MVP — Manny's Variety Picks
Bettor comparing two possible stake amounts

Flat Betting vs Kelly: Which Staking Plan Fits You?

Bettor comparing two possible stake amounts

Flat betting risks the same amount on every wager. The Kelly criterion risks a share of your bankroll that grows with your estimated edge and shrinks when the edge is thin. Flat staking is simpler and more forgiving of bad estimates. Kelly grows a bankroll faster in theory, but only if your probabilities are accurate, and in practice most bettors should use a fraction of it. This guide shows the math, a worked example, and a way to choose.

Mannysvariety
Size Bets With Better Insight
Manny's Variety uses sport-specific AI engines to generate data-driven picks for NBA, MLB, and NFL bettors.
Explore Manny's Variety

TL;DR

  • Flat betting: the same stake every bet, usually 1% to 3% of your starting bankroll. It is easy to run and hard to wreck with one bad estimate.
  • Kelly: stake size scales with your edge and the odds. It maximizes long-run growth, but it assumes you know your true win probability, which no bettor does.
  • Fractional Kelly (half or quarter) gives up some growth in exchange for much smaller swings and protection against overestimating your edge.
  • Pick by your inputs: if you do not track a calibrated edge on each bet, use flat units. If you do, use a fractional Kelly with a hard cap.

Table of Contents

What flat betting is and how it works

Flat betting, also called level staking, means every wager is the same size. You define a unit, then bet one unit on everything, or a fixed number of units on a small set of grades (for example, 1 unit on standard plays and 0.5 on lower-confidence ones).

A common starting point is a unit of 1% to 3% of your bankroll. With a $1,000 bankroll, that is $10 to $30 per bet. Most flat bettors keep the unit fixed for a set period, such as a month or a season, and then reset it to the new bankroll. Resetting on a schedule, rather than after every bet, is what separates flat staking from percentage staking.

Why bettors like it

  • Nothing to calculate. You never need a probability estimate to size a bet.
  • Clean records. Results in units are easy to compare across weeks, sports and bet types. Our guide to measuring sports betting ROI with a spreadsheet builds on exactly this kind of flat-unit log.
  • Error tolerance. If your read on a game is wrong by a few points of probability, the damage is the same small, fixed stake.

Where it falls short

Flat staking treats a bet with a thin edge the same as a bet with a large one. It also does not scale on its own. If your bankroll doubles and you keep the same dollar stake, you are under-betting relative to your capital. If it halves, you are over-betting. That is why a periodic reset matters.

What the Kelly criterion is

The Kelly criterion is a formula for sizing a sequence of bets by maximizing the long-term expected value of the logarithm of wealth. In plain terms: it picks the stake that makes your bankroll grow fastest over many bets, assuming your probabilities are right.

For a simple win-or-lose bet, the formula is:

f = (b × p − q) ÷ b*

  • f* is the fraction of your bankroll to stake
  • b is the net payout per dollar risked (decimal odds minus 1)
  • p is your estimated probability of winning
  • q is 1 minus p

If the result is zero or negative, the bet has no edge by your numbers, and Kelly says to stake nothing.

A worked example

Say you are offered a spread at -110. A $110 stake wins $100, so b is about 0.909. The break-even win rate at -110 is 52.38%, which you can verify with our walkthrough on how to convert odds to probability.

Now suppose your model says the true chance is 55%:

  • b × p = 0.909 × 0.55 = 0.500
  • q = 0.45
  • f* = (0.500 − 0.45) ÷ 0.909 = about 0.055

Full Kelly says to stake about 5.5% of your bankroll. On $1,000 that is $55. Half Kelly is about 2.75%, or $27.50.

Now change the estimate to 53%. The edge nearly disappears, and f* falls to roughly 0.6%, or about $6. A flat bettor would still stake the same $20 or so on both. That is the core trade: Kelly responds to edge, flat does not.

A worked example — overview diagram

Why edge must be real

Kelly is only as good as the probability you feed it. Sportsbooks build a margin into every line, so implied probabilities across all outcomes add up to more than 100%. Sports betting textbooks point out that a real edge exists only where the book’s implied probability undervalues the actual chance after that margin is accounted for. Our explainer on expected value betting shows how to calculate that gap before you size anything.

Flat betting vs Kelly side by side

Factor Flat betting Full Kelly Fractional Kelly (¼ to ½)
Stake size Fixed unit Varies with edge and odds Varies, scaled down
Needs a probability estimate No Yes, for every bet Yes, for every bet
Growth potential Moderate Highest in theory Lower than full, still strong
Bankroll swings Smallest Largest Moderate
Sensitivity to a wrong estimate Low Very high Moderate
Effort Minimal High High
Best for Beginners, anyone without a calibrated model Rarely used in full Bettors with a tracked, tested edge

Why full Kelly is risky in practice

The math behind Kelly assumes you know the true probability. Real bettors never do. E.O. Thorp, who pioneered applying Kelly to real gambling, is cited as arguing that the main risk lies in the human tendency to overestimate one’s advantage. When your estimate is too high, Kelly pushes stakes too high, and drawdowns arrive quickly.

Simulations back this up. One analysis of bet sizing under uncertainty shows why reducing the Kelly fraction helps when your probability is only an estimate. Pinnacle’s write-up on applying fractional Kelly to sports betting notes the common complaint that Kelly makes bankroll growth erratic, with profits interrupted by sometimes significant losses.

Two more practical problems:

  • Correlated bets. Kelly in its basic form assumes independent bets. Two wagers on the same game, or a parlay with related legs, are not independent. Research on Kelly for non-mutually exclusive bets exists because the simple formula does not cover these cases.
  • Psychology. A full Kelly bettor can see large percentage drops in a bankroll during a normal losing run. Many people abandon the plan at the worst moment.

Fractional Kelly: the practical middle

Fractional Kelly means betting a set share of the Kelly stake: half, a third or a quarter. You compute f* as usual, then multiply.

Why it works:

  • It cushions overestimates. If your 55% is really 53%, a half-Kelly stake is much less harmful than a full one.
  • It smooths the ride. Stakes are smaller, so losing streaks cost less.
  • It keeps edge sensitivity. Bigger edges still get bigger stakes.

A workable setup for a bettor with a tracked edge:

  1. Estimate your win probability from a model or a tested process.
  2. Compute full Kelly with the formula above.
  3. Multiply by 0.25 to 0.5.
  4. Cap any single bet at a fixed share of bankroll, such as 2% to 3%.
  5. Recalculate against your current bankroll on a schedule, not after every result.

The cap matters. Even a quarter-Kelly stake can look large on a bet where you feel very sure. Feeling sure is exactly when estimates drift high.

How to choose: a decision guide

Ask these four questions in order.

1. Do you have a probability for every bet?

If you pick games on feel, trends or a tout’s say-so, you do not have a probability, and Kelly has nothing to work with. Use flat units.

2. Is your edge tested over enough bets?

A hot month proves little. Our piece on how many bets make a reliable sample explains why a few dozen results can mislead. If your tracked sample is small, stay flat until it grows.

3. Can you stomach the swings?

If a 20% bankroll dip would make you chase losses or quit, choose flat staking or quarter Kelly with a cap. A plan you abandon is worse than a slower plan you keep.

4. Are your bets independent?

If you often bet several markets in one game, or stack parlays, reduce your stakes further or treat the combined exposure as one bet.

Quick profiles

  • New bettor, no model: flat 1% unit, reset monthly.
  • Intermediate bettor, logging results: flat units with a half-unit tier for lower-confidence plays.
  • Bettor with a calibrated model and 500+ tracked bets: quarter to half Kelly, 2% to 3% cap.

For the broader framework around any of these, see our guide to bankroll management for betting.

Our perspective

We see the same pattern in how bettors use picks. The staking debate gets most of the attention, yet the input behind it decides the outcome. Kelly rewards accuracy and punishes optimism, so a staking plan is only as honest as the record behind it.

That is why we publish every pick and grade it in public. Our tracked record stands at a 63.5% win rate across more than 1,600 picks, with a net return of 443.9 units. Those figures are in units, because units are what a flat bettor can apply directly and what a Kelly bettor can compare against their own estimates. Past results do not guarantee future ones, and a record is not a substitute for your own bankroll limits.

Our perspective — overview diagram

Our recommendation is simple. Start flat. Log every bet. Move to a fractional Kelly only after you have enough tracked results to trust your probabilities, and keep a hard cap even then. A modest stake you can repeat for a full season beats an aggressive one you abandon in week three.

We also lean on evidence-based betting for the same reason: sizing should follow measurement, never the other way around.

— Manuel

Putting a plan into practice

Here is a concrete starting framework for a $1,000 bankroll.

  1. Set the unit. 1% is $10. Write it down.
  2. Define tiers. Standard play, 1 unit. Lower-confidence play, 0.5 units. No bet larger than 2 units.
  3. Log everything. Date, sport, market, odds, stake, result, units won or lost.
  4. Review monthly. After 100 bets, check your results against the break-even rate for the odds you took.
  5. Reset the unit only on a calendar date.
  6. Test Kelly on paper. Calculate what a quarter-Kelly stake would have been on each logged bet and compare the paper results with your flat results. Switch only if the paper results hold up across a large sample.

Taxes and records

Staking records do double duty. In the U.S., gambling winnings are generally taxable, and Form W-2G covers reportable payouts. The IRS instructions tie loss deductions to your winnings and require accurate records, so a clean bet log helps at tax time too. Talk to a tax professional about your own situation.

A note on responsible play

Odds are built to favor the book, and betting is best treated as entertainment rather than income. Responsible gambling guidance recommends setting a budget for each session, sticking to it and avoiding chasing losses. Neither flat betting nor Kelly removes that risk. If gambling stops being fun or you are betting money you cannot afford to lose, call 1-800-GAMBLER, the National Problem Gambling Helpline.

Where Manny’s Variety fits

Staking works best when the picks behind it are graded openly. Manny’s Variety Picks gives you daily AI-driven picks, props and parlays from sport-specific engines, with a permanent archive of every pick so you can check the record yourself. You can see how the service works on the how it works page.

Plans, as published on the pricing page:

  • Free: $0, with limited features
  • Core: $39.99 per month
  • Elite: $79.99 per month

Pair any of them with the flat-unit plan above, log your results, and judge the picks by the numbers.

FAQ

Is flat betting better than Kelly?

For most bettors, yes. Flat betting needs no probability estimate and tolerates errors in judgment. Kelly can grow a bankroll faster in theory, but only when your probabilities are accurate, which is hard to prove without hundreds of tracked bets.

What is fractional Kelly?

Fractional Kelly means staking a set portion of the full Kelly amount, such as half or a quarter. It trades some growth for smaller drawdowns and a buffer against overestimating your edge. Many bettors who use Kelly choose a fraction for that reason.

What percentage of my bankroll should I bet?

Flat bettors commonly use 1% to 3% per bet. A Kelly bettor’s percentage changes with each edge, so a cap of 2% to 3% on any single wager is a sensible guardrail. Smaller is safer if you are new or your sample is small.

Can I use Kelly on parlays?

It is risky. Parlay legs are often correlated, and the basic Kelly formula assumes independent bets. If you do use it, stake less than the formula suggests and treat all exposure in one game as a single position.

How do I know my edge is real?

Track every bet and compare your results with the break-even rate for the odds you took. You need a large sample before the result means much. Until then, treat any estimated edge as provisional and stay with flat units.

Sources