Why a real edge still spends months losing
The hardest part of value betting is not finding the edge. It is living through the part of the distribution where the edge is invisible, and not mistaking that period for evidence that the edge was never there.
§01Losing runs are not a malfunction
Take a bettor with a genuine 55% win rate on even-money bets. The chance of losing any single bet is 45%, so the chance of losing five in a row is about 1.8% — small for one attempt, but across a few hundred bets there are hundreds of overlapping five-bet sequences, and a five-loss run is close to certain to appear at some point. Longer runs appear too, just less often.
The arithmetic of independent events is unforgiving in exactly this way. Streaks feel like information because they are memorable and because human pattern recognition is tuned to find causes, but a losing run at a positive edge carries no information about the next bet.
| Run length | Chance it starts on a given bet | Expected occurrences in 500 bets |
|---|---|---|
| 3 losses | 9.1% | about 45 |
| 4 losses | 4.1% | about 20 |
| 5 losses | 1.8% | about 9 |
| 6 losses | 0.83% | about 4 |
| 8 losses | 0.17% | about 0.8 |
§02How many bets before a record means anything
Distinguishing a 55% win rate from 50% at a conventional confidence level requires on the order of a thousand bets. Distinguishing 52% from 50% requires several times that, and no recreational bettor will ever place that many. This is the single most important reason to measure closing line value rather than win rate: closing line value has perhaps a tenth of the variance of results, so a readable signal emerges from tens of bets instead of thousands.
| Edge | Bets for a win-rate signal | Bets for a CLV signal |
|---|---|---|
| 2% | several thousand | a few hundred |
| 3% | about 2,000 | a couple of hundred |
| 5% | about 800 | around 100 |
| 10% | about 200 | around 30 |
These are order-of-magnitude figures intended to make a point, not precise requirements. The point is that reading results directly is a bad instrument, and reading them over small samples is worse than not reading them at all.
§03Drawdown arithmetic in units
Express results in bankroll units rather than money and the picture becomes comparable across stakes. A bettor with a 100-unit bankroll who stakes 2 units per bet and hits a twenty-bet losing run has lost 40% of the bankroll, which is a materially different event from a bettor staking 1 unit and losing 20%.
- Bankroll
- 100 units
- Stake per bet
- 2 units
- A 10-bet losing run
- −20 units (20%)
- A 20-bet losing run
- −40 units (40%)
- Units recoverable at +0.1 EV per bet
- about 400 bets
- What the arithmetic implies
- stake small enough that 20 losses are survivable
The last two rows are the whole argument for staking well below what the edge alone would justify. A large stake converts a survivable drawdown into an unrecoverable one, and no edge protects against having no bankroll left to bet.
§04Bankroll units, and the honest answer about ruin
Risk of ruin is the probability that a series of negative swings empties the bankroll before the edge has time to express itself. It depends on the edge, the stake size and the bankroll, and it is never zero while the stake is a meaningful fraction of the bankroll. Two habits reduce it: measuring the edge conservatively, and staking as though the edge were smaller than measured.
It is worth being blunt about the direction of the risk. A value bettor who is wrong about the edge and stakes aggressively does not merely fail to profit; they lose the bankroll faster than a bettor with no edge at all, because aggressive staking on a bad estimate compounds the error.
The only guarantee
The mathematics of variance guarantees only one thing: that a large enough stake on an uncertain edge will eventually find the part of the distribution where the bankroll ends. Everything else is a probability.
Read the rules of the operator you actually use
Variance is dealt with by the operator as money, and the sponsored partner link below opens the account and payment section of the partner operator. Check the withdrawal times, the limits and the terms before you deposit anything — they apply to every bet in your sample.
Affiliate disclosure and risk warning
Every affiliate link on this page and in the header is a sponsored link to a partner operator, and we may be paid if you open an account through it, at no extra cost to you. That link pays us; it does not price the market for you, and it is never a recommendation to bet. Nothing on this page is betting, financial or legal advice, and no price, model or result on it is a prediction. 18+ only. Betting is gambling: a measured edge is an estimate and the estimate can be wrong, the operator's margin is built into every price you are offered, and a selection with positive expected value can still lose — including many times in a row. Gambling can cause serious financial harm, including debt and damage to relationships and mental health. Never stake money you cannot afford to lose, never borrow to bet, and never increase a stake to chase a loss. Free, confidential support is available in most countries from national gambling-harm helplines, for bettors and for the people around them.