Why your strategy trades differently live than in the Strategy Analyzer (and how to know if it can be fixed)

The scene is always the same: you enable the strategy on a live account (or live demo), you wait for the entry the Strategy Analyzer would have taken… and it takes a different one. Or none. Or a late one. Or one in the wrong place. And you start wondering whether the backtest was a lie.

Before you throw away the strategy — or the computer — here's the thing: in the vast majority of cases the backtest isn't "lying"; there's a specific, findable technical problem. These are the five causes behind almost every case I've seen, including my own strategies.

Cause 1 — Forming bars vs closed bars

The Strategy Analyzer processes closed bars: when your code reads a bar, it's complete. In real time, your code receives ticks from a bar still forming: the High, Low and Close of "this bar" change with every tick.

If your logic builds levels (an opening range, a session high, an average) by reading the current bar without checking whether it's closed, live it will be using data the backtest never saw. Typical symptom: entries systematically one bar early or late, or breakout levels that don't match the analysis. The Calculate setting (OnBarClose, OnPriceChange, OnEachTick) decides when your code runs — and changing it changes the strategy.

Cause 2 — The backtest's fairy-tale fills

The analyzer fills orders at the exact crossing price, with no slippage and no queue. Live, a market order fills with slippage, and a limit order sometimes doesn't fill even when price touches it. This doesn't change trade times: it dirties prices and results. If your trades match in time and direction but the numbers don't quite square, you're here — and this isn't a bug, it's the toll of trading for real.

Cause 3 — You're not looking at the same data

Your provider's historical data and the live feed don't always tell the same story: different session template (RTH vs 24h), contract merge policy, minute-bar backtests vs tick-by-tick reality… The same logic on different data is a different strategy. Make sure the analysis uses the same instrument, the same session template and — if possible — tick replay.

Cause 4 — Reconnections and states

Live trading has things the backtest never meets: connection micro-drops, restarts, weekends. The NinjaTrader settings that manage this (ConnectionLossHandling, StartBehavior, max restart attempts) decide whether, after a drop, the strategy recalculates, waits until flat, or simply disables itself. Typical symptom: "it didn't enter today but the backtest did" right after a drop, or finding the strategy switched off without knowing since when.

Cause 5 — Order management doesn't behave the same

Managed stops and targets (SetStopLoss, SetProfitTarget) anchor to the real fill and usually behave. Problems come with manual management: breakevens computed from theoretical prices instead of the real fill, exits using the wrong call or overload, orders stepping on each other. The backtest forgives constructions that live trading punishes.

Diagnostic checklist

When is it a bug and when is it just "real life"?

Small price differences will always exist: that's slippage and latency, and no fix in the world removes them. What's not normal — and can be fixed — is different trades: different times, different direction, entries the analysis takes and live doesn't (or the other way round). That's structural, it has a concrete cause in the code or configuration, and it can be located.