Quarter-Kelly Staking Explained: How Much to Bet on a +EV Pick
Finding a real +EV bet is the hard part. Most bettors treat sizing it as an afterthought – bet a “unit,” bet what feels right, bet more because you’re confident. That’s a mistake, and it’s a bigger one than picking a slightly worse line. An edge you size wrong can lose money just as reliably as no edge at all.
There’s a formula for this, and it’s been around since the 1950s: the Kelly criterion. It tells you the mathematically optimal fraction of your bankroll to risk on a bet, given your edge and the odds. Almost nobody should use the full number it produces. This post covers why, and why quarter-Kelly is what sharp bettors actually use instead.
The short answer
The Kelly criterion calculates the bet size that maximizes the long-run growth rate of your bankroll. The formula is:
f* = (bp – q) / b
Where f* is the fraction of your bankroll to bet, b is the decimal odds minus 1 (your net payout per dollar risked), p is your true win probability, and q is 1 – p (your true loss probability).
Full Kelly is mathematically correct but practically dangerous, because it assumes you know your true edge with certainty. You don’t – you have an estimate. Quarter-Kelly means betting 25% of what the full formula tells you, and it’s the standard sizing approach among professional bettors for one reason: it cuts bankroll volatility dramatically while giving up only a small fraction of long-run growth.
Working the formula
Say you’ve identified a moneyline bet at +130 (decimal odds 2.30, so b = 1.30) that you believe is genuinely +EV – your model or your Kalshi-anchored fair value puts the true win probability at 46%, versus the market’s implied 43.5%.
p = 0.46, q = 0.54, b = 1.30
f = (1.30 x 0.46 – 0.54) / 1.30
f = (0.598 – 0.54) / 1.30
f = 0.058 / 1.30
f = 0.0446, or about 4.5% of your bankroll
Full Kelly says bet 4.5% of your bankroll on this single pick. On a $2,000 bankroll, that’s a $90 bet. Quarter-Kelly takes that same 4.5% and divides it by four – about 1.1% of bankroll, or roughly $22.50 on the same $2,000 bankroll.
That gap – $90 versus $22.50 on the exact same bet – is the whole argument for fractional Kelly, and it gets worse the more confident you get. The tighter your believed edge, the bigger the full-Kelly number gets, and the more full Kelly punishes you when that edge turns out to be smaller than you thought.
Why not just bet full Kelly
Full Kelly is only optimal if your probability estimate is exact. In sports betting, it never is. Every “true win probability” is a model output or a fair-value estimate, not a fact, and even a small error compounds badly at full Kelly stakes.
Here’s the part that surprises most bettors: overestimating your edge doesn’t just shrink your growth rate at full Kelly, it can flip your expected growth negative even when a real edge exists. If you think your edge is 4.5% and it’s actually 2%, full Kelly sizing based on the inflated estimate can produce a bankroll trajectory that trends toward zero over time, using bet sizes that would have been fine at the correct, smaller edge.
Full Kelly is also brutal to sit through even when your estimate is correct. The formula that maximizes long-run growth also produces enormous short-term swings – drawdowns of 50% or more are a normal, expected part of a full-Kelly staking plan, not a sign that something’s broken. Almost nobody has the bankroll discipline (or the stomach) to keep betting full Kelly-sized stakes through a drawdown like that, which means in practice, almost nobody who claims to bet full Kelly is actually doing it consistently.
The math behind quarter-Kelly
Kelly growth rate is a concave function of bet size – it rises fast as you scale up from zero, peaks at the full-Kelly fraction, then falls off sharply past that peak. The key fact that makes fractional Kelly work: the curve is much flatter on the way up to the peak than it is coming down on the other side.
Betting a quarter of full Kelly gives up a relatively small amount of theoretical long-run growth compared to full Kelly – commonly cited estimates put it in the range of losing 25-35% of the maximum theoretical growth rate. In exchange, it cuts bankroll variance dramatically, often by 75% or more, and it builds in a real margin of safety against the fact that your probability estimate is not perfect. That trade – give up a modest amount of theoretical upside, keep most of the growth, remove most of the risk of ruin from a bad estimate – is why quarter-Kelly (and half-Kelly, for bettors with tighter, better-tested edges) is the practical standard instead of full Kelly.
A simple staking process
Step 1 – Get your true win probability. This has to come from somewhere real: a model, a fair-value benchmark like Kalshi pricing, or a documented process with a track record. Guessing your edge and plugging it into Kelly just moves the guessing earlier in the process.
Step 2 – Calculate full Kelly using the formula above.
Step 3 – Divide by four. That’s your bet size as a fraction of current bankroll.
Step 4 – Recalculate from your current bankroll, not your starting bankroll, every time. Kelly staking is proportional by design – stakes shrink automatically during a downswing and grow automatically during an upswing, which is part of what protects you from ruin.
Step 5 – Cap it. Most professional bettors set a hard ceiling – commonly 2-3% of bankroll per bet – regardless of what the formula says, specifically to protect against a bad probability estimate on any single pick.
Common sizing mistakes
Flat betting the same dollar amount regardless of edge size treats a 1% edge and a 6% edge identically, which throws away the entire point of proportional staking. Betting a fixed percent of bankroll without adjusting for edge size makes the same error in a different form. And sizing up after a string of wins – “I’m up, I can afford to bet bigger” – is a bankroll-management mistake dressed up as confidence; your edge on the next bet doesn’t change because your last few bets won.
How Automatehive Edge fits into this
Kelly sizing is only as good as the win-probability input, which is why the fair-value estimate behind it matters more than the formula itself. Edge anchors its fair value to Kalshi prediction market pricing rather than a sportsbook consensus, because Kalshi is a two-sided exchange with structurally lower vig – a cleaner benchmark for estimating true probability than a sportsbook line built to balance action. Every alert Edge sends comes with an implied probability and a recorded price, publicly tracked against the closing line, so the input you’d plug into a Kelly calculation isn’t a private number you have to take on faith.
The takeaway
The Kelly criterion tells you the bet size that’s mathematically optimal if your edge estimate is exactly right. It never is, which is why quarter-Kelly – not full Kelly – is what disciplined bettors actually use. It gives up a modest slice of theoretical growth in exchange for a bankroll that can survive the fact that every edge estimate carries some error. Find the edge first. Size it like you’re not completely sure about it, because you’re not.
See it in practice: Automatehive Edge posts every +EV pick publicly with an implied probability and a Kalshi-anchored fair value attached, plus an unfakeable CLV record graded against the closing line. See the live track record.
Not betting advice. Bettors must be 21+. Bet responsibly — only wager what you can afford to lose. Gambling problem? Call 1-800-GAMBLER.