NFL Bankroll Management: Unit Sizing, the Kelly Criterion and Surviving Variance

Editorial Team «nfl Betting Strategies» June 18, 2026 Reading time: 18 min

The Reason 90% of NFL Bettors Go Bust Before Week 10

In November 2019, I watched a friend blow through 800 pounds in a single NFL Sunday. He’d started the day with a plan — or so he told me — but after losing his first two spread bets he doubled his stake on the afternoon slate, lost again, then tripled it on the Sunday night game. By Monday morning his account balance read zero and he hasn’t placed an NFL bet since. His handicapping wasn’t terrible. His bankroll management was non-existent.

Billy Walters once said it plainly: the way the system is set up, most people have no chance to win. Zero. He wasn’t talking about the odds being unbeatable — he was talking about behaviour. The house edge on a standard NFL spread bet is small, roughly the gap between 50% and the 52.38% breakeven rate at -110 odds. That’s a margin a skilled bettor can overcome with discipline. But discipline means nothing if you’re resizing your bets based on emotion rather than mathematics.

The UK gambling market processes about 290 million online sports bets every month. The overwhelming majority of those bets are placed without any coherent staking plan. People wager what feels right, what they can “afford to lose” in the moment, or — worst of all — whatever amount is needed to chase back yesterday’s losses. That approach turns a game of thin edges into a coin flip with a loaded coin, and the coin is loaded against you.

What follows is the system I’ve used for nine years of NFL betting. It isn’t glamorous. It won’t double your money in a weekend. But it will keep you in the game long enough for your analytical edge to compound — and compounding, not any single brilliant pick, is where the real money in sports betting comes from.

Setting Your Starting Bankroll: How Much Is Enough?

The question I get asked most often at the start of every season: “How much should I put aside for NFL betting?” The honest answer is: only money whose complete loss would cause you zero financial hardship. Not mild inconvenience — zero hardship. If losing the entire amount would delay a rent payment, reduce your savings cushion below a comfortable level, or cause stress in your household, the number is too high.

Once you’ve determined that ceiling, I recommend starting with a round figure between 200 and 1,000 pounds for your first serious NFL season. The exact amount matters less than your commitment to treating it as a fixed pool — not a top-up account you replenish from your current account when things go sideways. The total UK gambling market generates 16.8 billion pounds in gross gambling yield annually, and a meaningful chunk of that comes from recreational bettors who never set a hard limit.

A starting bankroll of 500 pounds, for example, gives you enough room to bet in meaningful units (we’ll define those next) while keeping individual wager sizes small enough that a bad week doesn’t crater your confidence. At one unit equalling 2% of your bankroll, that’s 10 pounds per bet. You can place five bets a week for seventeen weeks and have enough cushion to absorb a 40% drawdown without going bust — which, as I’ll show in the variance section, is a drawdown that any competent bettor should expect to experience at least once per season.

One rule I never break: the bankroll is separate from my personal finances. It sits in its own mental (and sometimes physical) compartment. When it’s gone, it’s gone until next season. That firewall is the single most important structural decision you can make.

The Unit System: Sizing Bets as a Percentage, Not a Feeling

“I fancy this one more, so I’ll put a bit extra on it.” That sentence has cost bettors more money than any bad pick ever has. Sizing bets by intuition is how controlled risk-taking turns into gambling — and I use that word deliberately, because there’s a difference between betting with a system and gambling on a feeling.

The unit system eliminates the guesswork. One unit is a fixed percentage of your current bankroll. The standard recommendation — and the one I use — is 1-3% per bet, with 2% as the default. On a 500-pound bankroll, one unit is 10 pounds. Every standard bet is one unit. If I have higher conviction on a particular game, I’ll go to 1.5 or occasionally 2 units, but I cap the maximum at 3% of my bankroll regardless of how strong the edge looks. The reasoning is mathematical: even a bet with a genuine 60% probability of winning still loses 40% of the time, and a 3-unit loss on a bad day shouldn’t leave you reeling.

The flat-staking approach — same unit size on every bet, regardless of confidence — is the simplest version and, frankly, it works well for most people. The slight refinement I make is a tiered system: 1 unit on standard plays, 1.5 units on strong plays (where my model gives me 3+ percentage points of edge over the implied probability), and 2 units on the rare occasions where I see 5+ points of edge. Over a season, I place roughly 70% of my bets at 1 unit, 25% at 1.5, and maybe 5% at 2 units.

What I never do: bet more than two units on any single NFL game, regardless of the situation. One game is one game. A sample of one tells you nothing about your process. The discipline is in the consistency, not the individual outcome, and your unit system is the mechanism that enforces it. Keeping this rigid protects your bankroll during the weeks when the games don’t go your way — and there will always be weeks like that.

The Kelly Criterion Applied to NFL Spread Betting

The Kelly Criterion sounds like something a maths professor invented to torture undergraduates, but it’s actually one of the most practical tools in a bettor’s arsenal. Developed by John Kelly at Bell Labs in 1956, it answers a deceptively simple question: given my edge and the odds on offer, what percentage of my bankroll should I stake to maximise long-term growth?

The formula itself is straightforward. Kelly fraction equals (bp – q) / b, where b is the decimal odds minus 1 (your net profit per pound staked), p is your estimated probability of winning, and q is the probability of losing (1 – p). On a standard NFL spread bet at 1.91 decimal odds, b = 0.91. If you estimate your probability of covering at 55%, the Kelly fraction is (0.91 x 0.55 – 0.45) / 0.91 = 0.056, or 5.6% of your bankroll.

Now, 5.6% sounds reasonable until you consider what happens if your probability estimate is even slightly off. If your true edge is 53% rather than 55%, the Kelly fraction drops to 3.5%. If you’re actually at 51% — barely above the breakeven threshold of 52.38% — Kelly tells you to bet less than 1% of your bankroll. The problem is that nobody knows their true win probability to within a percentage point. Estimation error is the norm, not the exception.

This is why I — and most serious bettors I know — use fractional Kelly, typically half-Kelly or quarter-Kelly. Half-Kelly on the 55% example gives you a stake of 2.8% of your bankroll, which aligns neatly with the 2-3% unit range I outlined above. The growth rate is slower than full Kelly, but the ride is dramatically smoother. Full Kelly can produce drawdowns of 50% or more in a bad stretch; half-Kelly keeps those drawdowns manageable while still capturing most of the long-term compounding effect.

If you want to go deeper into the relationship between Kelly sizing and long-term profit tracking, I’ve written a separate piece on building a bet tracking spreadsheet that includes a Kelly calculator column you can add to your weekly log.

Understanding Variance: Why a 55% Bettor Can Still Lose 8 Straight

Here’s a number that unsettles people: a bettor with a genuine 55% win rate — which would make them one of the sharpest NFL bettors on the planet — has roughly a 6% chance of losing eight consecutive spread bets at standard -110 odds. That’s not a once-in-a-lifetime event. Over a 17-week season placing four or five bets per week, you should expect at least one losing streak of five or more games, and a streak of eight is plausible in any given year.

Variance is the statistical term for this phenomenon, and it is the single biggest reason why good bettors with sound bankroll systems still feel like they’re failing midway through the season. US sports betting revenue hit a record 16.96 billion dollars in 2025 — a 22.8% increase year on year — and a significant portion of that revenue comes from bettors who abandoned their system during a rough patch and started chasing losses with oversized stakes.

The way I’ve learned to coexist with variance is by reframing the goal. I’m not trying to win every week. I’m trying to make correct decisions every week — decisions that have positive expected value regardless of the outcome. A bet placed at +3 when the true line should be +2 is a good bet even if the underdog loses by ten. Over 200 or 300 bets, the quality of those decisions shows up in the results. Over 10 bets, it might not.

Practically, this means I never evaluate my performance on fewer than 100 bets. My first in-season review happens around Week 8 or 9, when I’ve typically accumulated enough data to see whether my process is sound. Before that, the results are dominated by noise, and making staking adjustments based on noise is worse than making no adjustments at all. The temptation to react to short-term results is the gravitational pull that drags most bettors off course, and the bankroll system is your anchor against it.

Drawdown Recovery: Adjusting Units After a Losing Streak

Your bankroll will shrink. The question is what you do when it does. Most bettors respond to a drawdown in one of two ways, and both are wrong. The first is to increase unit size to “make it back faster.” The second is to freeze completely, stop betting, and miss the rebound that often follows a losing stretch. The correct response sits between those extremes.

I use a recalculation trigger. If my bankroll drops by 20% from its peak — say from 500 pounds to 400 — I recalculate my unit size based on the new, lower balance. One unit at 2% of 400 is 8 pounds instead of 10. This reduces my exposure during a cold stretch while keeping me in the game. If the bankroll recovers past a new high, I recalculate upward. The system breathes with the balance, expanding during winning runs and contracting during losing ones.

The 20% trigger is a personal preference. Some bettors recalculate weekly regardless of performance, which is simpler but can lead to very small unit sizes after a bad patch. Others recalculate only after every 50 bets, which keeps unit sizes stable but doesn’t protect against sharp drawdowns. I’ve found 20% to be the right balance between responsiveness and stability, but the exact number matters less than the commitment to having a trigger at all.

What never changes: I don’t increase units after a loss to chase the drawdown, and I don’t decrease units after a win out of fear. The recalculation is mechanical, not emotional. If the maths says my new unit is 8 pounds, that’s what I bet, regardless of how I feel about the upcoming slate. Feelings are useful in many areas of life. Bankroll management is not one of them.

A 17-Week Bankroll Plan for the NFL Regular Season

The NFL regular season runs from early September to early January — 18 weeks, with each team playing 17 games and getting one bye. That’s a defined window, which makes it easier to plan your bankroll than in sports with year-round schedules. Here’s the framework I use each season.

Before Week 1, I set my starting bankroll and calculate my unit size. I allocate a maximum number of bets per week — for me, that’s four to six during the regular season, rising to two or three per week during the playoffs when the slate is smaller but each game carries more analytical value. This means I expect to place roughly 80-100 bets across the regular season, plus another 10-15 in the postseason.

Weeks 1 through 4 are my calibration phase. I bet at standard (1 unit) across the board, regardless of conviction level. Early-season data is unreliable — rosters are still gelling, coaching schemes are evolving, and the previous year’s statistics are a noisy predictor of current performance. I use this stretch to observe, take notes, and build my model rather than to swing for profit.

Weeks 5 through 12 are the core of the season. By now I have enough current-season data to distinguish genuine team quality from early noise. This is where I deploy tiered staking — 1 unit on standard plays, up to 2 units on strong edges. I also begin tracking my closing line value, which is the single best predictor of whether my bets will be profitable over time. About 2.7% of British adults experience gambling-related harm, and the mid-season grind is where poor habits tend to surface. If your betting starts feeling like stress rather than analysis, that’s a signal to review your process or step back entirely.

Weeks 13 through 18 and the playoffs are where the season comes together or falls apart. Lines get sharper as the market incorporates a full season of data, which means edges are thinner. I tighten my criteria in this stretch, betting only on the strongest setups. If my bankroll is healthy (above or near my starting balance), I maintain unit sizes. If I’m in drawdown, I continue with recalculated units and resist the urge to press. The worst possible outcome isn’t a losing season — it’s a losing season followed by a reckless bet that wipes out your ability to come back next year.

Emotional Discipline: Stopping Tilt Before It Drains Your Balance

Tilt is a poker term, but it describes NFL bettors perfectly. It’s the state where frustration from recent losses overrides your analytical process and you start making decisions designed to feel better rather than decisions designed to win. I’ve been there. Every bettor has. The question is whether you recognise it in time.

My personal warning signs: placing a bet within ten minutes of a bad result, backing a team I haven’t researched because “they’re due,” or increasing my unit size because I’m “confident I can get it back tonight.” If any of those impulses surfaces, I close my betting apps and don’t reopen them until the next day. No exceptions. The ten-minute rule alone has saved me from more bad bets than any analytical insight I’ve ever had.

The structural defence against tilt is pre-commitment. Before each NFL week, I write down my planned bets — side, stake, and reasoning — before any games kick off. If a bet isn’t on the list, I don’t place it. Live additions are allowed only if new information (a late injury, a weather change) fundamentally alters the value of a market I’ve already analysed. “This game looks good” is not new information. It’s an emotion wearing a thinking cap.

Finally, keep a simple mood log alongside your bet tracker. One word per day: “calm,” “frustrated,” “chasing,” “sharp.” Over a season, you’ll start to see patterns. Mine show that my worst weeks financially almost always coincide with entries marked “frustrated” or “chasing.” The correlation is strong enough that I now treat those mood flags as automatic bet-reduction triggers, dropping my maximum bets for the week from six to three. The bankroll is a number. Your emotional state is the variable that determines whether you follow the system or abandon it.

NFL Bankroll Management Questions Answered

Written by the editors at nfl Betting Strategies.