Your edge dies from sizing mistakes, not bad picks
You can be right about games and still go broke. That’s not a motivational poster. It’s math.
Most bettors obsess over “who covers” and barely think about “how much.” Then they wonder why a month of good reads turns into a flat bankroll. Or worse, why one ugly Saturday wipes them out.
Stake sizing is where recreational bettors get crushed because it feels boring. It’s not. It’s the difference between compounding an edge and donating it.
Three approaches cover 95% of what you’ll ever need:
- Flat staking: same bet size every time (or same “unit”). Simple, sturdy, and hard to screw up.
- Kelly Criterion: bet a % of bankroll based on your edge and the odds. Mathematically optimal for long-run growth, psychologically brutal.
- Fractional Kelly / hybrids: Kelly’s brain with flat staking’s seatbelt.
This guide shows you the math, then turns it into rules you can actually follow—especially when you’re down, tilted, or staring at a “can’t miss” -110.
If you’re fuzzy on terms like EV and edge, fix that first. This post assumes you can speak the language: CLV, EV, ROI: 9 Betting Terms People Keep Butchering.
Flat staking: the “I won’t blow up” default
Flat staking means you bet the same amount every time. Usually it’s a “unit,” like 1% of your bankroll, or a fixed dollar amount you refuse to change.
Example: bankroll = $2,000. You choose 1 unit = 1% = $20. Every wager is $20. Bears -110, $20. Yankees +130, $20. Total 47.5 alt line at -105, $20. No drama.
Why it works:
- It’s hard to overbet. Most bankroll deaths come from a few oversized bets, not a thousand small ones.
- It fits real-world constraints. Limits, slow withdrawals, and uneven bet volume don’t wreck the plan.
- It protects you from your own confidence. Your “10/10 lock” is usually just you being emotional.
Why it’s not “optimal”:
Flat staking ignores edge. If you truly have a bigger edge on one bet than another, you’re underbetting the best spots and overbetting the marginal ones. That costs growth.
But don’t confuse “not optimal” with “bad.” Flat staking is often the best choice if:
- You don’t trust your edge estimates (you shouldn’t, at first).
- You bet across multiple sports/markets with uneven model quality.
- You struggle with tilt, chasing, or changing stakes mid-week.
One strong tweak: flat staking with periodic recalibration. Keep $20 bets for the month, then if your bankroll becomes $2,400, bump unit size to 1% = $24 next month. That keeps your risk roughly proportional without emotional resizing every day.
Kelly Criterion: the math that scares people for a reason
Kelly tells you how much of your bankroll to bet when you have an edge. It maximizes long-run logarithmic growth. Translation: if your edge estimates are perfect and you can survive the swings, Kelly prints.
Here’s the core formula for a single bet:
f* = (b·p − q) / b
- f* = fraction of bankroll to bet
- b = net odds (decimal odds − 1)
- p = your win probability
- q = 1 − p
Example 1: -110 spread
-110 implies decimal odds 1.909. So b = 0.909.
Say your number makes the bet a 55% win probability (p = 0.55, q = 0.45).
Compute it:
f* = (0.909 × 0.55 − 0.45) / 0.909
0.909 × 0.55 = 0.49995
0.49995 − 0.45 = 0.04995
0.04995 / 0.909 ≈ 0.055
Kelly says bet 5.5% of bankroll. On a $2,000 roll, that’s $110.
Example 2: +150 underdog
+150 is decimal 2.50, so b = 1.50.
Say you make the dog 45% to win (p = 0.45, q = 0.55).
f* = (1.50 × 0.45 − 0.55) / 1.50
1.50 × 0.45 = 0.675
0.675 − 0.55 = 0.125
0.125 / 1.50 = 0.0833
Kelly says bet 8.33%. That’s $166 on a $2,000 roll.
That number probably made your stomach turn. Good. That’s the catch: full Kelly is aggressive. It assumes your p is correct. In betting, your p is never that clean.
Fractional Kelly: same math, fewer bankroll funerals
Fractional Kelly means you take the Kelly fraction and cut it down—most commonly half Kelly (0.5×) or quarter Kelly (0.25×).
Using the -110 example where full Kelly said 5.5%:
- Half Kelly = 2.75% of bankroll
- Quarter Kelly = 1.375% of bankroll
On a $2,000 bankroll:
- Full Kelly: $110
- Half Kelly: $55
- Quarter Kelly: $27.50
Why fractional Kelly makes sense in sports betting:
- Your edge estimate is noisy. Even good bettors misprice games. Full Kelly punishes overconfidence brutally.
- Lines move and your bet quality varies. You’re not always getting the same EV even if you think you are.
- Limits and liquidity exist. Kelly wants to scale up fast as bankroll grows; books don’t always let you.
Here’s the important part most people miss: fractional Kelly doesn’t just reduce risk linearly. It reduces drawdowns a lot more than it reduces growth. That trade is usually worth it because staying alive matters more than theoretical optimality.
Also, fractional Kelly fixes a human problem: you’ll actually follow it. A plan you follow at 80% efficiency beats a perfect plan you abandon the first time you hit a 12-bet losing streak.
If you want a practical default for someone who’s profitable but not running a PhD-grade model: quarter Kelly is a damn good place to live.
Hybrid staking rules that work in the real world
If you’re betting for years, you end up with a hybrid. Pure systems look clean on paper and messy in your account history.
Here are three hybrid rules I’ve seen actually hold up when the variance hits.
1) Flat units + edge tiers
You keep a base unit, then scale modestly based on how strong the edge is.
- Small edge: 1 unit
- Medium edge: 1.5 units
- Large edge: 2 units
Example: bankroll $2,000, unit $20.
- 1 unit = $20
- 1.5 units = $30
- 2 units = $40
This approach avoids the “Kelly wants me to bet $166” problem while still rewarding your best spots.
2) Fractional Kelly with hard caps
Run quarter Kelly, but never exceed:
- Max bet: 2% of bankroll on any single wager
- Daily exposure: 6% of bankroll total
Those caps save you when your inputs lie to you (they will), and when you get correlated exposure without noticing (same team, same injury, same weather, same market).
3) Kelly for “true probabilities,” flat for everything else
If you have a market where you genuinely model win probability well (say, moneylines in a niche league you track obsessively), use fractional Kelly there.
For markets where you’re basically price-shopping and reading the screen (totals, props, derivative alt lines), go flat. You’re still getting value, but your confidence in p is lower, so you size conservatively.
If you want help turning EV into a repeatable staking plan with guardrails, Betting Assistant does exactly that—max bet rules, stop-loss notes, and a consistent framework so you don’t freestyle your way into overbetting.
Real numbers: what happens to a bankroll with different sizing
Let’s use one simple, realistic setup: you bet -110 sides. Your true win rate is 54% (that’s strong—most people aren’t there). You place 500 bets.
At -110, your profit per win is 0.909 units, and your loss per loss is 1 unit.
Expected profit per bet in units:
EV = p×0.909 − (1−p)×1
EV = 0.54×0.909 − 0.46×1
0.54×0.909 = 0.49086
EV = 0.49086 − 0.46 = 0.03086 units
That’s about 3.1% of your unit per bet.
If you flat bet 1 unit each time, expected profit over 500 bets:
500 × 0.03086 = 15.43 units
If your unit is 1% of bankroll (and you keep it roughly updated), that’s meaningful growth. Not sexy. Real.
Now the dangerous part: variance. Over 500 bets at 54%, your “normal” swings include ugly stretches. A 10-bet losing streak happens more often than your brain wants to admit.
With flat 1% stakes, a 10-loss streak is roughly a 10% drawdown (ignoring small compounding effects). That stings, but you’re alive.
With full Kelly (and using a typical edge estimate that implies 5%+ stakes), that same streak can be a 40–50% drawdown. That’s where you start changing your process, skipping plays, doubling to “get even,” and doing all the stuff that kills you.
Kelly maximizes growth if you keep betting through the drawdowns exactly as prescribed. Most humans don’t. That’s why half/quarter Kelly or a capped hybrid often outperforms “full Kelly” in the real world—because you actually survive long enough to realize the edge.
Common staking mistakes that torch good bettors
You can have a real edge and still screw yourself with these. I’ve done some of them. You’ve probably done the rest.
- Betting bigger because you’re “due.” The market doesn’t care about your last five losses. Your edge doesn’t increase because you’re annoyed.
- Changing unit size mid-downturn. Cutting stakes after losses locks in the pain and slows recovery. Pick a recalibration schedule (weekly/monthly) and stick to it.
- Using Kelly with made-up probabilities. If your “p” is just vibes, Kelly becomes a weapon pointed at your bankroll. Kelly is unforgiving to overestimation.
- Ignoring correlation. Three bets can be one bet wearing different hats. Team total over, QB passing yards over, and game over? That’s basically one script. Size accordingly.
- No caps, no limits, no plan. If your staking system doesn’t include max bet and max daily exposure, you’re one heater away from thinking you’re invincible—and one cold snap away from panic.
- Chasing CLV with reckless sizing. Closing line value matters, but you still need a bankroll plan. A great number with suicidal sizing still ends badly.
Also: parlays. Most parlays are sucker bets, and they wreck bankroll management because they hide stake size behind “small risk.” A $25 parlay at +900 isn’t “only $25” if you fire 12 of them a weekend. If you insist on building parlays, at least learn to avoid correlated traps: Parlay Builder: Catch Correlated Legs Before You Donate (2026).
If you want more strategy guides like this, browse /blogs/strategy/.
Responsible gambling note: Bet sizes should never put rent money in play. If staking discipline feels impossible, take a break and lower limits before you lower standards.
