How We Backtest Crypto Trading Signals
The Problem With Most Backtests
Every crypto signal service claims they're "backtested." Most of them are lying — or at least misleading you. Here's what they don't tell you:
- Survivorship bias. They show you the strategy that worked. They don't show you the 99 variations they tested before finding one that fit the historical data.
- No transaction costs. That 200% return shrinks fast when you add slippage, funding rates, and exchange fees.
- No liquidation modeling. A 10x leveraged strategy that "returned 500%" probably got liquidated three times along the way. They just don't model it.
- Cherry-picked date ranges. Start your backtest the day before a bull run, end it at the top. Instant alpha.
Atreidis was built to solve this. Our backtesting engine models the things most tools conveniently ignore.
What Makes Atreidis Different
Real Funding Rates
If you trade perpetual futures on Hyperliquid, you pay (or receive) funding every hour. Over a multi-week position, this adds up. Our engine models Hyperliquid's actual funding rate structure — hourly settlement at 0.00375% per interval. For comparison, we also support Binance-style 8-hour funding.
Most backtesting tools treat perps like spot with leverage. That's wrong. Funding rates can eat 30%+ of your returns annually.
Liquidation Risk
We model maintenance margins and liquidation penalties. If your position's unrealized loss breaches the maintenance margin threshold, you get liquidated — just like on the real exchange. No hand-waving, no "we assume you add margin."
Hyperliquid uses a 0.3% maintenance margin and 0.3% liquidation penalty. Our engine applies these exactly. You'll see liquidation events in your backtest results — because they happen in real trading too.
Directional Control
Most signal tools give you a binary: buy or sell. Atreidis lets you configure:
- Which signal colors open long positions
- Which signal colors close long positions
- Which signal colors open short positions
- Which signal colors close short positions
- Independent leverage for each direction
This means you can model conservative strategies (long-only on green, close on yellow) or aggressive ones (short on red, close on green, with higher leverage). Same signal, different risk profiles.
Decimal Precision
Financial calculations in floating-point JavaScript are a nightmare. 0.1 + 0.2 = 0.30000000000000004. In a leveraged position, these rounding errors compound into material discrepancies.
We use decimal.js for all arithmetic — the same arbitrary-precision library used by financial applications. Every P&L calculation, every margin check, every funding accrual is precise to the last decimal.
The Signal: Four Dimensions, One Output
The Atreidis signal combines four independent market dimensions:
Volume — 3-day On-Balance Volume tracks whether money is flowing in or out. Requires 2 consecutive confirmations before flipping direction, reducing false signals.
Momentum — Normalized MACD scaled 0-100. When momentum is strong during a bullish volume signal, the confirmation requirement is waived — capturing fast moves.
Trend — 120-day SMA for macro direction. Price crosses trigger immediate changes, with buy-the-dip and trailing stop logic built in.
On-Chain Risk — Bitcoin on-chain data creates a Core Risk gauge (0-100), flagging extreme market conditions before they show up in price.
These four dimensions produce a single color-coded output:
- Green: All signals aligned bullish. Full conviction.
- Yellow: Early warning signs. Reduce exposure.
- Orange: Multiple signals turning bearish. Defensive posture.
- Red: Signals aligned bearish. Exit or go short.
The result: 8-12 signal changes per year. Not noise — conviction.
How to Use the Backtest Modeler
The modeler is free and requires no signup. Here's what you can test:
Perpetual Futures Strategy
Configure leverage (1-100x), funding rates, maintenance margins, and liquidation penalties. Choose Hyperliquid or Binance-style defaults, or set custom parameters. See exactly how funding and liquidation would have impacted your strategy.
Spot Strategy
No leverage, no funding, no liquidation risk. Pure signal-following on spot markets. Best for understanding the signal's raw quality without leverage noise.
Buy-and-Hold Comparison
Every strategy is compared against buy-and-hold over the same period. If your signal-based strategy doesn't beat holding, you're paying fees for nothing. We show you the truth.
Metrics You'll See
- Total Return — Strategy P&L as a percentage
- Max Drawdown — Worst peak-to-trough decline
- Win Rate — Percentage of profitable trades
- Calmar Ratio — CAGR divided by max drawdown (higher = better risk-adjusted returns)
- Profit Factor — Gross profit / gross loss
- Equity Curve — Visual growth of $10,000 starting capital over time
What We Don't Do
- We don't optimize to fit history. The signal methodology is fixed. We don't run 10,000 parameter combinations to find one that looks good on past data.
- We don't hide bad periods. The backtest modeler lets you select any date range. Pick the worst bear market. We want you to see it.
- We don't fabricate metrics. Every number on our landing page comes from an actual backtest run. The engine code is the same one you use in the modeler.
Try It Yourself
The best way to evaluate a signal is to break it. Pick adversarial date ranges. Crank up leverage. See where it fails. If a strategy can't survive your stress test, it won't survive the market.
No signup. No API keys. No custody. Just data.