agiprolabs/claude-trading-skills/exit-strategies

Exit Strategies

Entries are easy, exits are everything. A mediocre entry with a disciplined exit will outperform a perfect entry with no exit plan. This skill covers systematic, rule-based exit methods for crypto and Solana token trading.

Why Exits Matter

• Entries determine _if_ you participate. Exits determine _how much_ you keep.
• Most traders spend 90% of effort on entries and 10% on exits — invert this.
• Without defined exits you rely on emotion, which guarantees inconsistency.
• Every trade should have three exits defined before entry: stop loss, take profit,

and trailing stop.

Exit Categories

1. Stop Loss — Risk Management Exits

Predefined price level where you close the position to cap downside.

ATR-based stop (recommended default):

import pandas_ta as ta
atr = df.ta.atr(length=14)
stop_loss = entry_price - (atr.iloc[-1] * 2.0)  # 2x ATR below entry

Multiplier guide:

• 1.5× — Tight. High win rate needed. Good for scalps.
• 2.0× — Standard. Balances noise filtering with risk.
• 3.0× — Wide. For swing trades in volatile conditions.

See references/stop_loss_methods.md for complete methodology.

2. Take Profit — Target Exits

Predefined levels where you lock in gains.

Fixed risk/reward targets:

risk = entry_price - stop_loss_price
tp_2r = entry_price + (risk * 2)  # 2:1 R:R
tp_3r = entry_price + (risk * 3)  # 3:1 R:R
tp_5r = entry_price + (risk * 5)  # 5:1 R:R

Scaled exit framework (recommended for meme/PumpFun tokens):

Market cap milestone exits:

For PumpFun and meme tokens where R:R ratios are less meaningful:

milestones = [
    {"mcap": 50_000,  "sell_pct": 0.25, "label": "Cover cost"},
    {"mcap": 100_000, "sell_pct": 0.25, "label": "Lock profit"},
    {"mcap": 500_000, "sell_pct": 0.25, "label": "Major profit"},
    # Hold 25% as moonbag with trailing stop
]

See references/take_profit_strategies.md for full methodology including Fibonacci extension targets and volume-based exits.

3. Trailing Stop — Trend-Following Exits

Dynamic stops that follow price upward but never move down.

Percentage trailing:

def percentage_trailing_stop(
    current_price: float,
    highest_since_entry: float,
    trail_pct: float = 0.10,
) -> tuple[float, bool]:
    """Return (stop_level, triggered)."""
    highest = max(highest_since_entry, current_price)
    stop = highest * (1 - trail_pct)
    return stop, current_price <= stop

ATR trailing (Chandelier Exit):

def chandelier_exit(
    highs: list[float],
    atr_value: float,
    multiplier: float = 2.5,
    lookback: int = 22,
) -> float:
    """Highest high over lookback minus ATR * multiplier."""
    highest_high = max(highs[-lookback:])
    return highest_high - (atr_value * multiplier)

EMA trailing:

# Exit when close < EMA for M consecutive bars
ema = df.ta.ema(length=20)
below_ema = df["close"] < ema
consecutive_below = below_ema.rolling(3).sum() == 3  # 3 bars below

Typical EMA periods: 10 (scalp), 20 (day trade), 50 (swing).

See references/trailing_stops.md for Parabolic SAR, SuperTrend, and step trailing.

4. Time-Based Exits

Exit if the trade hasn't moved in your favor within a defined window.

bars_since_entry = current_bar - entry_bar
if bars_since_entry > max_hold_bars and current_pnl <= 0:
    exit_reason = "time_stop"

Guidelines:

• Scalp: 5–15 minutes
• Day trade: 4–8 hours
• Swing: 3–5 days
• PumpFun snipe: 2–10 minutes (token-specific)

Time stops prevent capital from sitting in dead trades.

5. Signal-Based Exits

Exit when the indicator that generated the entry signal reverses.

# RSI reversal exit
rsi = df.ta.rsi(length=14)
if position == "long" and rsi.iloc[-1] > 70:
    exit_reason = "rsi_overbought"
# MACD crossover exit
macd = df.ta.macd()
if macd["MACDs_12_26_9"].iloc[-1] < macd["MACDh_12_26_9"].iloc[-1]:
    exit_reason = "macd_bearish_cross"

Signal exits work well when combined with trailing stops — the signal triggers tightening the trail rather than an immediate full exit.

6. Liquidity-Based Exits

Exit when volume or liquidity deteriorates, signaling reduced ability to exit cleanly.

recent_vol = df["volume"].rolling(10).mean().iloc[-1]
baseline_vol = df["volume"].rolling(50).mean().iloc[-1]
if recent_vol < baseline_vol * 0.3:  # Volume dropped to 30% of baseline
    exit_reason = "liquidity_deterioration"

Critical for low-cap Solana tokens where liquidity can evaporate rapidly.

PumpFun-Specific Exit Rules

PumpFun tokens have unique dynamics requiring specialized exit logic.

Pre-Graduation Exits

Tokens on the bonding curve before reaching 85 SOL fill:

bonding_fill_pct = current_fill_sol / 85.0
if bonding_fill_pct > 0.90:
    # Near graduation — decide: hold through or exit before
    # Graduation creates volatility spike, both up and down
    pass
if bonding_fill_pct < 0.50 and time_since_entry > 300:  # 5 min
    exit_reason = "stalled_bonding_curve"

Volume Decay Exits

buy_vol_1m = get_buy_volume(token, "1m")
buy_vol_5m = get_buy_volume(token, "5m") / 5  # Normalize to per-minute
if buy_vol_1m < buy_vol_5m * 0.3:
    exit_reason = "buy_volume_decay"

Time Decay for PumpFun

Most PumpFun tokens that will succeed show momentum within the first few minutes:

Combining Exit Rules

A complete exit plan layers multiple rules. Here is a recommended template:

exit_plan = {
    "hard_stop": {
        "type": "fixed_percentage",
        "value": 0.20,  # -20% max loss
        "priority": 1,   # Checked first, always honored
    },
    "atr_stop": {
        "type": "atr_trailing",
        "multiplier": 2.5,
        "atr_length": 14,
        "priority": 2,
    },
    "take_profit": {
        "type": "scaled",
        "tranches": [
            {"at_rr": 2, "sell_pct": 0.25},
            {"at_rr": 4, "sell_pct": 0.25},
            {"at_rr": 8, "sell_pct": 0.25},
        ],
        "priority": 3,
    },
    "time_stop": {
        "type": "max_bars",
        "value": 50,
        "condition": "if_not_profitable",
        "priority": 4,
    },
}

Priority hierarchy: Hard stop > ATR trailing > Take profit > Time stop.

The hard stop is always active and never overridden. The ATR trailing stop activates after the first take-profit tranche fills. The time stop only fires if the trade is not yet profitable.

Common Exit Mistakes

Integration with Other Skills

• `position-sizing` — Size the position based on the stop loss distance.

position_size = (account_risk * account_balance) / (entry - stop_loss)

• `risk-management` — Exits are the mechanism that enforces risk limits.
• `pandas-ta` — Use ATR, EMA, RSI, MACD for signal-based and trailing exits.
• `slippage-modeling` — Estimate execution cost of the exit to set realistic targets.
• `liquidity-analysis` — Verify exit liquidity before entering a position.

Files

References

• references/stop_loss_methods.md — Complete stop loss methodology and anti-patterns
• references/take_profit_strategies.md — Scaled exits, R:R targets, Fibonacci extensions
• references/trailing_stops.md — Trailing stop implementations and parameter guidance

Scripts

• scripts/exit_simulator.py — Simulate and compare exit strategies on synthetic price data
• scripts/stop_loss_calculator.py — Calculate stop levels, position sizes, and R:R targets