← Back to CopyToLive
📐 Methodology
Plain-English explanation of how a strategy goes from "idea in our farm" to "available for you to copy". No black boxes — every filter and rule is documented here.
The 4-Stage Pipeline
1
Generate (1,000+ candidates)
Our farm continuously generates strategy variants by sweeping parameter spaces (entry triggers, exit rules, timeframes, indicators). Tens of thousands of variants are generated per asset per week.
2
Backtest Filter — Tier-4 Hard Gate
Each candidate must pass all four filters on a 1-3 year historical window:
- PFProfit Factor between 2.0 and 8.0 (below = unprofitable, above = likely overfit)
- TradesAt least 300 trades in backtest window (statistical significance)
- DDMaximum drawdown ≤ 25% on $1k starting capital
- WRWin rate ≥ 30% (very low WR with high R:R is fine if PF still ≥ 2)
Reject rate ~99.5% of generated candidates fail Tier-4.
3
Paper Validation (Forward Test)
Strategies that pass Tier-4 enter paper trading mode with live market data for at least 30 days. Real-time execution, real slippage, real fees — but no real capital. We compare paper PnL vs backtest expectation; deviation > 30% triggers re-evaluation.
4
Live Attach (Limited Capital)
Strategies passing paper validation are attached to a small live capital ($100-500 in our internal account). After 14 more days of consistent live behavior, the strategy is published to subscribers.
Each public strategy on /track-record.html shows both backtest AND live performance side-by-side. We never hide the live underperformance.
What We Do NOT Do
- ❌ We do NOT trade on your behalf with our capital — you trade with YOUR Hyperliquid account using YOUR API key.
- ❌ We do NOT cherry-pick backtest periods (every strategy uses the same continuous window).
- ❌ We do NOT show only winning strategies (failed ones are archived publicly with reason).
- ❌ We do NOT promise specific returns ("3% per week", "100% per year"). Anyone who does is lying.
- ❌ We do NOT use survivorship bias — strategies removed from the active list keep their historical record visible.
How To Read /track-record.html
Each strategy card shows:
- Profit Factor (PF): gross profit / gross loss. PF=2 means $2 won per $1 lost. Higher is better, but PF>8 often signals overfit.
- Win Rate (WR): % of trades that close profitable. NOT the most important metric — high WR with tiny wins + huge losses = bad.
- Max Drawdown (DD): largest peak-to-trough decline. If you cannot stomach a -25% drop, this strategy is not for you.
- Total Trades: sample size. <100 = noise; >300 = meaningful.
Realistic Expectations Calculator
Based on top-decile live performance over the past 12 months (subject to change without notice):
- Conservative profile (5x leverage, 10% daily loss cap): historical range +1% to +5% monthly, drawdowns up to -10%.
- Moderate (10x leverage, 15% daily loss cap): historical range +3% to +12% monthly, drawdowns up to -20%.
- Aggressive (20x leverage, 25% daily loss cap): historical range +5% to +25% monthly, drawdowns up to -40%, occasional liquidation events.
These ranges include losing months. Top strategies have 3-5 negative months in a typical year.
What Happens When a Strategy Stops Working?
- Live 30-day rolling PnL drops below 50% of backtest expectation → flagged "Underperforming" on /track-record.html
- Live 60-day PnL stays below 30% of expectation → strategy is paused (no new signals).
- Subscribers following only paused strategies are notified via email and offered free month extension.
- Paused strategy stays visible with full history — we do NOT delete bad records.
Methodology version 1.0 · Last updated 9 May 2026 · Material changes announced via /changelog and email 30 days in advance.