What is the Kelly Criterion and should I use full Kelly for position sizing?

The Kelly Criterion calculates the optimal bet size based on your win rate and average win/loss ratio. The formula is Kelly% = (Win% x Avg_Win/Avg_Loss - Loss%) / (Avg_Win/Avg_Loss). However, full Kelly is extremely aggressive. Always use Half-Kelly or Quarter-Kelly in practice, and only when you have accurate win rate and payoff estimates. If those are uncertain, default to Fixed Fractional sizing.

How much should I risk per trade?

For most traders, risk 1-2% of total portfolio per trade (moderate profile). Conservative traders should risk 0.5%, while aggressive traders can go up to 2-3%. Never risk more than 5% per trade, as losses become exponentially harder to recover from — a 50% loss requires a 100% gain to break even.

What should I do during a drawdown?

Follow a structured drawdown management schedule: at 5-10% drawdown, reduce position sizes by 25%; at 10-15%, reduce by 50% and stop opening new positions; at 15-20%, close the weakest 50% of positions and switch to paper trading; at 20%+, close all positions and conduct a mandatory strategy review with a cooling period.

Risk Management | CLI Trader

This skill is the mandatory gatekeeper for all trade workflows. No position should be opened, sized, or modified without passing through the risk validation framework defined here. It covers the full risk hierarchy — from per-trade sizing to portfolio-level exposure limits — and includes special guardrails for autonomous trading agents.

When to Use This Skill

Before opening ANY new position (mandatory)
When sizing a position for a specific trade setup
When setting or adjusting stop-loss levels
When evaluating risk/reward of a proposed trade
When assessing portfolio-level risk exposure
When managing a drawdown or losing streak
When evaluating leverage on any position
When an autonomous agent is about to execute a trade
When reviewing risk parameters for an existing strategy

What This Skill Does

Position Sizing: Calculates optimal position size using Kelly Criterion, fixed fractional, or volatility-adjusted methods
Stop-Loss Design: Determines stop-loss placement using ATR, technical levels, or time-based methods
Risk/Reward Assessment: Validates minimum R:R ratios by strategy type
Portfolio Risk Check: Enforces concentration limits, correlation-adjusted exposure, and max drawdown thresholds
Drawdown Management: Applies position reduction schedules during drawdowns
Leverage Framework: Assesses leverage risk with liquidation price calculations
Agent Guardrails: Enforces autonomous agent-specific risk limits and kill switches
Risk Validation Gate: Runs mandatory pre-trade checklist — blocks execution if any critical check fails

How to Use

Position Sizing

Size a position for [asset] with entry at [price] and stop at [stop price]

How much [asset] should I buy with a $50,000 portfolio risking 2% per trade?

Risk Assessment

Assess risk/reward for going long [asset] at [entry] targeting [target] with stop at [stop]

Portfolio risk check — show my current exposure and concentration

Drawdown Management

I'm down 15% this month. What should I adjust?

Agent Risk Check

Validate this agent trade: buy 0.5 BTC at $67,000 with 3x leverage

Data Sources

With MCP/CLI tools connected:

Empyrical MCP — Portfolio risk metrics (Sharpe, Sortino, max drawdown, VaR)
yFinance MCPs (tooyipjee, maxscheijen, Adity-star) — Historical volatility, price data for ATR calculation
CoinGecko MCP / CoinGecko Price MCP — Crypto asset volatility, correlation data
Binance MCP / Bybit MCP / OKX MCP — Real-time position data, margin levels, liquidation prices
OpenBB CLI — Portfolio analytics, risk decomposition
DexScreener / DeFiLlama — DeFi protocol TVL, liquidity depth for slippage estimation

Without tool access: Ask the user to provide:

Portfolio size (total capital)
Current open positions and their sizes
Entry price, stop-loss level, and target for the proposed trade
Asset’s recent volatility (or approximate ATR)
Current drawdown percentage (if any)
Leverage being used (if any)

Proceed with full analysis using provided data. Note which metrics are calculated vs. user-provided.

Methodology

Step 1: Establish the Risk Hierarchy

Risk is managed at four levels. Higher levels override lower levels.

Portfolio Level (max 20-25% total risk budget)
  └── Asset-Class Level (max allocation per asset class)
       └── Strategy Level (max allocation per strategy)
            └── Per-Trade Level (max 1-2% risk per trade)

Hard limits by account type:

Level	Conservative	Moderate	Aggressive
Per-trade risk	0.5%	1-2%	2-3%
Max correlated exposure	5%	10%	15%
Max single-asset exposure	10%	20%	30%
Max total portfolio heat	10%	20%	30%
Max drawdown before pause	10%	15%	25%

Step 2: Calculate Position Size

Choose the appropriate sizing method based on the strategy:

Method A: Fixed Fractional (Default — Simplest)

Position Size = (Account × Risk%) / (Entry - Stop)

Example:
  Account: $100,000
  Risk per trade: 1% = $1,000
  Entry: $50.00
  Stop: $47.00
  Risk per unit: $3.00
  Position Size: $1,000 / $3.00 = 333 shares
  Position Value: 333 × $50.00 = $16,650 (16.65% of account)

Method B: Volatility-Adjusted (ATR-Based)

Stop Distance = N × ATR(14)  [typically N = 1.5 to 3.0]
Position Size = (Account × Risk%) / (N × ATR)

Example:
  Account: $100,000
  Risk per trade: 1% = $1,000
  BTC ATR(14): $2,500
  N = 2.0
  Stop Distance: 2.0 × $2,500 = $5,000
  Position Size: $1,000 / $5,000 = 0.2 BTC

Method C: Kelly Criterion (For Strategies with Known Edge)

Kelly% = (Win% × Avg_Win / Avg_Loss - Loss%) / (Avg_Win / Avg_Loss)

Half-Kelly (recommended) = Kelly% / 2

Example:
  Win rate: 55%
  Average win: $2,000
  Average loss: $1,000
  Win/Loss ratio: 2.0
  Kelly% = (0.55 × 2.0 - 0.45) / 2.0 = 0.325 = 32.5%
  Half-Kelly = 16.25%

WARNING: Full Kelly is extremely aggressive. Always use Half-Kelly or Quarter-Kelly in practice. Kelly requires accurate win rate and payoff estimates — if these are uncertain, default to Fixed Fractional.

Method D: Equal Risk Contribution (Portfolio Context)

Weight_i = (1 / Volatility_i) / Σ(1 / Volatility_j)

Useful for: Multi-asset portfolios where you want each position
to contribute equal risk.

Step 3: Determine Stop-Loss Placement

Select stop method based on strategy type:

Strategy Type	Primary Stop Method	Backup Method
Trend-following	ATR trailing (2-3× ATR)	Swing low/high
Mean reversion	Fixed % (tight, 1-2%)	Time-based (3-5 bars)
Breakout	Below breakout level	ATR-based
DeFi yield	IL threshold + TVL drop	Smart contract risk event
Prediction markets	Probability threshold	Time-based (event deadline)
Scalping	Fixed ticks/pips	Time-based (seconds/minutes)

ATR-based stops (preferred for most strategies):

Long Stop = Entry - (Multiplier × ATR14)
Short Stop = Entry + (Multiplier × ATR14)

Conservative: 3.0× ATR (wider, fewer stopped out)
Standard:     2.0× ATR (balanced)
Aggressive:   1.5× ATR (tighter, more stopped out)

Trailing stop adjustment:

Trailing Stop = Highest High since entry - (Multiplier × ATR14)
Update: Only move stop UP (longs) or DOWN (shorts), never against the trade

Step 4: Validate Risk/Reward Ratio

Minimum acceptable R:R by strategy type:

Strategy	Min R:R	Typical Win Rate	Expectancy Check
Trend-following	2:1	35-45%	Wins are large, losses are small
Mean reversion	1.5:1	55-65%	Higher win rate compensates
Breakout	3:1	25-35%	Most breakouts fail — need large winners
Scalping	1:1	60-70%	Volume of trades compensates
DeFi yield	2:1	N/A (yield vs IL)	Annual yield > expected IL + risk
Prediction markets	1.5:1	Depends on edge	Expected value must be positive

Expectancy formula:

Expectancy = (Win% × Avg Win) - (Loss% × Avg Loss)

Must be positive. If negative, DO NOT TAKE THE TRADE.

Example:
  Win rate: 40%, Avg win: $3,000
  Loss rate: 60%, Avg loss: $1,000
  Expectancy = (0.40 × $3,000) - (0.60 × $1,000) = $1,200 - $600 = $600
  Per-trade expected value: +$600 ✓

Step 5: Portfolio-Level Risk Controls

Run these checks against the entire portfolio before approving a new trade:

## Portfolio Risk Checklist

1. CONCENTRATION CHECK
   - No single position > 20% of portfolio (crypto: 10%)
   - No single sector/protocol > 30% of portfolio
   - Top 3 positions < 50% of portfolio

2. CORRELATION CHECK
   - Sum of highly correlated positions (ρ > 0.7) < 25% of portfolio
   - Example: BTC + ETH + SOL are highly correlated — treat as one bucket
   - Crypto + tech stocks often correlate — check cross-asset correlation

3. PORTFOLIO HEAT
   - Total open risk (sum of all position risk%) < 20% of portfolio
   - Calculation: Σ(position_size × distance_to_stop / portfolio_value)

4. LIQUIDITY CHECK
   - Position size < 1% of asset's daily volume (equities)
   - Position size < 0.5% of asset's daily volume (crypto)
   - DeFi: Position < 5% of pool liquidity

5. MARGIN/LEVERAGE CHECK
   - Total margin used < 50% of available margin
   - No single position using > 5x leverage (crypto: 3x max recommended)
   - Maintenance margin buffer > 50% above liquidation

Step 6: Drawdown Management

When portfolio hits drawdown thresholds, automatically reduce risk:

Drawdown	Action	New Per-Trade Risk
0-5%	Normal trading	1-2%
5-10%	Reduce position sizes by 25%	0.75-1.5%
10-15%	Reduce position sizes by 50%, no new positions	0.5-1%
15-20%	Close weakest 50% of positions, paper trade only	0% live
20%+	FULL STOP — close all positions, mandatory review	0%

Recovery rules:

After a 10%+ drawdown, require 5 consecutive profitable paper trades before resuming live
After a 20%+ drawdown, mandatory 2-week cooling period + strategy review
Scale back into live trading at 50% normal size, increase by 25% every 5 profitable trades

Step 7: Leverage Risk Framework

If leverage is being used, add these additional checks:

LIQUIDATION DISTANCE
  Liq Price (Long) = Entry × (1 - 1/Leverage + Maintenance Margin%)
  Liq Price (Short) = Entry × (1 + 1/Leverage - Maintenance Margin%)

  MINIMUM: Liquidation price must be > 2× ATR away from current price

LEVERAGE LIMITS BY ASSET CLASS
  Large-cap crypto (BTC, ETH): Max 3x
  Mid-cap crypto (top 50):     Max 2x
  Small-cap / meme coins:      NO LEVERAGE (1x only)
  US equities:                  Max 2x (Reg-T)
  Futures:                      Follow exchange margins, max 5x effective
  DeFi (on-chain):              Max 2x (IL + liquidation compound risk)

LEVERAGE RISK MULTIPLIER
  Adjust position size: Leveraged Size = Base Size / Leverage
  Example: 3x leverage → position is 1/3 of what unleveraged sizing suggests

Step 8: Agent-Specific Risk Guardrails

When an autonomous trading agent is executing, apply these additional constraints:

## Agent Risk Limits (on top of all above rules)

1. PER-TRADE LIMIT
   - Agent max per-trade risk: 0.5% (half of human limit)
   - Agent max position size: 5% of portfolio

2. DAILY LIMITS
   - Max daily loss: 2% of portfolio → auto-pause agent
   - Max trades per day: 10 (prevents runaway loops)
   - Max total daily volume: 10% of portfolio

3. APPROVAL GATES
   - Trades > 1% of portfolio: require human approval
   - New asset (never traded before): require human approval
   - Leverage > 2x: require human approval
   - Any trade outside predefined strategy parameters: require human approval

4. KILL SWITCHES
   - Drawdown > 5% in single session: auto-halt, notify human
   - 3 consecutive losses: pause for 1 hour, notify human
   - API error or unexpected response: halt immediately
   - Market volatility spike (>3 std dev move): pause, notify human

5. AUDIT TRAIL
   - Log every trade decision with reasoning
   - Log every risk check pass/fail
   - Log portfolio state before and after each trade

For comprehensive agent safety architecture, see Agent Trading Guardrails.

Step 9: Mandatory Risk Validation Gate

THIS CHECKLIST MUST PASS BEFORE ANY TRADE IS EXECUTED.

Every trade — manual or agent — must clear this gate. If any CRITICAL item fails, the trade is blocked.

## Pre-Trade Risk Validation

### CRITICAL (any failure = trade blocked)
- [ ] Position size ≤ max per-trade risk% of portfolio
- [ ] Stop-loss is defined and placed
- [ ] Risk/reward ratio meets minimum for strategy type
- [ ] Portfolio heat stays under limit with this trade added
- [ ] No single-asset concentration limit breached
- [ ] If leveraged: liquidation price > 2× ATR from current price
- [ ] If agent trade: within agent-specific risk limits
- [ ] Sufficient liquidity (position < volume threshold)

### WARNING (flag but allow if acknowledged)
- [ ] Correlation check: not adding to already-correlated cluster
- [ ] Drawdown status: not in reduced-risk mode
- [ ] Timing: not trading into major news event / high-volatility period
- [ ] Slippage estimate: acceptable for position size

### INFO (log for audit)
- [ ] Strategy type identified
- [ ] Entry rationale documented
- [ ] Time horizon specified
- [ ] Exit plan (target + stop) documented

Anti-Patterns

DO NOT do these — they are the most common risk management failures:

Averaging down without a plan: Adding to a losing position increases risk. Only add if the original thesis is intact AND a specific add-down level was pre-planned.
Revenge trading: After a loss, immediately entering a new trade to “make it back.” Always follow the drawdown management schedule.
Moving stops further away: Once a stop is set, it can be tightened but NEVER widened. Moving a stop further from entry is equivalent to increasing risk after the trade is already going against you.
Risk of ruin: Risking >5% per trade makes account recovery exponentially harder.
- 10% loss requires 11.1% gain to recover
- 25% loss requires 33.3% gain to recover
- 50% loss requires 100% gain to recover
Ignoring correlation: Having 5 “diversified” positions that are all correlated to BTC means you effectively have one large BTC bet.
Overleveraging on “high conviction”: Conviction is not a risk parameter. Size positions based on math, not feelings.
No stop-loss: Every position needs a defined exit point. “I’ll watch it” is not a risk management strategy.

Validation Checkpoints

Input Validation

Portfolio size provided or fetched from connected exchange
Entry price, stop-loss, and target clearly defined
Asset volatility available (ATR or implied vol)
Current portfolio positions known (for concentration checks)
Strategy type identified (determines R:R minimums and stop methods)

Output Validation

Position size calculated with specific method cited
Dollar risk amount and percentage stated explicitly
Stop-loss level and method documented
R:R ratio calculated and meets minimum threshold
Portfolio-level impact assessed (heat, concentration, correlation)
All items in Risk Validation Gate checked
If leveraged: liquidation price calculated and buffer confirmed
If agent: agent-specific limits verified

Example

User: “I want to go long ETH at $3,500 with my $50,000 portfolio. How should I size this?”

Analysis:

## Risk Assessment: Long ETH at $3,500

### Position Sizing (Volatility-Adjusted)
- Portfolio: $50,000
- Risk per trade: 1% = $500
- ETH ATR(14): $180
- Stop method: 2× ATR below entry
- Stop price: $3,500 - (2 × $180) = $3,140
- Risk per unit: $3,500 - $3,140 = $360
- Position size: $500 / $360 = 1.39 ETH
- Position value: 1.39 × $3,500 = $4,861 (9.7% of portfolio)

### Risk/Reward Assessment
- Target: $4,200 (2× ATR above entry or prior resistance)
- Reward per unit: $4,200 - $3,500 = $700
- Risk per unit: $360
- R:R = $700 / $360 = 1.94:1 ✓ (meets 1.5:1 minimum for trend-following)

### Portfolio Impact
- Current portfolio heat: 8% (existing positions)
- New trade adds: 1% risk
- Post-trade heat: 9% ✓ (under 20% limit)
- ETH correlation with existing BTC position (ρ = 0.85): combined crypto = 14% ⚠️ (near 15% limit)
- Concentration: 9.7% in ETH ✓ (under 20% limit)

### Risk Validation Gate
- [x] Position size (1%) ≤ max risk (2%) ✓
- [x] Stop-loss defined: $3,140 ✓
- [x] R:R (1.94:1) ≥ minimum (1.5:1) ✓
- [x] Portfolio heat (9%) < limit (20%) ✓
- [x] Concentration (9.7%) < limit (20%) ✓
- [x] Not leveraged ✓
- [x] Liquidity OK (ETH daily volume >> position size) ✓
- [⚠️] Correlation: crypto exposure at 14%, approaching 15% limit

### Recommendation
APPROVED with WARNING: Position is within all risk parameters.
Monitor total crypto exposure — one more correlated position would breach the 15% limit.

Position: Buy 1.39 ETH at $3,500
Stop-loss: $3,140 (2× ATR)
Target: $4,200
Risk: $500 (1% of portfolio)