Position Sizing with Kelly Criterion for Crypto Trading

How much should you risk per trade? Too little and you barely grow. Too much and one losing streak wipes you out. Kelly Criterion gives a mathematically optimal answer.

What Is the Kelly Criterion?

The Kelly Criterion is a formula that tells you what fraction of your capital to bet on each trade, given your win rate and payoff ratio. It was developed by John Kelly at Bell Labs in 1956 and is used by professional gamblers, traders, and investors.

The formula:

Kelly % = W - (1 - W) / R

Where:
W = Win probability (e.g., 0.55 for 55% win rate)
R = Win/Loss ratio (average win ÷ average loss)

One thing the formula’s name hides, and most articles get wrong: Kelly % is the fraction of your capital to put at risk, not the size of your position. If your stop-loss is 10%, a position’s notional is ten times the capital you are risking on it. We convert between the two explicitly below.

A Real Example

Using our BB Squeeze SHORT preset’s current backtest (live run, 2026-08-12, 2,416 trades on the top-50 universe):

Win rate (W): 53.02%
Average win: 5.13% of position
Average loss: 4.96% of position
Win/Loss ratio (R): 5.13 / 4.96 = 1.034

Kelly % = 0.5302 - (1 - 0.5302) / 1.034
Kelly % = 0.5302 - 0.4543
Kelly % = 0.076 (7.6% of capital at risk per trade)

Full Kelly says put 7.6% of capital at risk per trade. With a 10% stop-loss that means a position notional of ~76% of your account (risk ÷ stop distance). In practice, no one uses full Kelly.

Why Full Kelly Is Dangerous

Full Kelly optimizes for maximum long-term growth rate, but the ride is brutal:

Expected drawdowns: 50-90% drawdowns are mathematically normal at full Kelly
Assumption of perfect edge: Your actual win rate and payoff ratio have estimation error
No margin of safety: Any overestimation of your edge leads to massive losses

The solution: Fractional Kelly

Most practitioners use 1/4 to 1/2 Kelly:

Fraction	Capital at Risk (our example)	Character
Full Kelly	7.6%	Extremely risky
1/2 Kelly	3.8%	Aggressive
1/4 Kelly	1.9%	Moderate
1/10 Kelly	0.76%	Conservative
1/20 Kelly	0.38%	Very conservative

What We Actually Use

The simulator’s default sizing (its own methodology page states this) is:

Capital: $10,000
Position: $60 margin at 5x leverage = $300 notional
Stop-loss: 10% → capital at risk per trade = $30 = 0.3% of the account
Against the 7.6% full Kelly above, that is roughly 1/25 Kelly

Why so conservative?

Multiple simultaneous positions: We can have up to 100 positions open. Even at 1/4 Kelly (1.9% at risk each), 15 simultaneous positions would put ~29% of capital at risk — and since crypto positions are highly correlated, that behaves closer to one oversized bet than to 15 independent ones. The rigorous treatment (Thorp’s multivariate Kelly) sizes positions off the covariance between them, and it shrinks every position sharply versus the one-asset formula.
Estimation uncertainty: Backtested win rates can differ from real-world results due to execution costs, market regime changes, and other factors. Conservative sizing protects against this uncertainty.
Correlation risk: In crypto, when BTC drops, most altcoins drop together. Simultaneous losses across 50 positions are not independent events.
Survival priority: The first rule of trading is to survive. A 50% drawdown requires a 100% gain to recover. A 20% drawdown only requires 25%.

Position Sizing with Leverage

A critical mistake: forgetting that leverage multiplies both exposure and risk.

# Without leverage
Margin = $200, notional = $200
Risk if 10% SL hit = $20 (0.2% of $10,000 capital)

# With 5x leverage
Margin = $200, but controls $1,000 notional
Risk if 10% SL hit = $100 (1% of $10,000 capital)

How leverage actually enters the math — it doesn’t divide Kelly. Kelly gives you the capital to put at risk; the stop-loss converts risk into notional; leverage only determines how much margin that notional requires:

Capital at risk   = Kelly fraction × capital
Position notional = capital at risk ÷ stop-loss distance
Margin required   = notional ÷ leverage

Quarter Kelly on $10,000 (1.9% at risk):
  risk     = $190
  notional = $190 / 0.10 = $1,900
  margin   = $1,900 / 5  = $380

(A correction from an earlier version of this article: it divided the Kelly fraction by the leverage multiplier. That double-counts — the stop-loss already fixes how much of the notional you can lose, so leverage changes the margin requirement, not the risk. Dividing by it again produced position sizes ~10x smaller than the framework it described.)

Practical Position Sizing Framework

Here’s a step-by-step approach:

Step 1: Calculate Your Edge

Use at least 500 trades of data (backtest or live):

Win rate
Average win %
Average loss %

Step 2: Apply Kelly Formula

Kelly % = W - (1-W) / R

Step 3: Choose Your Fraction

Aggressive (1/4 Kelly): You’re confident in your edge, have limited positions, low correlation
Moderate (1/10 Kelly): Multiple positions, some correlation, moderate confidence
Conservative (1/20 Kelly): Many simultaneous positions, high correlation (crypto), uncertain edge

Step 4: Convert Risk into a Position

Notional = (Kelly fraction × capital) ÷ stop-loss distance. Then margin = notional ÷ leverage. Leverage sets the margin requirement — it does not change the risk the stop-loss already fixed.

Step 5: Verify with Maximum Drawdown

Run a Monte Carlo simulation with your position size:

Simulate 10,000 random sequences of trades
Check: What’s the worst-case drawdown?
If max drawdown > 30%, reduce position size

The Mathematical Edge of Survival

Consider two traders:

Trader A (aggressive): 10% per trade, 70% win rate

After 10 consecutive losses (probability: 0.0006%): -65% drawdown
Needs +186% to recover

Trader B (conservative): 2% per trade, 70% win rate

After 10 consecutive losses: -18% drawdown
Needs +22% to recover

Both have the same edge. But Trader B survives any streak the market throws at them. In our backtest, we found stretches of 10 consecutive losing days. At 2% per trade, these periods are uncomfortable but survivable.

Key Takeaways

Kelly Criterion gives the mathematically optimal bet size, but full Kelly is too aggressive for real trading
Use fractional Kelly — 1/10 to 1/20 is appropriate for crypto with multiple positions
Always account for leverage when calculating position sizes
Correlation matters — 50 crypto positions are not 50 independent bets
Survival beats optimization — a small consistent edge with proper sizing beats a large edge with reckless sizing

PRUVIQ’s default simulation uses fixed $60 positions at 5x leverage — with a 10% stop that is ~0.3% of a $10,000 account at risk per trade, roughly 1/25 Kelly at current estimates. See our risk management approach and strategy library.