Freqtrade Paper Trading
PAPER trading is most useful as a forward test of the complete system, not as a demo account used to celebrate a few winning trades.
What PAPER can test
It can expose signal frequency, position overlap, exit behavior, runtime failures and whether a strategy behaves differently as market conditions change. It also creates a time-ordered record that cannot be optimized retroactively without starting a new version.
What PAPER cannot prove
Simulated fills can differ from live fills. Slippage, latency, liquidity, fees and exchange behavior can reduce live performance. A small profitable sample is therefore evidence about that sample, not proof of a durable edge.
Build a minimum forward record
When should PAPER testing stop?
Not after an arbitrary number of profitable days. The useful threshold depends on trade frequency, regime coverage and the decision being made. A five-trade sample can reveal implementation bugs, but it cannot establish stable expectancy. More observations reduce uncertainty; they do not guarantee future profitability.
Metrics to retain
- Closed and open trades.
- Net PnL after modeled costs.
- Drawdown and losing sequences.
- Win/loss distribution rather than win rate alone.
- Strategy version and configuration changes.
- Downtime, errors and missed data.
See our implementation
Our experiment pages publish sanitized aggregate PAPER metrics automatically while the trading server remains isolated.
Open experiments →Scope: PAPER trading in Freqtrade
This page focuses on running and evaluating Freqtrade in PAPER/dry-run mode. For the broader methodology of forward simulation and its evidentiary limits, use our platform-neutral PAPER trading guide.