strategy
Flat Betting
Also known as: Flat Staking, Unit Betting
Flat betting is the practice of staking the same amount on every wager, regardless of confidence or recent results. It maximizes resistance to tilt and minimizes variance compared to confidence-weighted staking schemes.
Flat betting means every bet is the same size. The bettor sizes a unit (commonly 1-2% of starting bankroll) and stakes one unit on every play. Higher-confidence picks don’t get bigger bets; lower-confidence picks don’t get smaller bets. The simplicity is the feature.
Why it works: behavioral discipline beats theoretical optimization for almost all bettors. Confidence-weighted staking (betting more on “lock” plays) sounds smart but trips on three reliable problems. First, bettors are bad at calibrating their own confidence; the picks they feel best about don’t actually hit at higher rates than their other picks. Second, raising stakes after losses or “due to be hot” picks introduces emotional bias. Third, oversized bets on the wrong picks compound into bankrull-killing drawdowns that flat staking would have survived.
Worked example. Two bettors run identical 500-bet seasons at 55% win rate against -110. Bettor A uses flat 2% staking. Bettor B uses “confidence” staking, ranging from 1% to 5%, with the larger bets on plays they feel best about. Their conviction is uncorrelated with actual outcomes. Bettor A finishes the season up roughly 4.5 units. Bettor B’s results scatter widely: sometimes up 6 units (luck), sometimes down 3 (bad bigger bets), with much higher variance. Bettor B doesn’t have an edge over Bettor A; they have more noise.
Elite Picks defaults all P&L reporting to flat-unit accounting. Fractional Kelly is available as an advanced alternative for users with calibrated probability estimates, but flat is the recommended starting framework. Discipline first, optimization second.
Example
A bettor uses 1 unit ($50) per bet for every play, whether the model rates it as 2% edge or 8% edge. Across 500 bets, total exposure is $25,000 and variance is determined entirely by win rate, not stake variation.