Backtest vs live trading: why the results differ
Every strategy has two track records: the one in the backtest and the one on the account. They differ for four distinct reasons, and a trader who cannot name which one is at work cannot fix it. This is the map.
Why the gap exists
A backtest is scored on a single ordering of historical trades, with fills the script assumed and signals the script could see in hindsight. Live trading has one ordering that has not happened yet, fills the market decides, and signals that exist only when a bar closes. The gap is not a bug; it is four separate effects, and each has a page below.
Execution: spread, slippage and commission the backtest assumed away. Repainting and lookahead: signals scored on information that was not available at the time. Sampling: one path through history treated as the whole distribution. Overfitting: parameters tuned to the past until the edge is the noise.
The pages, in reading order
- Why a Pine backtest won't match live
- Realistic commissions, slippage and fills in Pine
- How to verify a backtest is real
- Is your backtest overfit?
- How long should you backtest?
- Is twelve months enough?
- In-sample vs out-of-sample testing
- Walk-forward vs Monte Carlo
- Choosing parameters without overfitting
- Pine mistakes that only surface live
- Repainting and barstate.isconfirmed
- Our forecast vs live results: the scorecard
- Verify an automated trade really happened
- What 12,000 backtests reveal about payouts
The one test that catches most of it
Run the strategy live on a demo or a small account for a month and compare the trade list to the backtest for the same month, trade by trade. Fills, timing and count will differ; the question is by how much and in which direction. If live results sit inside the range a Monte Carlo resample of the backtest produces, the strategy is behaving. If they sit outside it, one of the four effects is larger than modelled, and the trade list tells you which.
Related
- Balanced vs growth: the price of safety
- Best TradingView indicators for prop-firm trading
- Prop firm rules that change your strategy
FAQ
Why does my strategy perform worse live than in the backtest?
Four candidates: execution costs the backtest ignored, repainting or lookahead in the signals, a single historical path treated as the whole distribution, or parameters overfit to the past. A trade-by-trade comparison of one live month against the same backtest month shows which.
How much worse should I expect live results to be?
Enough to matter but not enough to change the sign. A strategy with a profit factor of 1.6 in a clean backtest commonly runs 1.3 to 1.4 live once realistic execution is modelled. A backtest that survives only at 1.1 will not survive live.
Is a 12-month backtest enough?
It depends on trade count, not calendar time. A strategy with 300 trades in twelve months has a usable sample; one with 30 does not. Monte Carlo resampling turns the single path into a distribution regardless of window length.
Firm rules verified September 2026. Prop firm rules change frequently; confirm on the firm's site before purchasing an evaluation.