer data produces an unresolved result, not a chosen win. The foundational strategy is a long-only moving-average crossover with fixed costs in R; financing data is absent, so this is not a complete simulation of every execution condition. Compare against the baseline and report sample size.
Try the connected toolSources
Methodology guide, not attributed to an unverified expert. Educational material is not personal advice. Content revision: