Key takeaways
- Fractional Kelly staking sizes bets as edge divided by odds, with quarter Kelly the standard because win probability estimates are noisy.
- Flat 1% staking is the safer default when your edge estimates come from an unaudited record rather than a model you can verify.
- Set drawdown triggers in advance: cut unit size 25% at a 15% drawdown and halve stakes at 25% before re-auditing your CLV record.
- A two-sheet Excel calculator with an inputs sheet and a bet log reprices every stake automatically as your running bankroll moves.
- Correlated same-game bets share an outcome, so size the basket as one position or your effective Kelly fraction silently doubles.
Bankroll management sports betting comes down to three numbers: your current bankroll, your edge, and the fraction of Kelly you stake. This guide gives you the exact Excel formulas for fractional Kelly unit sizing, drawdown-triggered bet reduction, and CLV-adjusted staking, so you can build the calculator yourself in under an hour.
What is bankroll management in sports betting?
Bankroll management in sports betting is a staking system for a dedicated pool of risk capital, separate from your checking account and sized so a 20-unit drawdown doesn't change how you live. The picks aren't the problem; the bet sizing is. Two bettors can take the same +EV plays and get opposite results if one stakes 5% per game and the other stakes a fraction of Kelly.
Kelly staking sizes bets as edge divided by odds, maximizing log growth rather than raw profit. The original formulation is John Kelly's 1956 paper, "A New Interpretation of Information Rate" (Bell System Technical Journal, via Princeton). Full Kelly assumes your edge estimates are exact, so most sharps stake a quarter or half of it.
If you want the shorthand version of these terms before building the calculator, the betting cheat sheet covers them.
How much of your bankroll should you bet per game?
On a $5,000 bankroll, flat 1-2% staking means $50-$100 per bet regardless of edge. Fractional Kelly scales stake to your estimated advantage. Both work; they answer different questions. Flat staking controls variance when your edge estimates are unreliable. Kelly maximizes growth when they are trustworthy.
Flat-percentage staking worked example
At 1% ($50), a 10-loss streak costs $500, a 10% drawdown. At 2% ($100), the same streak costs $1,000, or 20%. To rebuild a 20% drawdown you need to win 25% of your remaining bankroll, not 20%. That asymmetry is why most sharps cap flat stakes at 1%.
Fractional Kelly vs flat staking
Full Kelly on a 52.5% win probability at -110 is (0.525 - 0.475)/0.909 = 5.5% of bankroll, or $275 on $5,000. Most bettors run quarter Kelly, about $69 here, because win probability estimates are noisy. The comparison:
| Method | Stake on $5,000 | 10-loss drawdown |
|---|---|---|
| Flat 1% | $50 | $500 |
| Flat 2% | $100 | $1,000 |
| Full Kelly (5.5%) | $275 | $2,750 |
| Quarter Kelly (1.4%) | $69 | $690 |
Quarter Kelly behaves like flat 1-2% but shrinks automatically when your edge estimate drops. The trade-off is that it depends entirely on honest inputs. If you grade your own picks, or subscribe to a service that publishes auto-graded results and calibrated confidence scores you can audit (our approach), quarter Kelly is usable. If your edge estimates come from a capper with deleted losses, flat 1% is the safer default.
How does Kelly criterion staking work for sports betting?
Kelly formula for American odds
Kelly gives the fraction of bankroll to stake as f = (bp − q) / b, where b is the decimal profit per unit, p your win probability, and q = 1 − p. At -110 odds, b = 100/110 ≈ 0.909. The breakeven win rate is 110/210 ≈ 52.4%, and long-term profitability at that price requires beating it consistently; the margin between a real 53% and 54% bettor is the entire edge (Kelly criterion overview and breakeven math).
Worked example: you grade a bet at 54% against -110. f = (0.909 × 0.54 − 0.46) / 0.909 = 0.0311, so full Kelly stakes 3.1% of bankroll. On a $10,000 roll that is $311, not a flat unit.
Why bettors use quarter-Kelly
Full Kelly assumes your win probability is exact. Overestimate 54% when you are really 52.5% and full Kelly overbets into a negative edge. Variance also bites: full Kelly drawdowns of 40-50% are mathematically normal. Quarter-Kelly divides f by four, so the example stakes 0.78% ($78 on $10,000), trading growth rate for survivable variance. Most sharps run quarter or half Kelly for exactly this reason. If your edge estimates come from a picks feed rather than your own model, discount further; MVP Capping publishes CLV and auto-graded results you can audit before trusting any probability input.
How do you build a bankroll management calculator in Excel?
Two sheets do the job: an inputs sheet for sizing, a bet log for tracking. Total build time is about 20 minutes.
Sheet 1: inputs and unit sizing
Put these labels in column A and values in column B: Bankroll (B1), Kelly fraction (B2, start at 0.25), Minimum edge to bet (B3, e.g. 1.5%), and Unit base (B4). Unit base is simply =B1*0.01, so one unit reprices automatically as the bankroll moves. Every stake on Sheet 2 references these cells, which means a bad week shrinks your bets before you have to think about it.
Sheet 2: bet log with Kelly stake formula
Columns: Date, Sport, Book, Odds (American), Your win probability, No-vig fair probability, Edge, Stake, Result, P/L, Running bankroll. Edge is =(D2 win prob - fair prob) in decimal terms; compute it as =(E2-F2) where E is your probability and F is the no-vig implied probability from the book's odds converted with =IF(D2>0,100/(D2+100),-D2/(-D2+100)), then de-vigged against the other side's price. The Kelly stake in H2 is =MAX(0,(B1*0.25*(E2*G2-1)/(G2-1))) scaled down by your fraction, capped with MIN(...,3*B4) so a fat edge never triggers an oversized bet. Running bankroll in K2 is =K1+J2, dragged down the column.
Copy-paste formulas
- Unit base:
=B1*0.01 - Implied probability:
=IF(D2>0,100/(D2+100),-D2/(-D2+100)) - Fractional Kelly stake:
=MIN(3*$B$4,MAX(0,$B$1*$B$2*(E2*(1/F2)-1)/((1/F2)-1))) - Running bankroll:
=K1+J2
The log is also where CLV gets honest. When you audit a picks service, you want the same discipline from them: every pick graded automatically, losses left visible, and confidence scores checked against outcomes. Our graded picks feed publishes exactly that, so you can drop each recommendation into this sheet and verify the edge before paying for the next month. If a service's track record cannot survive your own spreadsheet, it does not deserve your bankroll.
How do you track drawdowns and adjust unit sizes?
Drawdown trigger thresholds
A drawdown is your peak-to-trough loss in units, and it is the fastest signal that your edge estimate is wrong or your variance run is worse than modeled. Set two thresholds: at 15% of bankroll, cut unit size by 25%; at 25%, halve stakes and re-audit your CLV record before betting again. If you are beating closing lines but still bleeding units, the problem is likely bet selection at your price, not staking. Remember the baseline: at standard -110 juice you need roughly 52.4% to breakeven, per Pinnacle's betting resources.
Automating unit cuts in the sheet
Track peak bankroll in one cell (=MAX(B:B) on a running balance column), then compute drawdown as =(peak-current)/peak. Your stake cell becomes =base_unit*IF((peak-current)/peak>=0.25,0.5,IF((peak-current)/peak>=0.15,0.75,1)), multiplied by your fractional Kelly output. This keeps the cut mechanical, so a bad week cannot talk you into oversized recovery bets. MVP Capping logs every pick with auto-graded results and no deleted losses, so your drawdown math runs on an auditable record; the same principles apply to any tracked ledger, as covered at mvpcapping.com/guides. Where it falls short: the sheet assumes your win-probability inputs are honest, and no calculator fixes misgraded edges.
When should line movement change your bet size?
CLV as a stake-sizing signal
Closing line value is an outcome measure, not a pre-bet input. If your Excel calculator grades a past bet as beating close by 1.5 cents, that confirms your edge was real, but the line has already moved, so the +EV you sized on is gone. Never resize the current bet because a similar pick closed well last week. Log CLV per pick, review it monthly, and let a sustained negative CLV trend cut your base Kelly fraction from, say, 0.5 to 0.25.
Steam and RLM: pass vs press
Steam that you catch live is usable information: if the market moves before you bet, your number is stale and the edge in your model no longer matches the price. Recalculate Kelly at the new line or pass. Reverse line movement is weaker evidence; sharp money can be hedging or book-specific. Treat RLM as a reason to re-check the no-vig price, not to press. The only stake increase that is defensible is a fresh edge at the current price, sized by the formula you already built.
What bankroll management mistakes do losing bettors make?
Four mistakes show up repeatedly: over-sizing after wins, chasing losses, stacking correlated same-game bets, and staking records you cannot audit. The first two are emotional, the second two are structural, and both cost you units.
Correlated bets that double your exposure
A $500 bet on the Chiefs moneyline and a $500 bet on Chiefs team total are not two $500 bets. They share an outcome, so your true exposure is closer to $1,000 on one view. If your Kelly stake assumes independent bets, correlated legs silently double your effective fraction and your drawdown when the correlated result misses. Size the correlation basket as one position, then split it.
Staking picks you cannot audit
Any pick service that deletes losses or hides a full graded record makes your Kelly inputs guesswork. Before staking someone else's edge, demand auto-graded results and closing line value per pick. Problem gambling research shows chasing losses accelerates when bettors lack clear feedback on results (National Council on Problem Gambling). If a service will not show you its losses, treat the record as zero and stake accordingly.
Which tools help with bankroll management for sports betting?
Comparison criteria
Judge any picks service on three things: does it publish CLV on every graded pick, are confidence scores calibrated against real outcomes, and can you audit the full record before paying. Action Network offers odds and tracking but is adjacent, not a sharp feed. Direct options include EdgeAI Sports, Sharp App, The Over, and Sharp Edge; evaluate each against your staking plan, not their marketing.
Where MVP Capping falls short
Honest accounting: our picks are auto-graded with CLV tracked on every play and no deleted losses, but we do not ship a built-in Kelly calculator or drawdown dashboard, so you will run the Excel sheet from earlier sections alongside the feed. The Edge tier runs $29/mo, aimed at bettors staking $500+ per game who want the raw numbers behind each unit recommendation. If you bet internationally or under $500 bankroll, the US-heavy board limits value.
FAQ
Is a 1% or 2% unit size better for sports betting?
1% is the safer default. On a $5,000 bankroll, a 10-loss streak costs $500 at 1% versus $1,000 at 2%, and rebuilding a 20% drawdown requires winning back 25% of what remains. If your edge estimates are unreliable or come from an unaudited record, stay at 1%; fractional Kelly only makes sense when your inputs are honest.
What is a good starting bankroll for sports betting?
Size it as dedicated risk capital that a 20-unit drawdown cannot feel in your daily life. Practically, that means an amount where a 1% unit ($50 on $5,000, for example) still produces meaningful stakes after book limits and juice. Below a few thousand dollars, minimum bet sizes inflate your effective fraction, so flat 1% staking beats Kelly until the roll grows.
Should I recalculate my unit size after every bet?
Yes, if you use fractional Kelly, because the stake formula references current bankroll, so it reprices automatically as the running balance moves. If you stake flat units, recompute the unit base weekly or monthly instead. Never resize mid-session based on emotion; let the sheet's formula, capped at three base units, make the decision mechanically.
How many units down before I should take a break?
Set thresholds in advance: at a 15% drawdown, cut unit size by 25%; at 25%, halve stakes and re-audit your closing line value record before betting again. On a 1% unit basis, 25% of bankroll equals 25 units. The point of the break is mechanical, not motivational: it stops a bad run from triggering oversized recovery bets.
Does bankroll management work if I only bet parlays?
It works, but the inputs get harder. Parlay edges compound the margin on every leg, so your win probability estimates must be accurate per leg before the Kelly formula means anything. Correlated legs within one slate silently multiply exposure, so size the basket as a single position. If you cannot compute an honest edge per leg, stake flat 1% and skip Kelly.