Odds PiranhaBankroll Management & Unit Sizing
Survival · Updated August 2026 · ~6 min read
You can pick winners and still go broke. Betting too big means a normal cold streak — and cold streaks are guaranteed — can wipe you out before your edge ever pays off. Bankroll management is the discipline that keeps you in the game long enough for +EV to win. It's the least glamorous skill in betting and the one most responsible for who's still standing at the end of the season.
Start with units, not dollars
Your bankroll is the money you've set aside for betting — money you can afford to lose, entirely separate from rent and bills. A unit is a fixed slice of it, usually 1% of your bankroll. Thinking in units instead of dollars keeps your bet sizing rational and your emotions out of it.
$2,000 bankroll, 1% unit = $20 per unit. A "2-unit" bet is $40. If the bankroll grows to $3,000, a unit becomes $30 — you scale up with success and down with drawdowns automatically.
Flat vs. percentage staking
- Flat betting — the same 1 unit on every play. Boring, stable, and hard to blow up. The best default for almost everyone.
- Percentage staking — always risk a fixed % of your current bankroll. It naturally shrinks bets during losing runs and grows them during winning ones.
Avoid the trap on the other end: chasing losses by doubling up (the Martingale). It feels logical and it is how bankrolls die. One long streak — which will happen — and you're either broke or over the table limit.
The Kelly criterion
Kelly is a formula for the mathematically optimal bet size given your edge. For a simple win/lose bet:
f = (bp − q) ÷ b
b = decimal odds − 1 (profit per $1) · p = your true win probability · q = 1 − p
A +150 bet (b = 1.5) you believe wins 45% of the time (p = 0.45, q = 0.55). f = (1.5 × 0.45 − 0.55) ÷ 1.5 = 0.083 → full Kelly says risk 8.3% of your bankroll.
Full Kelly is brutally volatile and assumes your probability estimate is exactly right — it never is. Most serious bettors use quarter Kelly to half Kelly (25–50% of the number), which keeps almost all the growth with a fraction of the swings. When your edge estimate is shaky, size down.