Expected Value and the Kelly Criterion
Master probability-weighted returns and optimal position sizing for sustainable trading profits
Understanding Expected Value in Trading
Expected value (EV) is the foundation of profitable trading. It answers the fundamental question: “Over time, how much do I expect to win or lose per trade?” Whether you’re aware of it or not, every trade you take has an expected value—and understanding this concept separates consistent winners from random gamblers.
Expected value is calculated by multiplying the probability of each outcome by its payoff, then summing the results. In trading terms:
EV = (Probability of Win × Average Win) – (Probability of Loss × Average Loss)
A positive expected value means that over a large sample of trades, you should profit. A negative EV means you’ll lose money in the long run, no matter how good your individual results might look. This is why professional traders obsess over EV—it’s the mathematical guarantee of long-term success.
Many traders focus on their win rate, but this is misleading. You can have a 30% win rate and be highly profitable if your average wins are large relative to losses. Conversely, a 70% win rate with small wins and occasional large losses leads to ruin. EV captures this relationship and gives you the true picture of profitability.
Probability-Weighted Outcomes
To calculate expected value accurately, you need to assess both the probability of different price movements and the corresponding payoffs. This requires rigorous analysis of your historical trades and realistic forecasting.
Historical Analysis
Review your last 100+ trades (the larger sample, the better). Calculate:
- Win rate: percentage of winning trades
- Loss rate: percentage of losing trades
- Average profit per winning trade
- Average loss per losing trade
- Risk-to-reward ratio: average loss divided by average profit
Realistic Probability Estimates
Don’t use your win rate from a small sample as gospel. Market conditions change. Instead, consider:
- What market regime are we in? (trending, ranging, volatile)
- How does win rate vary by market condition?
- What assets or timeframes have higher edge?
- How do different setups perform?
Historical Data: 65 winning trades averaging +2.1% per trade, 45 losing trades averaging -1.8%
Calculations:
Win rate = 65/110 = 59.1%
Loss rate = 45/110 = 40.9%
EV = (59.1% × 2.1%) – (40.9% × 1.8%)
EV = 1.24% – 0.74% = +0.50% per trade
This means each trade generates 0.50% expected value. Over 12 trades per month, that’s 6% expected monthly return (not accounting for compounding).
The Kelly Criterion Explained
The Kelly Criterion is a mathematical formula that tells you the optimal fraction of your bankroll to risk on each trade to maximize long-term wealth growth while minimizing the probability of ruin. It was developed by John Kelly Jr. in 1956 and has become the gold standard for position sizing in professional trading.
Kelly % = (bp – q) / b
Where:
b = odds received (ratio of profit to risk)
p = probability of winning
q = probability of losing (1 – p)
What the Kelly Criterion Gives You
The Kelly Criterion outputs a percentage of your bankroll to risk. This percentage maximizes the geometric growth of your capital over time. For example, if Kelly says 5%, you should risk 5% of your total account on each trade.
Assume your strategy wins 55% of the time with a 2:1 reward-to-risk ratio (you win $2 for every $1 risked).
b = 2 (the 2:1 ratio)
p = 0.55
q = 0.45
Kelly % = (2 × 0.55 – 0.45) / 2 = (1.10 – 0.45) / 2 = 0.65 / 2 = 32.5%
Full Kelly says to risk 32.5% of your bankroll on each trade.
Fractional Kelly for Risk Management
Full Kelly is mathematically optimal, but it’s psychologically brutal and practically risky. If your edge estimate is wrong by a small margin, full Kelly can devastate your account. This is why professional traders use fractional Kelly—typically 25% Kelly or 50% Kelly.
Why Fractional Kelly?
In the real world, you don’t know your true win rate and edge with certainty. Your estimates are probabilistic. Fractional Kelly provides:
- Robustness: If your edge estimate is slightly wrong, fractional Kelly keeps you alive
- Psychological comfort: Smaller drawdowns are easier to handle emotionally
- Flexibility: You can increase Kelly fraction as you gain confidence in your edge
- Compounding: Even at 25% Kelly, you compound capital effectively
If you optimize your strategy on historical data, your win rate and edge will be artificially inflated. This causes you to overestimate Kelly. Always backtest on out-of-sample data and start with fractional Kelly in live trading.
Applying Kelly to Options Trading
Options trading complicates expected value because payoff distributions are non-normal and probabilities change as the market moves. However, the principles still apply.
Adjusting for Options Non-Linearity
For options, you need to model multiple price outcomes, not just win/loss. Consider:
- Price movement outcomes (e.g., 1%, 2%, 5%, 10% moves)
- Implied volatility changes
- Theta decay (time decay working for or against you)
- Greeks: delta, gamma, vega exposure
Selling a 30-day iron condor on a stagnant stock. You collect $200 premium and risk $800 to expiration.
Your analysis: 72% probability the stock stays inside the spread range, 28% probability it breaches for max loss.
EV = (0.72 × $200) – (0.28 × $800) = $144 – $224 = -$80
This trade has negative expected value! Skip it, even though it seems high-probability. The risk doesn’t justify the reward.
Volatility-Adjusted Position Sizing
For options, adjust your Kelly fraction based on implied volatility. High IV means lower expected return for short premium trades (your edge is smaller), so reduce position size. Low IV creates better value for long options, so you can size larger.
Bankroll Management: Beyond Kelly
The Kelly Criterion solves the position sizing problem mathematically, but professional traders layer additional constraints to manage risk:
Maximum Drawdown Limits
Even with Kelly, inevitable losing streaks occur. Set a maximum acceptable drawdown (typically 20-30% of capital). When you hit this limit, stop trading and reassess your edge. This prevents the “digging out of a hole” mentality that leads to ruin.
Per-Trade Risk Cap
Never risk more than 1-2% of your bankroll on a single trade, even if Kelly suggests more. This is your insurance policy against black swan events and model failures.
Sector and Correlation Limits
Position sizing must account for portfolio correlation. If you’re trading tech stocks, don’t have 80% of your capital in similar names. Diversify across uncorrelated assets to reduce tail risk.
Many professional traders use the “1-2% rule” as their primary risk constraint: never risk more than 1-2% of your total bankroll on any single trade. This ensures that even a sequence of losses won’t destroy your account, and it scales naturally as your bankroll grows.
Practical Implementation: A Complete Example
Let’s build a complete example from analysis to position sizing.
Your Trading Setup
- Bankroll: $50,000
- Strategy: Support/resistance swing trading on ES (micro E-mini S&P 500)
- Recent 60 trades: 35 wins (58%), 25 losses (42%)
- Average win: +$420
- Average loss: -$250
- Risk per trade: $250 (fixed)
- Reward per trade: $420 (fixed)
Calculate Expected Value
EV per trade = (0.58 × $420) – (0.42 × $250) = $243.60 – $105 = $138.60
This strategy has positive expected value of $138.60 per trade. Over 20 trades per month, you expect $2,772 profit.
Calculate Kelly Fraction
b = $420 / $250 = 1.68 (your reward-to-risk ratio)
p = 0.58
q = 0.42
Kelly % = (1.68 × 0.58 – 0.42) / 1.68 = (0.9744 – 0.42) / 1.68 = 0.5544 / 1.68 = 33% per dollar risked
Apply Fractional Kelly (50%)
50% Kelly = 16.5% per dollar risked
With a $250 fixed risk per trade, fractional Kelly is already incorporated. But let’s think about it in bankroll terms:
50% Kelly says risk 16.5% of bankroll per trade = 16.5% × $50,000 = $8,250
But the 1-2% rule overrides this: risk only 1-2% of bankroll = $500-$1,000 per trade
Final Position Size Decision
You’ll risk $500 per trade (1% of bankroll), below Kelly but above your current $250. This increases expected profit while staying within prudent risk limits.
Common Pitfalls to Avoid
Overfitting historical data: A strategy that perfectly fit the past rarely works the future. Test on multiple market regimes and out-of-sample periods.
Ignoring tail risk: Expected value assumes a normal distribution of outcomes, but markets have fat tails. One black swan event can wipe out years of profits if you’re over-leveraged.
Changing position size based on emotions: Stick to your Kelly calculation. After a win streak, don’t increase size out of overconfidence. After losses, don’t decrease out of fear.
Confusing correlation with causation: If two variables have correlated returns, you can’t just multiply their probabilities. Account for dependencies.
Key Takeaways
- Expected value is the average profit/loss per trade over infinite repetitions—it’s the true measure of strategy quality
- Calculate EV from honest historical data: (Win% × Avg Win$) – (Loss% × Avg Loss$)
- The Kelly Criterion is the optimal formula for position sizing: Kelly% = (bp – q) / b
- Use fractional Kelly (25-50%) in practice to account for estimation error and psychological factors
- For options, model multiple price/volatility outcomes, not just win/loss
- Layer additional constraints: max 1-2% per trade, max 20-30% drawdown, diversify across uncorrelated assets
- Recompute your expected value quarterly as market conditions and your edge change