Every result on this site comes from running freqtrade's own backtesting engine. We do not edit strategies, tune them for this test or pick the best time window after the fact. The point is a fair, like-for-like comparison.
Software
- Freqtrade 2026.10, the official release, run from Python.
- Strategies are taken unmodified from the community freqtrade-strategies repository, using whatever parameters their authors committed.
Data
- Exchange: Binance spot candles (OHLCV).
- Pairs: BTC, ETH, SOL, BNB, XRP, ADA, DOGE, AVAX, LINK and DOT, all against USDT. These are large, liquid coins that every supported exchange lists.
- Period: 1 June 2023 to 30 May 2025, about two years. This window includes the 2023–2024 rally, the spring 2025 sell-off and several sideways stretches.
- Timeframe: each strategy runs on its own native timeframe (for example 5m, 1h or 4h).
Account settings
| Setting | Value |
|---|---|
| Starting balance | 1,000 USDT |
| Stake per trade | 200 USDT (fixed) |
| Max open trades | 5 |
| Fee | 0.1% per side (Binance's standard spot fee), charged on entry and exit |
| Trading mode | Spot, long only, no leverage |
| Protections | None, unless built into the strategy |
Fixed stakes keep results comparable: a strategy's return is not inflated by compounding a lucky early streak. If losses take the balance below 200 USDT, the strategy can no longer open a trade and stops. That is why several of the worst results end at roughly −80%: the true loss would have kept growing.
Benchmarks
Each strategy is shown next to buy and hold: $1,000 put into Bitcoin on the first day and left alone. We also report an equal split across all ten coins. A strategy that trails these benchmarks did more work, and took more risk, for less money.
Metrics we show
- Return: final balance versus starting balance, after fees.
- Max drawdown: the largest fall in account value from a peak, as a percentage of that peak.
- Win rate: share of trades closed with a profit.
- Profit factor: gross profit divided by gross loss.
- CAGR, Sharpe and Sortino: as calculated by freqtrade.
What a backtest cannot tell you
Freqtrade's backtesting makes a few simplifying assumptions, documented in the official docs:
- Orders fill at the requested price with no slippage whenever the price is inside the candle's range.
- Entries are placed at the open of the candle after the signal.
- When a candle touches both the stoploss and the take-profit, the stoploss is assumed to happen first.
- Exchange trading limits are applied, but historic limits are not known.
Strategies on short timeframes with small profit targets are affected most. Treat their results as an upper bound.
Most importantly, past performance does not predict future returns. A two-year window is one market regime among many. Use these results to shortlist and understand strategies, then test them yourself in dry-run.
Strategies not yet included
Four of the strategies we selected trade on 1-minute candles (BinHV45, CCIStrategy, SmoothScalp and ReinforcedSmoothScalp). Two years of 1-minute data for ten pairs needs more memory than our test machine has, so they will be added once we can run them under the same conditions. FixedRiskRewardLoss demonstrates a custom stoploss technique and does not define its own timeframe, so it is not part of this comparison.
Updates
We plan to re-run every backtest periodically with newer data and add new strategies. Each strategy page shows the exact period it was tested on and the command to reproduce it.