Betting Units and Bankroll Management
A unit is a fixed share of your bankroll, and sizing bets in units instead of dollars is the discipline that keeps variance from wiping you out. Here is how units, fractional Kelly, and flat sizing fit together, and why Fairline tracks internal model observations at one unit.
Updated Jul 2026 · Part of the getting started series
A unit is a fixed fraction of your bankroll, usually 1%. Sizing bets in units instead of dollars is the most important discipline in betting, because it ties your stake to your bankroll and your conviction rather than to how you feel on a given night.
Why units, not dollars
Talking in units keeps strategy separate from the size of your wallet. A 2-unit play is a 2-unit play whether your bankroll is $500 or $50,000. It also makes results comparable over time, and it stops a hot or cold streak from quietly changing your bet size.
Fractional Kelly
The Kelly criterion is a formula for the mathematically optimal stake given your edge and the odds. Full Kelly is too aggressive for real betting, where your edge is an estimate rather than a known quantity, so serious bettors use a fraction of it and cap the result. Fairline’s Evaluate tool shows a fractional-Kelly stake for a specific bet, with the caveat that Kelly trusts the fair probability it is fed.
Why Fairline tracks internal observations at a flat 1u
Every internally flagged model-vs-price observation is tracked at a flat 1u, regardless of the size of the gap. These rows feed calibration and closing-line research and are not surfaced as bets. Fairline previously sized the research record on a ladder that scaled with the claimed edge, so a bigger gap carried a bigger assumed stake.
A 2026 review of the model’s own record found the model’s larger claimed edges were its worst-calibrated bets. The bigger the claimed edge, the worse that observation performed once graded against results. Scaling the research weight with claimed edge amplified exactly the rows that performed worst, and flat weighting removes that amplification. That review is why the ladder was retired for a flat 1u per observation.
The one thing you control
You cannot control variance, the schedule, or whether a star sits out. You can control how much you risk. Flat, unit-based sizing with a sensible cap is the difference between surviving a normal losing stretch (see variance) and busting out during one.