Why Most Bankroll Management Advice Gets You Broke
I blew through $3,200 in three months betting 5% flat stakes on what I thought were “value plays” before I understood the math behind bankroll management for sports bettors. Everyone tells you to bet 1-5% of your roll, but nobody explains that your edge size determines everything. A sharp with 8% edge betting 5% units is printing money. A square with 2% edge betting 5% units is heading toward ruin at terminal velocity. I tracked 847 bets across two seasons and learned this the hardest way possible: your edge determines your bet size, not some random percentage someone threw out in a forum thread.
The problem is that most bettors have no idea what their actual edge is. They see a line they like, convince themselves they have an advantage, and fire away with bet sizing pulled from thin air. I did this for months. Lost $1,900 before I even started tracking properly. Then I spent another four months documenting every bet, every closing line value movement, every result. The data told me something brutal: I had been risking 4-5% of my bankroll on edges that were probably closer to 1-2% when they existed at all.
The Pure Math Behind Kelly Criterion and Edge-Based Sizing
Here is the calculation that changed how I bet. The Kelly Criterion formula for sports betting is: f* = (bp – q) / b, where f* is the fraction of your bankroll to wager, b is the decimal odds minus 1, p is your probability of winning, and q is the probability of losing. Sounds academic, but watch what happens with actual numbers.
Say you find a bet at +150 (2.50 decimal). You estimate your true probability of winning at 45%. The sportsbook’s implied probability is 40% (removing vig). Your edge is 5 percentage points. Plugging into Kelly: b = 1.50, p = 0.45, q = 0.55. The formula gives you (1.50 × 0.45 – 0.55) / 1.50 = 0.085, or 8.5% of your bankroll. But here is what nobody tells you: if your probability estimate is wrong by just 3 percentage points, meaning your true edge is only 2% instead of 5%, Kelly drops to 2% of bankroll. Miss your edge estimate by 5 points and Kelly says bet nothing.
I ran this calculation backward on my first 200 tracked bets. Based on my actual results and closing line value, my real edge on plays I rated as “strong” was approximately 2.8%. I had been betting them at 5% of roll. Full Kelly for a 2.8% edge at standard -110 odds is about 2.9% of bankroll. I was overbetting by nearly double. Over 200 bets, that difference compounds into serious risk of ruin.
| Your Estimated Edge | Full Kelly Bet Size | Half Kelly Bet Size | Risk of Ruin (200 bets) |
|---|---|---|---|
| 1% | 1.0% | 0.5% | 18% |
| 2% | 2.1% | 1.0% | 8% |
| 3% | 3.1% | 1.6% | 3% |
| 5% | 5.2% | 2.6% | 0.4% |
| 8% | 8.4% | 4.2% | 0.01% |
The math is unforgiving because variance in sports betting is massive. Even with a legitimate 5% edge, you will see losing streaks of 8-12 bets regularly. If you are betting too big for your actual edge, one bad variance run ends your season.
What My 847 Tracked Bets Revealed About Edge Categories
After I started tracking properly, I categorized every bet into edge tiers based on closing line value and my confidence. Category A bets were plays where I got at least 8 cents of CLV at standard juice. Category B was 4-7 cents. Category C was 1-3 cents or a gut feel play where I could not quantify edge. Over nine months, here is what happened.
Category A plays: 118 bets, 56.8% win rate at average odds of +105, total profit $2,340 on $12,800 risked. That is 18.3% ROI. These bets earned 6-8 cents of CLV on average. Using a Kelly Calculator Sports tool with these inputs suggests my true edge was around 6-7%. I should have been betting 6-7% of bankroll on full Kelly or 3-3.5% on half Kelly. I was betting 3% flat across all categories, which meant I was underbetting my best spots and overbetting my worst.
Category B plays: 284 bets, 51.4% win rate at average odds of +102, total profit $890 on $17,600 risked. That is 5.1% ROI. My CLV averaged 3-4 cents. True edge probably 3-4%. Kelly says bet 3-4% full or 1.5-2% half. I was betting 3% flat, which was near the top end for this tier. Acceptable, but not optimal.
Category C plays: 445 bets, 48.5% win rate at average odds of -105, total loss $1,580 on $26,700 risked. That is -5.9% ROI. These were the gut plays, the game I watched and thought I had an angle, the revenge spots. I lost at worse than break-even rates. Betting 3% on plays with zero edge or negative edge burned through $1,580 that could have been allocated to Category A spots.
The lesson was surgical: my bankroll management for sports bettors framework needed to match bet size to demonstrated edge category, not use flat stakes across all plays. I should have been betting 4% on A plays, 2% on B plays, and 0% on C plays. Instead I bet 3% on everything and gave back 35% of my Category A profits to Category C nonsense.
Building Your Edge-Based Staking Framework
Here is the framework I use now after two years of tracking and adjustment. First, you must define your edge categories with hard criteria, not feelings. I use closing line value as my primary measure because it is objective. If you consistently beat the closing line by 5+ cents, you have real edge. If you are getting 2 cents or losing to the close, you probably do not.
Defining Edge Tiers With Measurable Criteria
Tier 1 (Premium Edge): Plays where you have 8+ cents of CLV, a significant market inefficiency, or proprietary data advantage. Your win rate should exceed 54% on standard -110 plays or you should be getting significant plus money. Bet size: 4-5% of bankroll on half Kelly. These are rare. I might find 2-3 per week during peak season.
Tier 2 (Solid Edge): Plays with 4-7 cents of CLV, line shopping advantages, or clear analytical edges confirmed by multiple models. Expected win rate 52-54% at standard juice. Bet size: 2-3% of bankroll. These are your bread and butter. I might have 6-10 per week.
Tier 3 (Marginal Edge): Plays with 1-3 cents of CLV, small market inefficiencies, or uncertain edges you want to track. Expected win rate 51-52%. Bet size: 1% of bankroll, tracking purposes only. These help you identify emerging edges but should not be big profit drivers.
Tier 4 (Entertainment/No Edge): Everything else. Bet size: 0% of serious bankroll. If you want to bet these games for fun, use separate entertainment money that you consider already spent. Do not confuse entertainment betting with investment betting.
Using this framework over the past eight months, my ROI jumped from 3.2% to 7.8% on basically the same volume. The difference was not finding better bets, it was allocating more capital to the bets I was already identifying correctly and cutting out the bets where I was fooling myself about having edge.
Adjusting For Correlation and Variance
One thing the basic Kelly formula does not account for is correlation between bets. If you are betting NFL sides and totals in the same games, your bets are correlated. A blowout might win your side but lose your over. A defensive struggle does the opposite. I tracked this across an entire season. When I had two positions in the same game, my effective variance was about 30% lower than two independent bets, but my edge was also diluted by about 15% because sharp line value in one market often means the other market is sharper too.
My adjustment: when betting correlated positions, reduce total exposure by 25%. If I would normally bet 4% on a Tier 1 side and 4% on a Tier 1 total in the same game, I bet 3% on each instead, for 6% total exposure rather than 8%. This kept me solvent during a brutal three-week stretch where game scripts were wildly unpredictable and correlations kept hitting in the worst possible combinations.
Variance in sports betting is also sport-specific. I tracked this meticulously. My NFL bets showed a standard deviation of about 1.08 units per bet. My NBA bets showed 1.15 units per bet. My MLB bets, especially totals, showed 1.22 units per bet. Higher variance sports demand lower bet sizing for the same edge. I now adjust my framework down by 10-15% for high-variance MLB and NBA totals compared to lower-variance NFL sides. You can calculate your own variance by tracking every bet and using standard deviation formulas, or you can visit Betting Data Lab for historical variance data by sport and bet type.
Where This Framework Breaks Down
This system works if you can accurately estimate your edge. That is the giant hole in the middle of everything. I thought I was estimating edge well until I compared my predictions to closing lines for 200 straight bets. I was calibrated terribly. Bets I rated as 60% winners hit at 52%. Bets I rated as 55% winners hit at 49%. I was systematically overconfident by 3-5 percentage points.
If your edge estimates are wrong, Kelly will destroy you faster than flat betting. A 5% bet on what you think is a 5% edge but is actually a -1% edge (meaning the book has the advantage) will drain your bankroll at an accelerating rate. Over 100 bets, you will lose roughly 10-15% of your starting roll. Make that mistake with 10% stakes and you are approaching ruin inside 50 bets.
The fix is painful: you must track closing line value religiously and calibrate your estimates against actual results. I use an EV Calculator to track expected value based on my probability estimates versus the actual odds offered, then compare that to results. After 200 tracked bets, patterns emerge. You will see where you are overconfident (usually favorites and home teams in my case) and where you are underconfident (usually divisional underdogs). Adjust your probability estimates based on this feedback, not based on results of individual bets. A +300 underdog that wins does not mean you found a great bet. It means you hit low probability variance. The data over hundreds of trials tells you if your edge estimates are real.
Bankroll Adjustment Rules You Cannot Ignore
Your bankroll changes every day. Winning streaks grow it, losing streaks shrink it. I see bettors make two mistakes constantly. First mistake: they set a starting bankroll, lose 30%, and keep betting the same dollar amounts. Now their “3% bets” are actually 4.3% of their remaining roll. Risk of ruin skyrockets. Second mistake: they set a starting bankroll, win 50%, and keep betting percentages of the new larger roll. Now variance can swing them more in dollars than their original starting roll, and they panic-sell during normal variance drawdowns.
My rule: recalculate bet sizes every two weeks based on current bankroll. If I am down more than 20% from peak, I treat the current balance as my new starting roll and begin a recovery phase with bet sizes reduced by an additional 25%. If I am up more than 50% from starting roll, I pull out 30% as profit and reset to a comfortable working bankroll. This keeps the psychological game manageable. A $5,000 starting roll that grows to $8,000 becomes a $6,000 working roll with $2,000 banked. Variance on the $6,000 feels less brutal than variance on $8,000, even though the dollar risk is similar.
| Bankroll Change | Action Required | Bet Size Adjustment |
|---|---|---|
| Down 10-20% | Review recent bets for edge degradation | Reduce by 10% |
| Down 20-30% | Enter recovery mode, strict Tier 1 only | Reduce by 25% |
| Down 30%+ | Stop betting, analyze tracking data | Full pause |
| Up 25-50% | Continue current plan | Maintain percentage stakes |
| Up 50%+ | Bank 30% profit, reset working roll | Reset to new baseline |
These rules saved me during a brutal stretch last season. I went on a 3-11 run that dropped my roll 22%. Old me would have kept firing 4% bets trying to win it back. New me dropped to 3% bets, cut all Tier 2 and Tier 3 plays, and focused only on my sharpest Tier 1 spots. It took six weeks to recover, but I recovered. Without adjustment rules, that 22% drawdown could have spiraled into 40%+ and threatened my entire season.
How Often Should You Bet Zero Percent?
The hardest discipline in bankroll management for sports bettors is betting nothing when you have no edge. I went three full days without placing a bet last month because I could not find a single play that met my Tier 1 or Tier 2 criteria. My brain screamed at me to find something. I had spent hours on research. I wanted action. But the ROI Calculator does not care about your feelings. Betting without edge produces negative expected value regardless of how much work you put in.
I tracked this for three months. Any week where I made at least one bet that did not meet my edge criteria, my overall ROI for that week dropped by an average of 2.1 percentage points. Weeks where I maintained strict discipline and only bet Tier 1 and Tier 2 spots, my ROI averaged 9.3%. Weeks where I let myself make “just one” entertainment bet or reached for a Tier 3 play that did not quite qualify, my ROI averaged 7.2%. That 2.1 point difference over a season is the gap between a winning year and a breakeven year on the same volume.
Zero percent is a bet size. Use it liberally.
FAQ: Bankroll Management By Edge Size
What if I cannot calculate my true edge accurately?
Start with half Kelly or smaller until you have 200+ tracked bets with documented closing line value. Most bettors overestimate their edge by 2-4 percentage points. Using half Kelly gives you room for error. As you accumulate data, your edge estimate will converge on reality. Until then, assume you have less edge than you think you do.
Should I ever use full Kelly bet sizing?
Almost never. Full Kelly maximizes long-term growth rate but produces massive volatility. A bad 20-bet run at full Kelly can cost you 40-50% of your bankroll even with a real edge. Half Kelly delivers 75% of the growth rate with half the variance. Quarter Kelly is even safer for part-time bettors who cannot handle the emotional swings of proper bankroll volatility.
How do I handle bankroll management across multiple sportsbooks?
Treat your total bankroll as one pool even if it is spread across books for line shopping. Your bet size percentages should come from total bankroll, not individual book balances. I keep 40% of my roll at my main book, 20% each at two secondary books for line shopping, and 20% in reserve. This ensures I can always fire my full calculated bet size at the best available line without being limited by individual book balances.
Explore more strategies in our NBA Player Prop Correlations: How Points, Rebounds, and Assists Move Together.