Kelly Criterion Strategy: Rules, Settings, Example

• The core idea: Use mathematical expectation (edge) to determine optimal bet size for each trade.
• Edge is calculated using win rate and average win/loss ratio from backtesting.
• Position size increases when edge is high, and decreases or is skipped when negative.
• Helps avoid overbetting (ruin) and underbetting (slow growth).
• The advantage exists because most traders misjudge sizing, risking too much or too little regardless of edge.
• Works best in strategies with consistent, quantifiable edge and reliable historical data.

Strategy Rules (Step-by-step)

Setup:

1. Determine your trading system with a clear entry/exit rule (e.g., moving average crossover, mean reversion, etc.).
2. Backtest your system over sufficient trades (ideally 100+). Record:
   • Win rate (W): Probability your trade is a winner.
   • Average win (R): Amount gained per winning trade.
   • Average loss (L): Amount lost per losing trade.
3. Calculate edge per trade: E = (W × R) – [(1 – W) × L].

Entry:

• Follow your strategy signal to enter trades as usual.
• For each trade, size your position using the Kelly formula:
• Kelly % = W – [(1 – W)/ (R/L)]
• Position size = Kelly % × your portfolio value. (Most use Half Kelly for safety; e.g., halve the result.)

Stop-loss:

• As dictated by your base strategy (e.g., swing low/high, ATR multiple, or fixed %).

Take profit:

• As dictated by your system (fixed R-multiple, trailing stop, etc.).

Trade management:

• Adjust position size dynamically as edge stats update.
• Reduce risk or skip trades if edge goes negative.

• Indicator settings: Not needed for sizing; rely on base strategy parameters.
• Timeframes tested: Valid on intraday (15m+) to daily charts.
• Assets tested: Stocks (AAPL, SPY), Crypto (BTC, ETH), Forex (EURUSD, GBPUSD).
• Session/Hours: Prefer liquid hours (e.g., NY open for stocks, London–NY overlap for FX, active crypto markets).

When It Works vs. When It Fails

Works best:

• Base strategy provides steady, positive edge over many trades.
• Market volatility is stable; slippage/fees are minor vs. profits.
• Statistics (win rate, average win/loss) are representative of forward performance.

Struggles:

• Markets with regime shifts causing edge reversal.
• Data mining bias or overfitting in backtest creates misleading edge.
• Sudden volatility spikes, major news events, or flash crashes.

Filters to avoid bad conditions:

• Pause trading during/around major news releases.
• If recent performance (last 10–20 trades) drops below historical edge, reduce or stop sizing.
• Use an ATR or volatility filter to skip extremely volatile sessions.

Risk Management (Beginner-safe)

• Position sizing: Never risk more than the calculated Kelly % (typical range 0.5–5% double-checked).
• Max open risk: Total risk on all open trades ≤2% portfolio, or lower with multiple concurrent trades.
• Daily loss limit: Stop trading for the day after 2R or 2 consecutive Kelly-sized losses.
• Fees/slippage: Factor exchange/broker costs and possible execution slippage. Consider halving Kelly output if execution is uncertain.

Example Trade (Walkthrough)

• Pair/Asset: BTC/USDT
• Timeframe: 1h
• Setup snapshot: After backtesting a trend-following breakout system, win rate is 48%, average win 2.0%, average loss 1.0% (R = 2, L = 1).
• Entry: BTC breaks above a key level; buy at $40,000.
• Stop-loss: $39,600 (0.01 BTC position size exposure; based on prior swing low).
• Take profit: $41,000 (2:1 reward–risk).
• Kelly calculation:
  • Win rate W = 0.48, R = 2, L = 1.
  • Kelly % = W – [(1 – W) / (R/L)] = 0.48 – [0.52/2] = 0.22 = 22%
  • Half Kelly used for safety: 11% of $10,000 account = $1,100 position (with 0.01 BTC notional exposure, aligned to stop-loss distance).
• Outcome: Trade wins; $220 gain (2R). Lesson: Conservative Kelly sizing allows compounding but avoids ruin, even after a string of losses.

Pros

• Quantitative, rule-based sizing removes emotion.
• Mathematically maximizes geometric growth.
• Adapts to both high and low-edge environments.
• Risk per trade adjusts dynamically.

Cons

• Needs accurate win rate and edge estimate—bad estimates can lead to overbetting/ruin.
• Large swings if using full Kelly—high drawdowns possible.
• Does not guarantee profits if system edge erodes.

Common Mistakes

• Chasing entries not validated by system statistics.
• Moving stops, which skews win/loss data and edge.
• Over-leveraging by ignoring sharp drops in edge.
• Trading during major news events or after abnormal moves.

Tips and Variations

• Apply a higher timeframe filter for trend consistency.
• Use adaptive ATR-based stops for evolving volatility.
• Combine with scale-outs on partial targets to stabilize edge.
• Set alerts for Kelly threshold or when system stops qualifying.
• Automate with journaling tools or spreadsheet calculators.

Tools You Can Use

• Charting: TradingView, MetaTrader, Thinkorswim.
• Screeners/Alerts: TradingView, TrendSpider, custom scripts.
• Journaling: Edgewonk, TraderVue, Excel/Google Sheets.
• Backtesting: Amibroker, QuantConnect, Tradestation, Python/pandas.

FAQs

• Does it work on crypto? Yes, provided you have reliable edge statistics and trade liquid pairs.
• What timeframe is best? Any, from intraday to daily, as long as your base system is robust.
• What win rate to expect? Typical systems range between 40–60%; Kelly adapts, but a minimum 33% with 2:1 win/loss ratio is preferred.
• Can I automate it? Absolutely; formulas can be coded into scripts, bots, or spreadsheets for live calculation.

Glossary

• EMA: Exponential Moving Average; smooths price for trend viewing.
• ATR: Average True Range; measures volatility.
• R-multiple: Reward-to-risk ratio per trade.
• Drawdown: Decline from equity peak to the next trough.