risk-management · advanced

The Kelly Criterion for Position Sizing

Kelly criterion trading helps you decide how much capital to risk when you have a measurable edge. It can improve position sizing, but only if you use conservative assumptions and control downside risk.

In this lesson, you will learn how the Kelly Criterion works, how to use it for trading position size, and why many professional traders use only a fraction of the suggested amount. The goal is not to maximize excitement; it is to protect capital while growing it efficiently.

What the Kelly Criterion Measures

The <strong>Kelly Criterion</strong> is a mathematical method for deciding what percentage of your capital to risk when the odds are in your favor. In simple terms, it asks: <strong>If I have an edge, what is the optimal position sizing to grow my account over many trades?</strong>

The method was originally used in probability and gambling, but it is also useful in trading because trading is a game of repeated decisions under uncertainty. Every trade has:

  • A chance of winning
  • A chance of losing
  • An average gain when it wins
  • An average loss when it loses
  • In <strong>kelly criterion trading</strong>, the key idea is that position size should increase when your edge is stronger and decrease when your edge is weaker. If there is no edge, the Kelly position is zero.

    This is different from simple fixed-risk rules such as risking 1% per trade. Fixed-risk rules are easier and often safer, but they do not adjust directly to the quality of the setup. Kelly attempts to connect risk size to the actual statistical strength of the strategy.

    However, Kelly is aggressive. Full Kelly can create large drawdowns, which means account declines from a peak to a lower value. For that reason, many traders use <strong>half Kelly</strong>, <strong>quarter Kelly</strong>, or even less.

    The Kelly Formula Position Size

    For trading, the most practical version of the Kelly formula is:

    <strong>Kelly % = W - [(1 - W) / R]</strong>

    Where:

  • <strong>W</strong> = win rate, or the percentage of trades that are profitable
  • <strong>R</strong> = reward-to-risk ratio, meaning average win divided by average loss
  • <strong>Kelly %</strong> = suggested fraction of account equity to risk per trade
  • This is the common <strong>kelly formula position size</strong> used when you know your win rate and average win/loss relationship.

    Example:

  • Win rate: 45%, so W = 0.45
  • Average winning trade: 2.2 times the average losing trade
  • R = 2.2
  • Formula:

    <strong>Kelly % = 0.45 - [(1 - 0.45) / 2.2]</strong>

    <strong>Kelly % = 0.45 - [0.55 / 2.2]</strong>

    <strong>Kelly % = 0.45 - 0.25 = 0.20</strong>

    The full Kelly result is <strong>20% of account equity risked per trade</strong>.

    That number is mathematically valid under the assumptions, but it is usually too high for real trading. A trader might use:

  • <strong>Half Kelly:</strong> 10% risk
  • <strong>Quarter Kelly:</strong> 5% risk
  • <strong>One-tenth Kelly:</strong> 2% risk
  • For most real traders, fractional Kelly is more practical because markets are uncertain, past results may not repeat, and execution costs reduce returns.

    Important: Kelly tells you how much to <strong>risk</strong>, not always how much notional position to open. <strong>Notional position size</strong> means the total dollar value of the trade. If your stop loss is far away, your notional size may be larger or smaller depending on the distance to the stop.

    Practical Trading Examples

    Example 1: Calculating risk from a strategy edge

    Suppose you trade a breakout strategy on major crypto assets. After reviewing 300 trades, you estimate:

  • Win rate: 40%
  • Average win: 3R
  • Average loss: 1R
  • An <strong>R-multiple</strong> compares profit or loss to the amount risked. If you risk $100 and make $300, that is a 3R win.

    Kelly calculation:

    <strong>Kelly % = 0.40 - [(1 - 0.40) / 3]</strong>

    <strong>Kelly % = 0.40 - 0.20 = 0.20</strong>

    Full Kelly says to risk 20% of equity per trade. That is very high. A more realistic choice might be quarter Kelly:

    <strong>20% × 0.25 = 5% risk per trade</strong>

    If your account is $20,000, then 5% risk equals:

    <strong>$20,000 × 0.05 = $1,000 risk</strong>

    If your stop loss is 8% away from your entry and you estimate fees plus slippage at 0.5%, your total risk distance is 8.5%.

    Position notional:

    <strong>$1,000 / 0.085 = $11,764</strong>

    So the trade size is about $11,764, not $1,000. The $1,000 is the amount you expect to lose if the stop is hit after costs.

    Example 2: A weak edge creates a small position

    Now assume another strategy has:

  • Win rate: 52%
  • Average win: 1.1R
  • Average loss: 1R
  • Kelly calculation:

    <strong>Kelly % = 0.52 - [(1 - 0.52) / 1.1]</strong>

    <strong>Kelly % = 0.52 - 0.436 = 0.084</strong>

    Full Kelly is 8.4%. Quarter Kelly is 2.1%.

    This shows why Kelly is useful: even though the win rate is above 50%, the edge is not huge because the average win is only slightly larger than the average loss.

    Example 3: No edge means no trade

    If a strategy has:

  • Win rate: 45%
  • Average win: 1.1R
  • Average loss: 1R
  • Kelly calculation:

    <strong>Kelly % = 0.45 - [0.55 / 1.1]</strong>

    <strong>Kelly % = 0.45 - 0.50 = -0.05</strong>

    A negative Kelly result means the strategy has no positive edge based on these inputs. The correct position size is <strong>zero</strong>. Do not force trades just because a setup looks interesting.

    If you place trades on an exchange such as [CoinW](https://www.coinw.com/en_US/register?r=3443555), Kelly should be calculated before the trade, not after emotions appear on the chart.

    Advanced Risks and Adjustments

    The Kelly Criterion is powerful, but it depends on accurate inputs. This is the main danger.

    1. Your win rate may be wrong

    A strategy with 50 historical trades is not enough to trust the estimate. A few lucky wins can make the Kelly number too large. More trades give better evidence, but even a large sample can fail if market conditions change.

    2. Average win and loss can shift

    Crypto and DeFi markets can move sharply. Slippage, liquidity gaps, failed transactions, bridge delays, and funding costs can turn a normal loss into a larger loss. Kelly assumes the loss size is controlled. If your real loss can be much larger than planned, reduce size.

    3. Trades may be correlated

    <strong>Correlation</strong> means different trades move together. If you take five long positions in similar tokens, they may all lose at the same time when the market drops. Kelly is usually calculated per bet or per strategy, but real portfolios often have overlapping risks.

    A practical adjustment is to apply a <strong>portfolio risk cap</strong>, such as:

  • No more than 5% total account risk across highly correlated trades
  • No more than 10% total risk across all open trades
  • Lower risk before major news, unlocks, or high-volatility events
  • 4. Full Kelly can cause severe drawdowns

    Full Kelly maximizes long-term mathematical growth under ideal assumptions, but it also creates painful account swings. A strategy can be profitable and still suffer losing streaks. If you risk too much, you may abandon the system before the edge has time to work.

    Many advanced traders prefer fractional Kelly because it sacrifices some theoretical growth in exchange for smoother performance. A common approach is:

  • Calculate full Kelly
  • Cut it to half, quarter, or less
  • Apply a maximum risk limit, such as 1% to 3% per trade for volatile assets
  • Recalculate only after enough new data is available
  • 5. Kelly should not replace stop losses

    Kelly helps decide size. It does not remove the need for trade invalidation. <strong>Invalidation</strong> means the price or market condition that proves your trade idea is wrong. You still need a planned exit, a maximum loss, and a clear reason for entering.

    A strong process looks like this:

    1. Define the setup.

    2. Estimate win rate and average win/loss from data.

    3. Calculate Kelly.

    4. Use fractional Kelly.

    5. Convert risk amount into position size using stop distance.

    6. Check portfolio-level risk.

    7. Place the trade only if the risk is acceptable.

    Key Takeaways

  • <strong>The Kelly Criterion estimates the percentage of capital to risk when you have a measurable edge.</strong>
  • <strong>The trading formula is Kelly % = W - [(1 - W) / R], where W is win rate and R is average win divided by average loss.</strong>
  • <strong>Full Kelly is usually too aggressive for real markets, so many traders use half Kelly, quarter Kelly, or less.</strong>
  • <strong>Kelly position size means risk amount, not automatically the total trade notional.</strong>
  • **Bad data, correlated trades, slippage, and large
  • Interactive lesson at /learn/lesson/the-kelly-criterion-for-position-sizing