Kelly’s Criterion Explained: Formula, Examples & Position Sizing Guide

Most trading education spends 90% of its time on one question: what to buy or sell. Almost none of it spends time on the question that actually determines survival, which is how much.

That imbalance shows up in the numbers. A SEBI study released in July 2025 found that individual traders’ net losses in equity derivatives widened 41% year-on-year to ₹1,05,603 crore in FY25, up from ₹74,812 crore in FY24, with more than 91% of individual F&O traders ending the year in the red. Many of these traders were not necessarily wrong about direction. They were wrong about size.

Kelly’s Criterion is the mathematical framework built specifically to answer the “how much” question. Originally derived for a problem that had nothing to do with markets, it has since become one of the most referenced position-sizing formulas in trading and investing literature. This guide breaks down where it came from, how the formula works, how to calculate it with real numbers, and where it fails if used carelessly.

What Is Kelly’s Criterion?

Kelly’s Criterion is a formula that calculates the mathematically optimal fraction of capital to risk on a bet or trade, in order to maximise the long-term growth rate of that capital.

It was not built by a trader. It was built by John L. Kelly Jr., a scientist at Bell Labs, who introduced it in a 1956 paper titled A New Interpretation of Information Rate, published in the Bell System Technical Journal. Kelly’s original problem was about information theory. How a gambler with inside knowledge of a noisy communication channel should size bets to grow capital at the maximum possible rate.

The leap from information theory to trading came later, largely through Edward O. Thorp, the mathematician known for card-counting in blackjack who formalised the connection in his 1992 paper “The Kelly Criterion and the Stock Market” (The American Mathematical Monthly). Thorp’s core argument: the same logic that tells a blackjack player how much to bet when the odds are in their favour can tell a trader how much of their capital to allocate to a position when their strategy has a statistical edge.

Why Position Sizing Matters in Trading

A trading strategy can have a genuine, positive statistical edge and still destroy an account. The mechanism is simple. Percentage losses compound against you faster than percentage gains recover them. A 50% drawdown requires a 100% gain just to break even.

This is precisely why the SEBI data above is so relevant to this topic. Over 91% of individual F&O traders lost money in FY25, and the average losses were not the result of a single bad call, they accumulated over many trades. Position sizing decisions, repeated hundreds of times, matter more to the final account balance than any individual entry signal.

Most retail traders size positions by instinct, by how confident a setup feels, or by a round-number lot size. Kelly’s Criterion replaces that instinct with a calculation forcing the trader to first quantify their actual edge before deciding how much capital that edge justifies risking.

How Kelly’s Criterion Works

Kelly’s Criterion assumes a trader or gambler has a quantifiable edge. A probability of winning that differs from breakeven, and a payoff that rewards being right. Given those two inputs, the formula calculates the fraction of capital that maximises the expected compounded growth rate of that capital over many repeated bets, not the expected value of a single bet.

This distinction matters. Betting too little under-uses a real edge and leaves growth on the table. Betting too much increases volatility and drawdown risk to the point where, even with a positive edge, the compounding math works against the trader over time. A large enough loss can erase the gains of several winning trades and take disproportionately long to recover from. Kelly’s formula identifies the size that sits at the balance point between these two failure modes.

Kelly’s Criterion Formula

The original formula is written as:

f* = (bp − q) / b
which simplifies to:
f* = p − (q / b)

Where:

• f\* = the fraction of capital to allocate
• p = probability of a winning trade
• q = probability of a losing trade (1 − p)
• b = the payoff ratio — how much is won per unit risked

Where W is the historical win rate of the strategy, and R is the win/loss ratio (average winning trade ÷ average losing trade).

Kelly’s Criterion Formula Explained with an Example

Consider a hypothetical trading strategy used here purely to illustrate the formula, not as a recommendation to trade it, with the following backtested statistics:

• Win rate (W) = 55%, or 0.55
• Average winning trade = ₹3,000
• Average losing trade = ₹2,000
• Win/loss ratio (R) = 3,000 ÷ 2,000 = 1.5

Applying the formula:

f* = 0.55 − [(1 − 0.55) / 1.5]
f* = 0.55 − (0.45 / 1.5)
f* = 0.55 − 0.30
f* = 0.25, or 25%

Full Kelly, in this illustrative case, suggests allocating 25% of trading capital to each position. As the next section on Full vs Half vs Quarter Kelly shows, most practitioners would never use this number directly but it is the starting point from which every other calculation is built.

How to Calculate Kelly’s Criterion

Calculating Kelly’s Criterion for a real strategy follows a repeatable process:

1. Build a sufficient trade sample. A handful of trades will not produce a statistically meaningful win rate. Most practitioners look for at least 30–50 closed trades from a consistent strategy before treating the numbers as reliable.
   
2. Calculate the win rate (W). Divide the number of winning trades by the total number of trades.
   
3. Calculate the win/loss ratio (R). Divide the average profit of winning trades by the average loss of losing trades.
   
4. Apply the formula. f* = W − [(1 − W) / R].
   
5. Check the sign. If f* comes out negative, the formula is stating that the strategy, as measured, does not have a positive edge and Kelly’s own logic says the correct position size is zero, not a small positive number.
   
6. Recalculate periodically. Win rate and payoff ratio are not fixed constants, they drift as market conditions and strategy performance evolve, so the calculation needs refreshing as new trade data comes in.

How Traders Use Kelly’s Criterion

In practice, Kelly’s Criterion is rarely used as a direct instruction to “risk this exact percentage.” It is more commonly used as a reference ceiling, a mathematically derived upper bound that traders then scale down from, based on their own risk tolerance and the reliability of their win-rate and payoff estimates.

A hypothetical trader running a systematic strategy again, illustrative only, not a recommendation, might use the Kelly output as one input into a broader position-sizing model that also accounts for a fixed stop-loss, maximum exposure per sector, and total portfolio risk across multiple simultaneous positions. This matters even more in instruments like index futures and options, where leverage means a Kelly-derived percentage translates into a much larger notional exposure, precisely the segment where SEBI’s FY25 data shows the highest concentration of retail losses.

Full Kelly vs Half Kelly vs Quarter Kelly

Full Kelly is the mathematically optimal fraction if the win rate and payoff ratio inputs are exactly correct. In real trading, they never are. They are estimates drawn from a limited, backward-looking sample, and markets change regime. Small errors in those estimates get amplified by the formula, and full Kelly sizing is known to produce sharp, uncomfortable drawdowns even for strategies with a genuine edge.

This is why Thorp and later researchers on the Kelly Capital Growth framework (MacLean, Thorp, and Ziemba’s work on fractional Kelly) generally advocate using a fraction of the full Kelly output, commonly half or a quarter, in real-world applications. The trade-off is straightforward: fractional Kelly gives up some theoretical long-term growth rate in exchange for a meaningfully smaller reduction in volatility and drawdown depth.

Continuing the earlier example (full Kelly = 25%):

For most retail applications and especially for leveraged instruments like F&O, given how concentrated losses are in that segment per SEBI’s data half or quarter Kelly is the more commonly discussed starting point in the literature, not full Kelly.

Advantages of Kelly’s Criterion

• Removes guesswork from sizing. It replaces “this trade feels right” with a calculation based on a defined win rate and payoff ratio.
• Prevents systematic under-betting. Traders with a genuine edge often under-size out of caution, leaving compounding growth on the table; Kelly quantifies what the edge actually justifies.
• Forces honest edge measurement. To use the formula at all, a trader must calculate their actual historical win rate and payoff ratio, a discipline many traders otherwise skip entirely.
• Adapts as the edge changes. Because it is a formula rather than a fixed rule, position size scales up or down as a strategy’s live performance data evolves.

Limitations of Kelly’s Criterion

• It depends on accurate inputs. Unlike a casino game with fixed, known odds, a trading strategy’s win rate and payoff ratio are estimates from a finite historical sample and estimation error feeds directly into the formula’s output.
• It assumes independence between trades. The classical formula assumes each bet is statistically independent. Real trading positions are often correlated. Multiple open positions in the same sector or direction behave more like one large bet than several independent ones, which the simple formula does not account for.
• Full Kelly can be extremely aggressive. As the earlier example showed, a 25% per-position allocation is significantly larger than most risk frameworks would consider prudent, particularly in leveraged instruments.
• It ignores practical trading frictions. Margin requirements, liquidity constraints, slippage, and transaction costs are not part of the base formula, yet all affect what is actually achievable.

Common Mistakes When Using Kelly’s Criterion

• Using an unreliable win rate. A win rate calculated from 10–15 trades, or from a cherry-picked backtest window, is not a stable enough input to build a position-sizing decision on.
• Ignoring a negative result. If the formula returns a negative f*, it is signalling no edge. Continuing to trade the strategy anyway defeats the purpose of running the calculation at all.
• Applying full Kelly to leveraged instruments. Given how concentrated retail losses are in derivatives per SEBI’s FY25 study, applying an un-fractioned Kelly output to F&O positions compounds an already high-risk setup.
• Treating the output as permanent. A Kelly calculation reflects the trade sample it was built from. It is not a one-time number to be set and forgotten.
• Sizing each position independently in a correlated portfolio. Applying Kelly per trade without checking overlap across simultaneously open positions can understate real portfolio-level risk.

Conclusion

Kelly’s Criterion is not a stock-picking tool, and it is not a signal to trade any specific instrument. It is a discipline for answering the question retail trading education most often skips. Given a measurable edge, how much of my capital does that edge actually justify risking? Used directly, full Kelly is often too aggressive for real-world trading. Used as a reference point scaled down to a half or quarter fraction, recalculated periodically, and applied alongside sound risk management, it gives traders a repeatable, data-grounded process instead of a guess.

Frequently Asked Questions (FAQs)