Bankroll management from first principles

Position sizing decides whether a positive-EV approach actually compounds or gets wiped out by variance before the edge plays out. Most retail bettors lose to position sizing before they lose to bad picks. Solving this is unglamorous and the highest-leverage thing you can do.

What a bankroll actually is

A bankroll is the amount of money you can afford to lose without affecting your life. Not the amount in your sportsbook app. The amount you have psychologically separated from rent, savings, food, and travel.

If a $5,000 loss would meaningfully change your life, your bankroll is not $5,000. Your bankroll is what is left after subtracting that constraint.

Unit sizing

Most operators express bet size as a percentage of bankroll. A 'unit' is your standard bet size. Common unit sizes are 1% to 3% of bankroll. A bettor with a $10,000 bankroll and 1% units bets $100 on a standard play.

BANKROLL          $10,000
  1% unit            $100
  2% unit            $200
  3% unit            $300

BANKROLL          $25,000
  1% unit            $250
  2% unit            $500
  3% unit            $750

Why fixed percentage and not fixed dollar? Because fixed-dollar sizing turns a bad month into a slowly accelerating disaster (your bets become a larger fraction of a shrinking bankroll). Fixed-percentage sizing is self-correcting. As the bankroll grows, units grow. As the bankroll shrinks, units shrink. Variance becomes survivable.

Why 1 to 3 percent?

The math behind unit sizing comes from the trade-off between rate of bankroll growth and probability of ruin. Bigger units compound faster on positive-EV bets but suffer larger drawdowns and a higher chance of going bust during a normal losing streak.

1% units are conservative. A bettor flat-betting 1% on +5% EV plays still grows the bankroll meaningfully over a season, with extremely low ruin risk. 3% units are aggressive enough to feel meaningful and aggressive enough to lose half the bankroll on a normal cold streak. Above 3%, you are probably overbetting.

The formal answer to 'how much to bet' is the Kelly criterion. See Kelly criterion. Kelly suggests fraction of bankroll proportional to your edge. For most amateurs, fractional Kelly (a fixed fraction of full Kelly, often a quarter) lands close to 1% to 2% on typical sports bets.

Conviction-proportional sizing

Flat unit sizing is a starting point, not the optimum. A bettor with a sound conviction estimate can size proportional to conviction: smaller bets on lower-conviction plays, larger bets on the highest-conviction plays.

Most retail bettors invert this. They risk small on the bets they actually believe in and chase larger on the bets they tilted into. The results are predictable.

WagerBird's confidence scoring exists to make conviction-proportional sizing concrete. The model rates every signal 25 to 100. A 95-confidence signal is sized larger than an 80-confidence signal. The discipline is built into the unit definition. See conviction-proportional sizing.

What a normal losing streak looks like

Even a profitable bettor will hit losing streaks that feel like the strategy is broken. Calibrate your expectations.

A bettor with a 55% hit rate on -110 lines (which is well above breakeven and rare in practice) will still go 0-5 on a five-bet stretch about 1.8% of the time. That happens roughly once a year for someone betting daily. A 4-12 cold stretch is normal variance, not strategy failure.

55% TRUE WIN-RATE BETTOR
  prob(0 wins in 5 bets)   = 0.45⁵ = 1.8%
  prob(<= 4 wins in 16)    ≈ 11%
  prob(<= 25 wins in 60)   ≈ 12%

  Translation: even a clearly winning bettor
  has about a 1-in-8 chance of looking
  underwater across any 60-bet stretch.

Tilt and exits

Two specific failure modes destroy more bankrolls than bad picks: tilt and the lack of an exit plan.

Tilt means deviating from your sizing rules to chase losses or press wins. The classic version is doubling unit size after a losing day to make it back. The math on this is brutal. Doubling units on a still-bad strategy doubles the rate of ruin without changing the long-run expected value. There is no version of this that ends well.

Exit plans are the rules you set before a session starts: max bets per day, max loss per day, max increase in unit size per session. Operators write them down. Casuals do not. The difference shows up in the equity curve.

Record keeping

Track every bet. Date, sport, market, odds, stake, result, closing line, sportsbook. The minute it stops being recorded, you stop being able to evaluate your performance honestly.

What you are looking for in the data:

• Hit rate by category (sport, market type, level of conviction).
• ROI by category. Hit rate without ROI is incomplete.
• Closing line value (how often did you bet a price better than the eventual close).
• Drawdown (the largest peak-to-trough loss).
• Bet size discipline (did your size correlate with conviction or with mood).

If you are subscribed to the WagerBird Terminal, the ledger surface tracks all of this automatically. If you are not, a spreadsheet works fine. The work is the same.

What to read next

Kelly criterion is the formal version of the question 'how much should I size this bet.' Variance and drawdown goes deeper on what a normal losing streak looks like and how to plan for it. Expected value is the input that makes any of this matter.