Fractional Kelly: A Smarter Way to Size Bets When Your Edge Is Fuzzy

A coin that quietly favors you

Imagine a friend offers you a game. The coin is slightly biased: it lands heads 60% of the time. You can bet any fraction of your bankroll — the money you have set aside for this game — and you win even money on heads. How much do you wager each toss?

Most people would not know what to do. In one real experiment, people were given $25 and a biased coin like this, with 30 minutes to play. The results were humbling: 28% of players went bust, and the average payout was just $91, even though the game was mathematically in their favor. Only 21% reached the maximum prize. The math was not the problem; deciding how much to bet was.

That question has a name: the Kelly criterion. It was developed in 1956 by John Kelly, a researcher at Bell Labs, who was originally trying to solve a problem about noise on telephone lines. It did not stay in the phone company for long. Gamblers, blackjack players, sports bettors and investors soon realized that the same logic could answer a very practical question: how much of your capital should you risk on an opportunity where you have an edge?

The formula, without the mystery

The Kelly criterion calculates a fraction of your bankroll to bet. The classic version uses three numbers: p, your probability of winning; q, your probability of losing, which is simply 1 minus p; and b, how much you win on average compared with how much you lose. If your average winner is $300 and your average loser is $200, then b is 1.5. The formula is:

Kelly fraction = (b × p − q) ÷ b

A negative answer means you do not have an edge. A positive answer tells you the bet size that, in theory, grows your wealth fastest over many repetitions.

Try it with the biased coin. Your win probability p is 0.6, q is 0.4, and because the bet is even money, b is 1. The formula gives (1 × 0.6 − 0.4) ÷ 1, which is 0.2. So the Kelly recommendation is to bet 20% of your bankroll on every toss.

Why not more? Because growth is about compounding, not just average wins. Betting 20% produces an expected geometric gain of about 2% per round, which compounds to a surprisingly large amount over hundreds of tosses. Betting too much makes your results swing wildly, and a few losses can shrink your bankroll so much that the compounding works against you.

The uncomfortable secret: you don't know your edge

The Kelly criterion is mathematically beautiful, but it has a catch. It assumes you know the true probabilities. That is a reasonable assumption for a coin you have tested for years. It is almost never true in investing.

Your win rate and your average win-to-loss ratio are estimates from past data. Those estimates carry error. In one worked example, a trader with a 55% win rate and a 1.5 reward-to-risk ratio gets a full Kelly suggestion of about 25% of the account per trade. Drop the estimated win rate by just five percentage points, and the suggested bet size falls to roughly 8% in the same example — a swing of more than three times. The same uncertainty that makes markets interesting is exactly what makes full Kelly fragile.

Worse, overconfidence hurts more than caution. If you overestimate your edge, the formula hands you a bet size that is too large. Betting more than the Kelly amount does not just reduce growth; it increases the risk of ruin. Underbetting, by contrast, costs you some growth but keeps you in the game.

Why fractional Kelly is the grown-up version

Full Kelly may be optimal in theory, but it is demanding in practice. A full Kelly bettor has roughly a one in three chance of halving the bankroll before doubling it. That kind of drawdown — the drop from a high point to a low point — makes people panic, abandon a sound plan, or simply lose sleep. It also assumes you can weather enormous swings that most humans, and most accounts, cannot.

Fractional Kelly means betting a fraction of the full Kelly number: half Kelly, quarter Kelly, or something in between. The trade-off is surprisingly gentle. Half Kelly keeps roughly 75% of the full Kelly growth rate while cutting a large share of the volatility. Quarter Kelly slows growth further, but it can bring position sizes close to the familiar “risk about 1% of your account” rule that many professionals already use.

This is not a fringe idea. In practice, many professional traders and investors deliberately use between 25% and 75% of the full Kelly figure. Gamblers who use Kelly in casinos report that fractional Kelly is far easier to stick with emotionally. The math puts a similar warning on leverage: when the Kelly fraction comes out above 1, it is telling you to borrow money to bet more than your whole bankroll. That can work in theory, but it magnifies estimation errors, margin calls, and every wrong guess.

Using Kelly without fooling yourself

If you want to use Kelly in real decisions, the practical version starts with humility. Base your numbers on a trustworthy sample — at least 50 to 100 closed trades, not a handful of recent wins. Estimate the win rate and the average win-to-loss ratio from actual results, not from hope.

Keep an eye on correlation. Kelly assumes each bet is independent. If you are holding two positions that tend to move together, such as two similar tech stocks, applying Kelly to each one separately can double or triple your true risk. Treat correlated positions as one combined exposure.

And remember that Kelly is not a permanent number. It changes as your edge changes. A short losing streak does not mean your edge has disappeared; recalculating everything after a few bad trades is a good way to sell low and buy high. Negative Kelly, on the other hand, is a clear signal: the strategy has negative expected value, and no amount of fancy position sizing will save it.

A thinking tool, not a crystal ball

The Kelly criterion deserves its reputation. No fixed-fraction betting strategy grows faster in the long run, as long as you know your edge and can survive the ride. But that “as long as” is doing a lot of work.

The real lesson is more useful than the formula itself. When your edge is uncertain, betting less is not cowardice; it is arithmetic. Full Kelly assumes a world with precise probabilities, and markets do not live there. Fractional Kelly is the practical compromise: it keeps most of the growth, reduces the chance of a painful drawdown, and gives you room to be wrong — which, as any investor will tell you, you certainly will be.

None of this is a recommendation for any particular bet. Your own goals, risk tolerance, and circumstances matter far more than any single percentage. Think of Kelly as a compass, not a command.