Backtests Can Flatter Trailing-Stop Results

A trailing-stop backtest built only from candle highs and lows can report exits that were impossible in live quote order. Intrabar sequence, bid/ask spreads, update frequency and platform latency determine whether the trail moved before price reversed.

MetaTrader states that its trailing stop reacts to incoming quotes and limits how often one position can be trailed. That is a sequence-dependent process. A single hourly candle showing both a high and a low does not reveal which came first. A backtester that assumes the favorable high occurred first may tighten the stop and then book an exit that real trading never could have produced.

Minimum evidence for a credible test

Test input	Why it matters
Tick or high-resolution data	Preserves path order
Bid and ask	Captures executable spread
Platform update rule	Matches trailing frequency
Slippage model	Avoids fill-at-trigger fiction
Out-of-sample period	Tests parameter stability

News periods deserve a separate test set. The August U.S. CPI release showed a 0.4% monthly rise, and the September Fed meeting includes projections. The BOJ decision follows two days later. These are exactly the conditions in which a bar-based assumption can be least reliable because price and spreads can change between recorded intervals.

Compare three results: a coarse candle test, a tick-level bid/ask test and a pessimistic execution test. Report the difference rather than publishing only the best curve. Include disconnected-terminal periods if the live implementation relies on a local trailing algorithm. A VPS reduces one operational risk but does not erase network or broker-side latency.

This 2026 MT5 trailing-stop guide shows the live feature. A demonstration of the menu is not a validation of historical performance.

The loop closes by matching the model to the mechanism. If the live trail needs quote sequence and terminal updates, the backtest must model both. Otherwise, the result measures an idealized rule that the account does not trade.