copy-trading Checked

Run scripts/evaluate_wallet.py for a comprehensive copy-trade suitability assessment.

### Stage 3 — Filtering

After evaluation, apply additional filters:

• **Consistency check**: rolling 7-day win rate should not swing below 40% in any period
• **Style compatibility**: understand whether the wallet is a sniper, scalper, or swing trader — your infrastructure must match their speed
• **Size compatibility**: if they trade 500 SOL per position and you have 10 SOL total, proportional sizing may be too small to cover fees
• **Sybil check**: use the sybil-detection skill to verify the wallet is not part of a wash-trading cluster

### Stage 4 — Monitoring

Once a wallet passes evaluation and filtering, set up real-time monitoring.

**Monitoring approaches (fastest to simplest):**

1. **Yellowstone gRPC**: sub-second latency, streams all transactions for subscribed wallets

2. **Helius Enhanced WebSocket**: near-real-time with parsed transaction data

3. **Polling via RPC**: getSignaturesForAddress every 5-10 seconds — simple but slower

See references/execution_strategy.md for implementation details on each approach.

### Stage 5 — Execution

When a monitored wallet executes a swap:

1. **Detect** the transaction (via monitoring infrastructure)

2. **Parse** the trade: token address, direction (buy/sell), size

3. **Validate** the token: check liquidity, holder distribution, honeypot risk

4. **Size** the position: proportional to your portfolio, not theirs

5. **Execute** via Jupiter aggregator with appropriate slippage tolerance

6. **Record** the copy trade with attribution to the source wallet

### Stage 6 — Risk Management

Copy trades require independent risk controls that do not depend on the copied wallet's behavior.

See references/risk_framework.md for the complete framework.

**Key limits:**

| Control | Recommended Value | Purpose |

|---------|------------------|---------|

| Max allocation per wallet | 10-20% of portfolio | Diversification across signal sources |

| Max concurrent copy positions | 3-5 | Prevent overexposure |

| Per-trade stop loss | -15% to -25% | Independent downside protection |

| Daily copy-trade loss limit | -5% of portfolio | Circuit breaker |

| Weekly copy-trade loss limit | -10% of portfolio | Longer-term circuit breaker |

Wallet Scoring for Copy Suitability

Composite score from 0-100 based on weighted criteria:

copy_score = (
    trade_count_score * 0.15 +
    win_rate_score * 0.20 +
    profit_factor_score * 0.25 +
    consistency_score * 0.20 +
    recency_score * 0.10 +
    human_probability_score * 0.10
)

**Component calculations:**

• **Trade count score**: min(trade_count / 200, 1.0) * 100 — maxes out at 200 trades
• **Win rate score**: max((win_rate - 0.40) / 0.30, 0) * 100 — scaled from 40% to 70%
• **Profit factor score**: min((pf - 1.0) / 3.0, 1.0) * 100 — scaled from 1.0 to 4.0
• **Consistency score**: (1.0 - std_dev_of_rolling_win_rate) * 100 — lower variance = higher score
• **Recency score**: max(1.0 - days_since_last_trade / 14, 0) * 100 — decays over 14 days
• **Human probability score**: (1.0 - bot_probability) * 100

**Interpretation:**

| Score | Rating | Action |

|-------|--------|--------|

| 80-100 | Excellent | Strong copy-trade candidate |

| 60-79 | Good | Suitable with monitoring |

| 40-59 | Marginal | Proceed with caution, reduce allocation |

| 0-39 | Poor | Do not copy |

Position Sizing for Copy Trades

Three approaches, from simplest to most nuanced:

### Fixed Amount

Use a constant SOL amount per copy trade (e.g., 0.5 SOL). Simple but ignores the wallet's conviction level.

### Proportional

Match the copied wallet's allocation as a percentage of their estimated portfolio:

your_size = (their_trade_size / their_estimated_portfolio) * your_portfolio

Requires estimating their total portfolio, which can be imprecise.

### Confidence-Scaled

Base amount multiplied by your confidence in the wallet:

your_size = base_amount * (copy_score / 100) * conviction_multiplier

Where conviction_multiplier is higher for wallets with longer track records.

Anti-Patterns to Avoid

### Copying Bots

Bots have sub-second execution and often use MEV strategies. You cannot match their latency. By the time you detect their trade, the opportunity is gone or you become the exit liquidity. Use the bot probability score to filter these out.

### Survivorship Bias

A wallet with 1000% returns from one PumpFun token is not necessarily skilled. Look for wallets with consistent performance across many tokens, not outlier wins. The max-single-trade-PnL filter catches this.

### Blind Following

Never copy a trade without understanding what the token is. At minimum, run basic safety checks: liquidity depth, holder concentration, contract verification. A 2-second check can prevent buying a honeypot.

### No Independent Exits

The copied wallet may have information you do not have. They may exit for reasons unrelated to the trade. Always maintain your own stop loss. Never rely solely on mirroring their exit.

### Correlation Risk

If you copy 5 wallets and