Flat Betting vs. Kelly Criterion: What Casual Bettors Should Actually Use

Every serious discussion of sports betting bankroll management eventually gets to two approaches: flat betting and the Kelly Criterion. They represent opposite ends of a spectrum — simplicity versus optimization — and the choice between them has real consequences for how your bankroll behaves over time.

Most bettors have heard of Kelly. Fewer understand what it actually requires to implement correctly. Fewer still understand why a theoretically suboptimal approach might produce better outcomes for them in practice.

Here's the full picture.

Flat Betting: The Baseline

Flat betting means wagering the same number of units on every bet, regardless of how confident you feel. If your standard unit is 1% of your bankroll, every bet is 1% — no exceptions.

The appeal is obvious. It's simple. It's emotionally neutral. It protects you from the cascading effects of overconfidence: the tendency to bet bigger on picks you "know are going to hit," which in practice means betting bigger on the picks where your analytical process breaks down most dramatically.

Flat betting also makes your record legible. If you bet 1 unit flat on everything and you're up 12 units after 200 bets, you know exactly what that means. You're profitable. If your win rate and unit sizing fluctuate, tracking true profitability becomes much harder.

The drawback is that flat betting treats all bets as equally valuable, which ignores the fact that some edges are genuinely larger than others. A pick where multiple signals converge strongly probably deserves more action than a thin-edge spot. Flat betting leaves that upside on the table.

The Kelly Criterion: The Theory

The Kelly Criterion is a formula developed by Bell Labs researcher John Kelly in 1956. It calculates the optimal fraction of your bankroll to wager on a bet to maximize long-run growth, given a known edge.

The formula is:

f = (bp - q) / b

Where: - f is the fraction of your bankroll to wager - b is the net odds received on the wager (so -110 gives you b = 10/11, or 0.909) - p is your estimated probability of winning - q is 1 - p (the probability of losing)

If you believe you have a 56% probability of winning a -110 bet, Kelly says bet approximately 8.5% of your bankroll on that game. If you have a 54% probability, Kelly says about 4.5%. If your estimated edge is 0%, Kelly says bet nothing.

The math is sound. Kelly betting maximizes the expected logarithm of bankroll growth, which means it produces the highest geometric mean return over many bets given accurate inputs. This is provably optimal in a theoretical sense.

The Problems Kelly Creates in Practice

**The inputs have to be accurate.** Kelly is only optimal when your probability estimates are correct. In sports betting, you are never certain your probability estimate is correct. You have a model, a signal framework, and a reasonable conviction — but you do not have exact probabilities. When you overestimate your edge (which most bettors do most of the time), Kelly tells you to bet too much. The further off your estimate, the worse the overbetting problem becomes.

**Kelly is brutally volatile.** Full Kelly produces enormous bankroll swings even when everything is working correctly. A series of losing bets — entirely normal in any betting process — produces rapid drawdowns because Kelly sizes up aggressively on perceived edges. Many professional bettors who have done the math on Kelly choose not to use it because they can't stomach the variance, or because the drawdowns would force them to reduce their bet size at exactly the wrong moment.

**Half Kelly is the professional standard, not full Kelly.** Most sharp bettors who use any form of Kelly use "half Kelly" or "quarter Kelly" — intentionally betting half or a quarter of what full Kelly recommends. This reduces variance significantly while preserving most of the long-term growth advantage. At half Kelly, you're still responsive to edge size, but you're not one bad week away from a 40% bankroll drawdown.

**Overconfidence compounds.** The bettor who genuinely has a 5% edge but estimates a 15% edge and bets full Kelly on that estimate will blow up their bankroll. Not might. Will. The bias doesn't have to be dramatic to be ruinous. A modest, consistent tendency to overestimate your edge — which is the default human state — makes full Kelly dangerous.

What Most Bettors Should Actually Do

For casual bettors, the answer is almost always some form of tiered flat betting. The logic:

Maintain a base unit size (1-2% of your bankroll is typical). For standard picks where you have reasonable confidence, bet one unit. For picks where signal alignment is particularly strong, allow yourself to go up to two or three units maximum. Never go above three units regardless of confidence level, because the confidence level above three units is usually overconfidence, not genuinely superior information.

This captures some of the benefit of variable staking — bigger bets when the edge is more convincing — without the catastrophic failure modes of full Kelly. It's also honest about the fundamental problem: your probability estimates are imperfect, and your bet sizing should not scale aggressively on imperfect inputs.

SharpSpots publishes picks with a one-to-five-star confidence rating. Our track record assumes flat betting (one unit per pick) for accounting clarity, but the star rating is a reasonable proxy for the signal strength that would justify a modest size increase in a tiered system.

A Word on Bankroll Size

Whatever staking system you use, the base unit size matters as much as the system itself. The standard guidance — which exists for good reason — is to treat your sports betting bankroll as money you are genuinely prepared to lose entirely without financial consequence. Not money you expect to lose, but money whose loss would not affect your life.

This isn't moralizing. It's mathematical. If the money you're betting affects decisions the way loss aversion does — making you chase, bet too large to recover, or abandon a working process after a rough week — then the bankroll is too large for your risk tolerance. The system you use becomes irrelevant when the emotional stakes are high enough to override your analytical process.

Set a bankroll that lets you execute your system without emotional interference, and then execute it for hundreds of bets before drawing conclusions.

The Bottom Line

Full Kelly is theoretically optimal and practically treacherous. Flat betting is theoretically suboptimal and practically robust. For most bettors, a tiered flat approach — one to three units per pick, with clear criteria for each tier — offers the best balance of edge-responsiveness and emotional manageability.

The research on Kelly suggests that the biggest risk isn't using the wrong formula. It's overestimating your edge when using any formula. Build that correction in from the start, and the specific staking method matters a lot less than you might expect.