Search for these patterns:
AFL check snippet:
// Verified: No look-ahead Buy = Cross( MACD(), Signal() ); // OK, uses current bar only
// NOT verified (look-ahead) Buy = Ref( Cross( MACD(), Signal() ), -1 ); // Signals based on NEXT bar? Wait: Ref(..., -1) is past? No: Ref(array, -1) is PREVIOUS bar. Ref(..., +1) is future. Be careful.
This is the holy grail. Verified code passes the "Future Leak Test" using StaticVarGet timestamps or the built-in equity() function to ensure signals are not using the close price of the current bar for entry.
Finally, plot the signals on a 5-year daily chart. Manually inspect 10 trades.
This is the most common killer of trading strategies. Look-ahead bias occurs when your AFL code uses future data to make a past decision. For example: amibroker afl code verified
// DANGEROUS (Unverified)
Buy = Ref(Close, -1) > EMA(Close, 20);
A verified code ensures that signals are calculated using only bars up to the current index. Professionals use Ref(..., -1) carefully and validate through ExRem and trade list reviews.
With ChatGPT and Copilot now writing AFL code, unverified scripts are flooding the trading community. AI often generates syntactically correct but logically flawed AFL—especially with complex state management (StaticVar, StaticVarGet).
Example of AI hallucinated, unverified code: Search for these patterns:
// AI writes this confidently, but it's broken.
Buy = Cross(StochK(), StochD()) AND Close > Highest(H, 50); // Uses future high!
AI doesn’t know that Highest(H, 50) includes the current bar unless shifted. Only a human with a verification checklist catches this. Going forward, the phrase “Amibroker AFL code verified” will become a premium service—requiring both human logic audits and automated test suites.
Before paying anyone, run these checks in Amibroker:
// 1. No repainting test Buy = YourCondition; Sell = YourExit; PlotShapes(Buy * shapeUpArrow, colorGreen); // Check if arrows appear on same bar as the trigger – not later.// 2. Look-ahead check Plot(Ref(C, 1), "Future Close", colorRed); // If this improves your strategy, you have look-ahead bias. AFL check snippet: // Verified: No look-ahead Buy
// 3. Out-of-sample test SetBacktestMode(backtestRegular); // Then split your data: optimize on 2010–2018, verify on 2019–2023.