First, it must not depend on a single currency pair, a single market regime or a single timeframe. A real edge should be able to work across multiple instruments and multiple timeframes, because market behaviour changes constantly. If a strategy works only on one pair or one specific setup, then there is always the risk that it is simply overfitted to historical noise.
UNIVERSAL APPLICATION
- Works on ALL currencies
- Works on ALL timeframes
- Does not depend on a specific market or condition
Second, the backtest itself should include the maximum amount of reliable historical data available. No cherry-picked periods. No bull years only. No favourable volatility environments only. A strategy should survive across different monetary cycles, crises, ranging markets, trend expansions, liquidity shocks and black swan events. The more historical environments an algorithm survives, the more confidence you can have that it is exploiting something structural and not temporary.
FULL DATA HISTORY
- Backtest on ALL available historical data
- Includes all market conditions
- Survives crises, high volatility, bearish and trending markets
Third, the backtest and the behaviour of real (Live) trades should correlate strongly across a meaningful sample of trades. This is one of the biggest differences between a theoretical system and a real trading framework. Many systems produce beautiful backtests but collapse entirely once exposed to live execution, spread variation, slippage, latency and changing volatility conditions. If live trades start to resemble the statistical behaviour of the historical model over time, then the probability that the edge is genuine increases significantly.
BACKTEST – LIVE CONVERGENCE
- The backtest must realistically simulate market conditions
- Live trade behaviour must align with the backtest
- A sufficient number of live trades for statistical reliability