Why Do Professional Traders Use Half-Kelly Instead of Full Kelly?

Half-Kelly—betting half of the theoretical Kelly fraction—is the closest thing to a universal standard among professional traders, hedge funds, and proprietary trading firms. If full Kelly maximizes long-term wealth but creates psychologically unbearable volatility, half-Kelly finds the sweet spot: it delivers roughly 75% of Kelly's long-term growth while cutting maximum drawdown in half (from 40% to 15-20%).

The gap between theory and practice exists because full Kelly assumes two conditions that real traders never meet: first, that you'll stick to your system through drawdowns that can exceed 40% of account equity, and second, that your backtest perfectly predicts live trading results. Neither assumption holds. A single deviation—quitting during a 35% drawdown, slightly overestimating your win rate, or hitting a market regime change—and full Kelly explodes your account. Half-Kelly is the practical answer: it retains the mathematical magic while introducing a buffer against human and market friction.

Quick definition: Half-Kelly is applying 50% of your calculated Kelly fraction. If your Kelly formula yields 30%, you risk 15% of account per trade. This reduces maximum drawdown from 40% to 15-20% while preserving 75% of Kelly's wealth accumulation rate.

Key takeaways

• Half-Kelly halves your position size compared to full Kelly, directly cutting maximum drawdown roughly in half
• The wealth accumulation rate drops only 25% compared to full Kelly (from optimal to ~75% of optimal), a favorable trade-off
• Half-Kelly is the institutional standard: pension funds, CTAs, hedge funds, and proprietary desks use it routinely
• The reduction in position size acts as insurance against edge estimation errors, market regime changes, and backtest overfitting
• Psychology and institutional constraints favor half-Kelly; full Kelly is a research benchmark, not a practical tool

The Math of Half-Kelly vs. Full Kelly

Start with a full Kelly calculation:

Full Kelly = (p × b - q) / b

Half-Kelly divides this by 2:

Half-Kelly = [(p × b - q) / b] / 2

Example: A trader has a 55% win rate (p = 0.55), a 1.5:1 odds ratio (b = 1.5), and a loss probability q = 0.45.

Full Kelly = (0.55 × 1.5 - 0.45) / 1.5
Full Kelly = (0.825 - 0.45) / 1.5
Full Kelly = 0.375 / 1.5
Full Kelly = 0.25 (25%)

Half-Kelly = 0.25 / 2 = 0.125 (12.5%)

Full Kelly says risk 25% of account per trade; half-Kelly says 12.5%.

On a $100,000 account:

• Full Kelly: $25,000 risk per trade
• Half-Kelly: $12,500 risk per trade

The position size is literally halved. But what about the long-term wealth accumulation?

Wealth Accumulation: The Trade-Off

Using the log-wealth formula from the Kelly derivation, the expected log-return per bet is:

E[log-return] = p × log(1 + b × f) + q × log(1 - f)

Let's compare full Kelly (f = 0.25) versus half-Kelly (f = 0.125) for our example:

Full Kelly (f = 0.25):

E = 0.55 × log(1 + 1.5 × 0.25) + 0.45 × log(1 - 0.25)
E = 0.55 × log(1.375) + 0.45 × log(0.75)
E = 0.55 × 0.3185 + 0.45 × (-0.2877)
E = 0.1752 - 0.1295
E = 0.0457 (or 4.57% per trade)

Half-Kelly (f = 0.125):

E = 0.55 × log(1 + 1.5 × 0.125) + 0.45 × log(1 - 0.125)
E = 0.55 × log(1.1875) + 0.45 × log(0.875)
E = 0.55 × 0.1722 + 0.45 × (-0.1335)
E = 0.0948 - 0.0601
E = 0.0347 (or 3.47% per trade)

Over 100 trades:

• Full Kelly: Wealth multiplier = e^(100 × 0.0457) = e^4.57 ≈ 97× growth
• Half-Kelly: Wealth multiplier = e^(100 × 0.0347) = e^3.47 ≈ 32× growth

Half-Kelly delivers 32× growth versus full Kelly's 97×. That's roughly one-third the growth. But here's the crucial point: over a typical trading career of 500+ trades, half-Kelly still produces compounding that vastly outpaces fixed dollar or overly conservative sizing. And crucially, you're still alive—you haven't quit mid-drawdown, and your account hasn't blown up.

Drawdown Reduction: The Psychological Shield

Maximum drawdown is the largest decline from peak-to-trough in account equity. Full Kelly exposes you to 30-40% drawdowns; half-Kelly reduces this to 15-20%.

Consider a streak of 10 consecutive losses (statistically plausible even with a 55% win rate):

Full Kelly (25% per trade, starting at $100,000):

After loss 1: $100,000 - $25,000 = $75,000 After loss 2: $75,000 - $18,750 = $56,250 After loss 3: $56,250 - $14,062 = $42,188 ... After loss 10: ≈ $5,620 (a 94% drawdown)

Half-Kelly (12.5% per trade, starting at $100,000):

After loss 1: $100,000 - $12,500 = $87,500 After loss 2: $87,500 - $10,938 = $76,562 After loss 3: $76,562 - $9,570 = $66,992 ... After loss 10: ≈ $26,629 (a 73% drawdown)

Wait—both look terrible! This reveals an important caveat: the drawdown reduction applies to typical drawdown sequences, not worst-case scenarios. A 10-loss streak is an extreme event. More typical are sequences of 3-5 consecutive losses interspersed with wins.

Let's simulate a more realistic 20-trade sequence with an actual 55% win rate: W, W, L, W, L, W, W, W, L, L, W, W, L, W, W, L, W, W, L, W (11 wins, 9 losses, the expected outcome).

Full Kelly: Wealth grows to $100,000 × 1.25^11 × 0.75^9 ≈ $170,000 (17% gain). Peak-to-trough varies but typically 15-25%.

Half-Kelly: Wealth grows to $100,000 × 1.125^11 × 0.875^9 ≈ $128,000 (28% gain). Peak-to-trough typically 8-12%.

In more realistic sequences, half-Kelly's drawdown is roughly 60-70% of full Kelly's drawdown. That's meaningful: a 12% drawdown is psychologically survivable; a 25% drawdown tempts you to question your system.

Institutional Adoption: Why the Pros Use Half-Kelly

Large institutions (pension funds, endowments, hedge funds, CTAs) universally use fractional Kelly, primarily half-Kelly. Why?

First, estimation risk. Your backtest says 55% win rate, but the true win rate (over infinite future trades) is 52%. This 3-percentage-point error barely matters with half-Kelly; it's catastrophic with full Kelly. By using half-Kelly, you're implicitly hedging against the gap between historical and future edge.

Second, market regime changes. Your system crushes in trending markets but bleeds in choppy markets. Full Kelly sizes for an average regime; half-Kelly survives a worse-than-average regime without ruin.

Third, compliance and risk management. Institutional portfolio managers report to boards and regulators. A 40% drawdown triggers scrutiny; a 15% drawdown is a "normal market correction." Half-Kelly satisfies both mathematical optimality and institutional risk constraints.

Fourth, delegation and consistency. If your fund employs multiple traders, each using half-Kelly, the aggregate portfolio risk is predictable and manageable. Full Kelly introduces tail-risk scenarios where multiple traders simultaneously hit a 40% drawdown (unlikely but possible), collapsing the fund.

Calculating Half-Kelly for Your System

The recipe is straightforward:

Step 1: Backtest 50+ trades of your system on out-of-sample data.

Step 2: Extract statistics.

• Win rate: p = (number of winning trades) / (total trades)
• Average winner: W (in dollars or points)
• Average loser: L (in absolute value, positive number)
• Odds ratio: b = W / L

Step 3: Calculate full Kelly.

Full Kelly = (p × b - (1 - p)) / b

Step 4: Divide by 2 for half-Kelly.

Half-Kelly = Full Kelly / 2

Step 5: Apply to account.

Risk per trade = Account Equity × Half-Kelly

Example: A trader has a $80,000 account and a system with p = 0.58, b = 1.8.

Full Kelly = (0.58 × 1.8 - 0.42) / 1.8 = (1.044 - 0.42) / 1.8 = 0.624 / 1.8 = 0.347 (34.7%)
Half-Kelly = 0.347 / 2 = 0.1735 (17.35%)
Risk per trade = $80,000 × 0.1735 = $13,880

She risks roughly $13,880 per trade using half-Kelly.

When to Use Half-Kelly vs. Full Kelly vs. Quarter-Kelly

Use full Kelly only if:

• You have 200+ backtested trades with statistically consistent results
• You've forward-tested (live or paper) and confirmed the backtest edge
• You have institutional backing or can absolutely tolerate a 40% drawdown without emotional deviation
• Your system has no evidence of curve-fitting or overfitting

Use half-Kelly if:

• You have 50-100 backtested trades
• You're trading with personal capital and need to sleep at night during drawdowns
• You work at an institution with risk management oversight
• You want a practical compromise between growth and survivability

Use quarter-Kelly if:

• You have fewer than 50 backtested trades
• Your backtest win rate is borderline (52-55%), leaving little room for estimation error
• You're just starting to forward-test a new system
• You want maximum insurance against overfitting and regime change

A typical progression: start with quarter-Kelly on a new system, advance to half-Kelly after 50 live trades confirming the backtest, advance to full Kelly only after 200+ live trades of consistent results (which many traders never reach).

The Adjustment for Slippage, Commissions, and Fees

Real trading includes friction costs. If your backtest ignores slippage and commissions, your measured edge is inflated. You must reduce half-Kelly accordingly:

Slippage assumption: 0.5 points average per trade (the difference between backtested fills and realistic fills).

Commission: $10 per round-trip trade.

On ES futures (where 1 point = $50, contracts are 1 quantity):

• Slippage cost: 0.5 points × $50 = $25
• Commission: $10
• Total per-trade friction: $35

If your backtest assumed a $200 average winner, the real winner is $200 - $35 = $165. Your effective odds ratio shrinks. Recalculate Kelly with the adjusted number, then apply half-Kelly.

Most professionals add a 10-20% haircut to the calculated Kelly, regardless of the derivation. This buffer for unknown unknowns is the practical equivalent of saying, "The formula is good, but I'm humble about how good."

Half-Kelly Decision Sequence

Position Limits with Half-Kelly

Half-Kelly tells you the fraction to risk, but you also need a hard cap on capital deployed per trade. A typical rule: Never deploy more than 15% of account in a single position.

Example: You have a $100,000 account, half-Kelly says 10% risk = $10,000. Your stop is 50 points on an ES contract.

Contracts = $10,000 / (50 × $50) = $10,000 / $2,500 = 4 contracts
Capital deployed = 4 × 1 ES multiplier × entry price

If ES is at 5000, 4 contracts ties up $20,000 notional (ignoring margin), or 20% of your account. This exceeds your 15% cap. You either reduce the risk to 7.5%, widen the stop, or skip the trade.

This hybrid approach—half-Kelly for risk allocation, plus a capital deployment cap—is how professionals actually operate. Kelly gives the risk; position limits give the absolute ceiling.

Tracking Half-Kelly in Your Trade Log

A sample spreadsheet row for half-Kelly execution:

Date	Account	H-Kelly %	$ Risk	Stock	Entry	Stop	Shares	Cap Deploy	Exit	P/L	New Acct
5/10	$100k	12.5%	$12,500	XYZ	$150	20	625	$93,750	$165	$9,375	$109,375

The "H-Kelly %" column is calculated quarterly from your rolling backtest; the "$ Risk" column is Acct × H-Kelly %; the "Cap Deploy" is Shares × Entry. If Cap Deploy ever exceeds 15% of Account, you reduce shares or pass on the trade.

Real-World Examples

Example 1: Swing trader, gradual progression. Maria backtests a 60-trade swing-trade system: 58% win rate, 1.7:1 odds ratio. Full Kelly = 0.344 (34.4%), Half-Kelly = 17.2%. She has a $50,000 account, so she'll risk $8,600 per trade on average. After 30 live trades with a 59% win rate (confirming the backtest), she's confident and increases to 18% half-Kelly. After 100 live trades, she advances to full Kelly (34.4%), but only on trades with highest confluence. Over 18 months, her account grows from $50k to $140k using half-Kelly, then accelerates with selective full-Kelly on high-conviction trades.

Example 2: Futures day-trader, institutional constraint. A proprietary trading firm allocates $500,000 per trader and mandates half-Kelly position sizing across all desks. A day-trader on the ES contract calculates half-Kelly = 8% of current P&L per trade. Over a day with 40 trades, her aggregate risk is 320% of daily P&L—which is the firm's risk appetite for a single trader. The firm's total exposure stays manageable across 20 traders, each running half-Kelly independently.

Example 3: CTA (commodity trading advisor), regulatory requirement. A registered CTA manages $50 million across 15 clients. Regulators and institutional LPs demand Sharpe ratios > 1.0 and maximum annual drawdown < 20%. The CTA uses quarter-Kelly (half of half) on most positions and full Kelly only on highest-edge systems, ensuring drawdown stays under 18% even in stressed markets. This "conservative Kelly" approach trades growth for institutional acceptability.

Common Mistakes

Calculating full Kelly but forgetting to divide by 2. You compute 30% full Kelly, then accidentally risk 30% per trade instead of 15%. You're running full Kelly without realizing it, and you hit a 35% drawdown that you weren't psychologically prepared for.

Using outdated Kelly from an old backtest. You calculated half-Kelly three years ago and haven't recalculated. Your system's edge has degraded, but you're still sizing as if it's optimal. Recalculate at least semi-annually, especially after regime changes (e.g., post-Fed policy shift, volatility spike).

Not accounting for commissions. Your backtest assumes $0 commission; your real account pays $10 per round-trip. You calculated Kelly on inflated edge numbers. Adjust your win rate and odds ratio downward to account for real costs.

Mixing half-Kelly on multiple concurrent positions. You hold three positions, each sized to 12.5% half-Kelly risk. Your aggregate risk is 37.5%—too much. Either reduce individual position sizes or limit concurrent positions to two.

Abandoning half-Kelly during a drawdown. You've committed to half-Kelly, but after 5 consecutive losses (-6.25% × 5 = -31.25% account in theory, less in practice), you panic and cut position size in half again. Now you're using quarter-Kelly at exactly the wrong moment (when your system is likely in a temporary drawdown, not broken). Stick to the framework or accept that you can't handle it.

FAQ

How does half-Kelly compare to fixed fractional sizing (e.g., 2% per trade)?

Fixed fractional (2%) allocates a fixed percentage regardless of your measured edge. Half-Kelly allocates a percentage based on your edge—higher for stronger systems, lower for marginal systems. Half-Kelly is more efficient if your edge estimate is accurate; fixed fractional is simpler and more forgiving if your edge estimate is uncertain.

Can I use different Kelly fractions for different trade types?

Yes. If you have two systems (one 58% win rate, one 52% win rate), calculate separate half-Kelly fractions for each. Apply the appropriate fraction based on trade type.

What if my half-Kelly is less than 0.1% (very small edge)?

This is a red flag. A half-Kelly of 0.1% means your edge is tiny—you're barely beating commissions. Before trading, confirm the backtest is sound and the edge is real. If it is, trading at 0.1% will be slow and painful; consider improving your system first.

Should I use half-Kelly or fixed fractional sizing if I'm unsure about my backtest?

If unsure, use fixed fractional (1-2%). It's simpler and doesn't require accurate edge estimation. Once you have 100+ live trades confirming your backtest, switch to half-Kelly.

How do I handle Kelly fraction when my trades have different stop losses?

Kelly assumes a consistent odds ratio. If your stops vary widely (5-50 points), your odds ratio varies too. Options: (a) calculate separate Kelly fractions for each stop-distance category, or (b) use average odds ratio across all trades. Most traders use the average approach for simplicity.

Is half-Kelly still optimal if I rebalance quarterly?

Yes. Rebalancing—recalculating Kelly based on the latest 50-100 trades—is essential. Your edge changes as markets evolve. Quarterly rebalancing ensures you're always sizing to your current (not historical) edge.

Related concepts

• The Kelly Criterion: Full Treatment
• Where the Kelly Criterion Comes From
• Quarter-Kelly for Conservative Traders
• Fixed Fractional Position Sizing
• What Is a Stop-Loss?

Summary

Half-Kelly is the practical middle ground between mathematical optimality (full Kelly) and human psychology (quarter-Kelly or fixed fractional). By allocating 50% of your calculated Kelly fraction, you reduce maximum drawdown from 40% to 15-20% while preserving roughly 75% of Kelly's long-term wealth accumulation. This trade-off—losing 25% of theoretical edge to gain psychological survivability—has made half-Kelly the institutional standard among hedge funds, CTAs, and professional traders.

To implement half-Kelly: backtest 50+ trades, calculate your win rate and odds ratio, plug into the Kelly formula, divide by 2, and size each trade to that fraction of current account equity. Add hard limits on capital deployment (no more than 15% of account per position) and rebalance quarterly as your edge evolves. Combined with discipline to stick to your system through 15-20% drawdowns, half-Kelly will compound your wealth at a rate that vastly exceeds fixed dollar or overly conservative sizing.

Half-Kelly is not the maximum-growth solution; it's the maximum-growth solution that humans can actually execute.

Next

→ Quarter-Kelly for Conservative Traders