How much to bet per play, why flat staking beats chasing, what fractional Kelly actually solves, and why a losing month says almost nothing about whether your process works.
Bankroll management is the part of betting that decides whether an edge ever reaches your pocket. You can be right more often than the market and still go broke, because the order in which wins and losses arrive is random and a big enough losing run will end you before the math has a chance to work.
The goal is not to maximise how fast you win. It is to make sure you are still betting when the edge pays out.
A unit is a fixed percentage of your bankroll, and for most people it should be between 1% and 2%. On a $2,000 bankroll that is $20 to $40 a bet. Everything gets measured in units after that, which matters more than it sounds: it stops you from making a bet bigger because you feel confident, and confidence is not a number your model produced.
The single most common way people lose is not bad picks. It is a normal-sized edge combined with wildly inconsistent bet sizes, so the losses land on the big bets and the wins land on the small ones.
Flat staking means every bet is one unit regardless of how good it looks. It is mathematically suboptimal — a genuine 8% edge deserves more money than a genuine 1% edge — but it is robust to the thing most likely to be wrong, which is your estimate of your own edge.
Variable staking only beats flat staking if your edge estimates are well calibrated. If they are not, sizing up on your 'best' plays means concentrating money exactly where you are most overconfident.
The Kelly criterion gives the bet size that maximises long-run growth: your edge divided by the odds. It is correct, and almost nobody should use it at full strength.
Full Kelly assumes you know your true probability exactly. You do not. If your real edge is half what you think it is, full Kelly bets are roughly double the growth-optimal size, and the drawdowns become brutal — 50% drops are routine even when the strategy is genuinely profitable.
Quarter Kelly or half Kelly captures most of the growth with a fraction of the swing. If you are going to size by edge at all, size conservatively and cap any single bet at around 2% to 3% of bankroll no matter what the formula says.
A bettor hitting 55% at -110 — which is a strong, sustainable edge — will still have losing weeks constantly and losing months regularly. Over a season, a run of eight or nine straight losses is not a warning sign; it is an expected event.
This is why people abandon working strategies. The drawdown arrives, it feels like evidence, and they change something. The only defence is deciding your staking rules before the run starts and writing them down.
Never chase a loss by increasing stake size — that converts a normal drawdown into a permanent one. Recalculate your unit no more than monthly, so a good week does not immediately inflate your risk. Keep bets to markets you can actually price. And track results in units, not dollars, so the record stays comparable as the bankroll changes.
If you are betting an amount that affects how you feel watching the game, the bet is too big regardless of what any formula says.