Home / Position Sizing / Half Kelly
Position sizingAdvanced

Half Kelly

Half Kelly bets half of the full Kelly fraction, 0.5 × f*, trading a small reduction in theoretical long-run growth for a large reduction in volatility, drawdown depth and sensitivity to a mis-estimated edge.

Quick answer: Half Kelly bets half of the full Kelly fraction, 0.5 × f*, trading a small reduction in theoretical long-run growth for a large reduction in volatility, drawdown depth and sensitivity to a mis-estimated edge.

In simple words

Half Kelly is the practical fix for full Kelly being too wild. You compute the Kelly fraction from your edge, then bet only half of it. The remarkable result is that halving the bet keeps most of the long-run growth while cutting the ups and downs roughly in half, so your drawdowns become far more bearable. Because your edge estimate is always uncertain, betting under Kelly also protects you from the danger of having overestimated it. It is the standard way serious traders make Kelly usable.

Purpose

Half Kelly exists because full Kelly is too volatile and too fragile to estimation error for real trading; halving the fraction preserves most growth while sharply improving survivability.

Visual explanation

Half Kelly

The growth curve near its peak is flat, so half Kelly keeps most growth while sitting far from the steep over-betting cliff.

Kelly Fraction & GrowthCapital (log) →Time / trades →Half KellyFull KellyOver-betting

Professional explanation

Why halving the bet keeps most of the growth

The long-run growth rate as a function of bet fraction is a curve that rises to a peak at full Kelly and then falls. Crucially, near the peak the curve is flat, so moving from full Kelly to half Kelly gives up only a small slice of the theoretical growth rate, on the order of a quarter, while it approximately halves the volatility of returns. This favourable trade, most of the growth for much less risk, is the mathematical heart of why half Kelly is preferred. You are stepping back from a sharp peak onto a broad, safer plateau at little cost.

Drawdown reduction is the real prize

Full Kelly is notorious for drawdowns of 50 percent or more, which are intolerable for most traders and often cause abandonment of the strategy at the worst moment. Because drawdown depth scales strongly with bet fraction, halving the fraction dramatically shallows the expected worst drawdown, turning a strategy that is theoretically optimal but practically unusable into one a human can actually run. Since a strategy only compounds if the trader keeps following it, the reduced drawdown of half Kelly can produce better realised results than full Kelly, whose volatility drives quitting.

Robustness against estimation error

The deeper justification for fractional Kelly is that the true edge is never known; W and R are noisy estimates. If the real edge is smaller than estimated, full Kelly is over-betting into the region where growth turns negative, whereas half Kelly leaves a margin of safety that keeps you on the productive side of the curve even if the edge was overstated. Given that the growth penalty for over-betting is far worse than for under-betting, deliberately sizing low is the rational response to uncertainty. Half Kelly is a hedge against your own optimism about your edge.

Half is a convention, not a magic number

There is nothing uniquely optimal about exactly one-half; it is a widely used convention that captures most of the benefit. Traders facing larger estimation error, fatter-tailed return distributions, or greater aversion to drawdown often go lower, to quarter-Kelly or less. The right fraction depends on how uncertain the edge estimate is and how much drawdown can be tolerated without breaking discipline. Half Kelly should be read as shorthand for a conservative fraction of Kelly, with the exact fraction chosen from the strategy's uncertainty and the trader's temperament.

It still inherits Kelly's structural assumptions

Halving the fraction reduces the damage from Kelly's flaws but does not remove them. The single-bet formula still ignores correlation, so half Kelly applied independently to several aligned positions can still over-bet the portfolio. It still assumes a stable edge and independent trades, and it still fails to model fat tails, gaps and margin realities. Half Kelly is a more survivable operating point on the Kelly curve, not a different, safer theory, so the correlation-aware and cost-aware caveats of full Kelly all continue to apply.

Where half Kelly sits relative to the 1 percent rule

For many realistic edges, even half Kelly implies a per-trade risk well above the 1-to-2-percent heuristic, which is why practitioners often treat the fractional-Kelly figure as an upper bound and still cap actual risk lower. When half Kelly and the percentage-risk rule disagree, the more conservative of the two usually wins in live trading, because the cost of over-betting is asymmetric and catastrophic. Half Kelly is best understood as pulling an aggressive theoretical size down toward the survivable range, meeting the percentage-risk discipline from above.

Formula

f_half = 0.5 × f* = 0.5 × (W − (1 − W) ÷ R)

f_half = the half-Kelly fraction actually risked; f* = the full Kelly fraction; W = win probability (0 to 1); R = payoff ratio, average win over average loss. Example: W = 0.5, R = 2 gives f* = 0.25, so f_half = 0.125, i.e. risk 12.5 percent, still large enough that most traders would go lower again. Any fraction below 1 (quarter-Kelly, etc.) follows the same pattern.

Full Kelly vs half Kelly

AspectFull KellyHalf Kelly
Bet fractionf*0.5 × f*
Long-run growthMaximum (100%)About three-quarters of maximum
Return volatilityHighestRoughly halved
Worst drawdownVery deep, hard to tolerateSubstantially shallower
Estimation-error safetyNone, over-betting is punishedMargin of safety if edge was overstated

Practical example

Illustrative example (Indian market)

A trader estimates W of 0.5 and R of 2 on a Nifty strategy, so full Kelly is 0.25, risking Rs 1,25,000 of Rs 5,00,000 per trade, which is reckless. Half Kelly is 0.125, or Rs 62,500, which is still very large and would produce severe drawdowns given that W and R are only estimates. Recognising this, the trader treats half Kelly as an upper bound rather than a target and caps actual risk at 1 to 2 percent, Rs 5,000 to Rs 10,000, far below even half Kelly. The lesson is that half Kelly halves the theoretical size and roughly halves the volatility, but when the edge is uncertain the responsible size sits well below even the half-Kelly figure.

For an F&O account running several correlated index positions, the single-bet half-Kelly figure still over-states safe size because it ignores correlation. A trader who halves each standalone Kelly but holds five aligned Nifty and Bank Nifty trades can still be over-bet at the portfolio level, so the correlation-aware size is smaller again than half Kelly implies.

Advantages

  • Keeps roughly three-quarters of full Kelly growth for about half the volatility
  • Sharply reduces drawdown depth, making the strategy humanly followable
  • Adds a margin of safety against an overestimated edge
  • Sits on a flat part of the growth curve, so small mis-sizing costs little
  • Simple to apply: compute Kelly, then halve it

Limitations

  • Still assumes a known, stable edge and independent trades
  • The single-bet form ignores correlation, so it can over-bet a portfolio
  • Even half Kelly often exceeds prudent per-trade risk for uncertain edges
  • One-half is a convention, not an optimum; some strategies need less
  • Does not model fat tails, gaps or margin constraints

Common mistakes

  • Treating half Kelly as safe rather than as a still-aggressive upper bound
  • Applying half Kelly per position without adjusting for correlation
  • Assuming halving removes Kelly's estimation-error and fat-tail problems
  • Using half Kelly on a lucky, overstated edge and still over-betting
  • Believing exactly one-half is optimal for every strategy
  • Ignoring that even half Kelly can imply drawdowns you cannot tolerate

Professional usage

Practitioners who use Kelly at all almost never run it in full; fractional Kelly, typically a half or a quarter, is the norm on quantitative desks. They exploit the flatness of the growth curve near the peak to give up little growth for much lower drawdown, and they treat the fractional figure as a ceiling that is further reduced for correlation, fat tails and estimation error. The governing principle is that surviving a mis-estimated edge matters more than capturing the last increment of theoretical growth.

Key takeaways

  • Half Kelly bets 0.5 × f*, keeping most growth for roughly half the volatility
  • Its main prize is far shallower, more tolerable drawdowns
  • It hedges against an overestimated edge, since over-betting is punished hardest
  • It still ignores correlation and fat tails, so cap actual risk below it

Frequently asked questions

What is half Kelly?
Half Kelly means betting half of the full Kelly fraction, 0.5 times f*. It gives up only a small part of the theoretical long-run growth while roughly halving return volatility and drawdown depth, which makes a Kelly-style strategy practical to run.
Why bet half of Kelly instead of full?
Because the growth curve is flat near its peak, so halving the bet costs only about a quarter of the growth but halves the volatility. It also protects against an overestimated edge, since betting above true Kelly pushes growth negative, and full-Kelly drawdowns are usually intolerable.
How much growth does half Kelly give up?
Roughly a quarter of the theoretical maximum growth rate, because the growth curve is flat near the peak. In exchange it approximately halves the volatility of returns and substantially shallows drawdowns, which is generally a very favourable trade.
Does half Kelly reduce drawdowns?
Yes, substantially. Drawdown depth scales strongly with the bet fraction, so halving the fraction turns full Kelly's typical 50-percent-plus drawdowns into far shallower ones. Since a strategy only compounds if you keep following it, that reduction often improves realised results.
Is exactly one-half the optimal fraction?
No. One-half is a convention that captures most of the benefit, not a proven optimum. Traders with larger estimation error, fatter tails or lower drawdown tolerance often use quarter-Kelly or less, so half should be read as shorthand for a conservative fraction of Kelly.
Does half Kelly fix Kelly's problems?
It reduces them but does not remove them. Half Kelly still assumes a stable, known edge and independent trades, still ignores correlation in its simple form, and still fails to model fat tails and gaps. It is a safer operating point on the same curve, not a different theory.
Why is under-betting safer than over-betting?
Because the growth curve falls steeply beyond the Kelly peak but gently below it, so the penalty for betting too much is far worse than for betting too little. Since the true edge is uncertain, deliberately sizing low, as half Kelly does, is the rational hedge.
Should I apply half Kelly to each position separately?
Not without adjusting for correlation. The single-bet formula assumes one independent bet, so halving each standalone Kelly across several aligned positions can still over-bet the portfolio. The correlation-aware size is smaller than the sum of independent half-Kelly figures.
Is half Kelly the same as risking 1 to 2 percent?
Usually not; for many edges even half Kelly implies risking well above 2 percent per trade. Because of estimation error and correlation, practitioners often treat half Kelly as an upper bound and cap actual risk lower, near the percentage-risk heuristic.
How do I calculate half Kelly?
Compute the full Kelly fraction f* = W minus one minus W over R, then multiply by 0.5. For W of 0.5 and R of 2, full Kelly is 0.25 and half Kelly is 0.125, or 12.5 percent, which most traders would still reduce further.
Can half Kelly still blow up an account?
It can if the edge was badly overestimated, if positions are correlated, or if fat-tailed losses exceed the model's assumptions. Half Kelly lowers but does not eliminate ruin risk, which is why it should be treated as a ceiling and combined with a firm per-trade cap.
Why do professionals prefer fractional Kelly?
Because surviving a mis-estimated edge matters more than capturing the last bit of theoretical growth. The flatness of the curve near the peak means fractional Kelly sacrifices little growth for large gains in robustness and drawdown control, which is the right trade for real money.
Does half Kelly work for option selling?
Poorly, without care. Option selling has fat-tailed losses that violate Kelly's assumptions, so even half of a high suggested fraction can be dangerous. The tail loss, not the frequent small win, governs safe size, so naive fractional Kelly still misleads here.
When should I use less than half Kelly?
When your edge estimate is very uncertain, your returns are fat-tailed, your positions are correlated, or you cannot psychologically tolerate deep drawdowns. In all these cases quarter-Kelly or a flat 1 to 2 percent cap is more defensible than half Kelly.

Voice search & related questions

Natural-language questions people ask about Half Kelly.

What is half Kelly?
It means working out the Kelly bet size and then betting only half of it. You keep most of the growth but the swings and drawdowns get much smaller.
Why bet half instead of full Kelly?
Because near the best size the growth barely changes, so halving your bet costs little growth but roughly halves the wild swings. It also protects you if you overrated your edge.
Does half Kelly make drawdowns smaller?
Yes, a lot smaller. Full Kelly can drop your account by half or more, which is why most people bet half or even less to keep it bearable.
Is half the perfect amount?
Not exactly. Half is just a popular, sensible choice. If your edge is uncertain or your losses can be huge, many traders go to a quarter or less.
Is half Kelly safe?
Safer than full Kelly, but still aggressive. For many edges it means risking well over two percent a trade, so treat it as a ceiling and often cap lower.
How do I work out half Kelly?
Find the Kelly fraction from your win rate and reward-to-risk, then just multiply it by a half. Many traders then cut it further to stay safe.

Sources & references

    Last reviewed 12 July 2026. Educational content only — not investment advice. Markets and rules change; verify current conventions with SEBI, NSE/BSE and your broker.

    ← Previous
    Kelly Criterion
    Educational content only — not investment advice. Examples use illustrative numbers and simplified models. Risk-management techniques reduce but never remove risk, and trading derivatives involves substantial risk of loss. See our Risk Disclosure and SEBI Disclaimer.