How this edition was verified
Two stages, stated separately because they measure different things. The engine that computes everything is the same C# code as the cTrader edition's engine, apart from its namespace line. That engine was replayed on 20,561 EURUSD M5 bars read from the TradingView chart model and compared with a probe copy of the original script: 46 internal series equal on every bar, both trendline collections equal across 1,439,270 field comparisons, and every break, signal, drawn line and label on the same bars. The report is on the cTrader page.
The NinjaTrader host is the second stage. With its diagnostic dump switched on it writes every committed bar with the values it drew, in exactly the columns the reference harness writes; those bars are replayed through the engine outside NinjaTrader, and the two files are diffed value by value. On 5,058 MNQ 1-minute bars all 119 series compared were equal on every bar, the trendline books included, slot by slot.
What that proves: the host feeds the engine exactly one closed bar per chart bar with the right open, high, low, close and volume, never double-feeds or skips a bar, and draws the lines and levels the engine produced. What it does not claim: there is no comparison against the original on futures bars, because TradingView's CME feed and NinjaTrader's feed do not carry the same bars, so no such comparison exists to make. And nothing here says anything about future performance.
NinjaTrader bars compared5,058
Series compared119, both trendline books included
Cells compared601,902
Mismatches0
Engine vs the original20,561 EURUSD M5 bars, 46 series and both line books
ConditionsMNQ 1-minute, NinjaTrader 8.1.8.2, September 2026
The full method, with the columns compared, is in the user guide: User Guide (PDF) and Installation and Quick Start (PDF).