Full Kelly is the math-optimal stake size. It's also the wrong choice for almost every real bettor. This lesson is the why — and how to land on the fraction (½, ¼, or somewhere in between) that actually keeps you in the game long enough for your edge to compound.
The Kelly recap
Kelly tells you what percentage of your bankroll to stake on a bet, given your edge. For a binary outcome, the formula simplifies to edge ÷ odds, and it produces the stake that maximizes the long-run growth rate of your bankroll.
"Long-run growth rate" sounds like exactly what you want. Why isn't it?
The variance problem
Full Kelly maximizes growth in expectation. The expectation hides a brutal truth about the path it takes to get there.
Picture a 55% bettor with a +3% edge per bet, using full Kelly. Mathematically they'll compound nicely. In practice, the path their bankroll takes is jagged. Drawdowns of 40, 50, even 60% off peak are routine. Not unusual events — routine.
Here's the math behind the discomfort. Kelly's growth rate is a function of your average outcome. But because betting outcomes are multiplicative (each bet's result compounds onto the new bankroll), variance hurts you asymmetrically. Lose 50%, win 50% back, and you're down to 75% of where you started — not even. Variance is a tax. Full Kelly minimizes that tax for someone with a perfectly known edge. You don't have a perfectly known edge.
The estimation problem
Kelly assumes your edge estimate is exact. If you think a bet is +3% EV, Kelly stakes as if it's +3%.
In reality, your estimate is noisy. The "true" edge might be +1%. Or +5%. Or 0%. You don't know.
If your real edge is half of what you thought, full Kelly is staking twice what it should. The drawdowns get worse, the growth gets worse, and the path becomes statistically much more likely to bankrupt you before the math saves you.
Why half-Kelly works
Here's the elegant result: half-Kelly captures roughly 75% of full-Kelly's long-run growth, while cutting volatility roughly in half.
That's not approximate; it's mathematical. Kelly's growth-rate curve is parabolic — it peaks at full Kelly and drops off slowly on either side. Staking half gives up only a quarter of the peak growth, but the variance scales linearly with stake size. The drawdown profile gets dramatically smoother.
In real terms: a half-Kelly bettor has the same expected wealth after a long sample as a full-Kelly bettor would have at about 75% of the time horizon — but they're far more likely to actually be there because they didn't quit during a 60% drawdown.
Quarter-Kelly and below
If your edge estimates are very uncertain (newer to the game, less proven model, fuzzier markets), quarter-Kelly is the conservative choice. You capture roughly 44% of full-Kelly's growth at roughly a quarter of the variance.
The principle: the more uncertain your edge, the smaller the fraction you should stake.
A rough scale:
- Full Kelly: only justified if you have a rock-solid edge estimate with low estimation error (high-volume arbing, certain promo conversions). Almost never the right call for prop, spread, or moneyline plays.
- Half-Kelly: the practical default for an experienced bettor with proven edge in their market.
- Quarter-Kelly: the right starting point for most new bettors, or for plays where your edge confidence is medium.
- Below quarter: for exploratory plays where you suspect edge but haven't proven it.
Sharp's recommended default
For most members, half-Kelly with the 5% cap. Your stakes will feel modest. They are. That's the point. The cap is the second safety net — Kelly's math can occasionally suggest big stakes on rare-edge plays; the 5% cap stops you from ever putting yourself in a position one bet can break.
After your first 200 logged bets with positive CLV, you can revisit the fraction. Most bettors who stay sharp end up lower, not higher, than where they started.
The discipline
The hardest part of fractional Kelly isn't the math. It's resisting the gut that wants to stake more on the "lock of the night." The lock isn't more locked because you feel strongly. It's the same edge it was when you computed it.
A Sharp bettor's discipline: the formula picks the stake size, not the feeling.