Two True Statements About the Same Strategy
A three percent edge is real. It is also close to useless over a year. Both statements are true and the apparent contradiction is resolved entirely by understanding variance.
Expected value describes a long-run average across a very large number of bets. Variance describes how far actual results stray from that average. In betting the variance per bet is enormous — typically comparable to the stake itself — so the path of actual results is dominated by variance for a very long time.
Our the value betting guide covers the practical method, and the detection guide covers how to find a genuine edge in the first place.
What Variance Actually Does to an Account
The arithmetic of variance is unglamorous and it explains the entire experience of betting with an edge.
With an even-money bet, the standard deviation of the result is on the order of the whole stake. A three percent edge is a tiny shift in the mean of a distribution whose spread is a hundred times wider. After any realistic number of bets, the mean is invisible inside the noise.
The consequence is that a bettor with a genuine edge will experience long losing runs that feel like proof the edge does not exist. This is not a failing strategy. It is what a positive expectation looks like from inside.
What Sample Size a Real Edge Requires
Distinguishing a small edge from random variation takes a very large number of observations, and the relationship is not intuitive.
| True edge | Approx. bets to distinguish from noise | Practical? |
|---|---|---|
| 0.5% | Many thousands | No |
| 2% | Several thousand | Rarely |
| 3% | Around a thousand | Marginal |
| 5% | Several hundred | Occasionally |
| 10% | Tens of hundreds | Feasible |
| 20%+ | A few hundred | Realistic |
Why a Larger Edge Matters More Than It Looks
The practical consequence of the sample-size problem is that the size of an edge matters enormously, and most bettors treat a two percent edge and a fifteen percent edge as equally valuable.
A fifteen percent edge can be demonstrated in a few hundred bets, which is a feasible sample. A two percent edge cannot be demonstrated in a practical career, and a bettor who is not sure which they have is not in a position to know.
This is a strong argument for passing on marginal bets entirely. A bet that is only slightly better than the price is a bet whose existence cannot be verified, and treating it as a real edge is the error. Our the trading guide covers the same reasoning applied to a different market.
Bankroll Size Decides Whether the Edge Survives
An edge is only realisable with a bankroll large enough that the worst plausible run does not end the process, and this is a numerical question rather than a judgement call.
With a three percent edge, a bettor must be able to survive a run of consecutive losses long enough for the mean to reassert itself. A bankroll that allows a hundred consecutive stakes to be lost is far short of what a three percent edge requires; one that allows several thousand is closer to adequate.
The consequence is brutal but simple. A small edge with a small bankroll is a losing proposition, because the probability of ruin before the edge manifests is high. Our the beginner bankroll guide covers the sizing, and the limits guide covers the point at which a bookmaker restricts a successful account.
A Realistic Expectation
The most useful thing a bettor can do is set an expectation that will not cause them to abandon a genuinely sound approach.
Assume that a three percent edge will show months or years of results indistinguishable from random, and that a losing run of several hundred bets is normal rather than evidence of failure. Judge the process over thousands of bets, not weeks.
Prefer fewer, larger bets where the edge is real, rather than many small ones where it is marginal. Record everything, including the bets passed, because the record of what you declined is the real test of whether your judgement is any good. Our the journal guide covers structuring that record.
- A 3% edge shows nothing for hundreds of bets
- Edge size matters more than existence, be honest about which you have
- Bankroll must survive a long bad run, size for it
- Pass on marginal bets, you cannot verify them
- Record what you declined, that is the real test
- Judge over thousands of bets, not weeks
- Set limits with our tools guide
Kelly Criterion Without the Mathematics
The most important practical result in this entire article requires no arithmetic at all, and it is worth stating on its own.
Bet a small fraction of your bankroll, chosen to be a proportion of the size of your edge rather than of your confidence. A bettor with a real five percent edge should stake a small percentage of the bankroll. A bettor with a real two percent edge should stake less, not more.
Almost every bettor does the opposite. They stake more when confident, which is precisely when their edge is largest and variance is least dangerous, and stake more after losses, which is precisely when the position is worst. Our the bankroll guide covers the sizing in full.
Why Most People Who Claim an Edge Do Not Have One
Claiming a betting edge is a claim about mathematics, and the mathematics are not especially generous to claims.
The most common sources of a believed edge that is not real are: evaluating a bettor's results rather than their decisions, which confuses luck with skill; and counting the bets they passed as well as the ones they took, which quietly removes most losing bets from the record.
Our the journal guide covers how to keep a record that survives this scrutiny, and the value guide covers how to test whether an edge exists before staking on it.
The Test That Distinguishes Skill From Luck
There is one procedure that reliably separates a bettor with an edge from one with a good run, and it is worth stating precisely.
Record every opportunity you declined, with the price and the reason. A bettor with a genuine edge will have a lower return when including declined bets than a bettor without one, because the declined bets are where their judgement was applying.
A bettor whose returns improve once declined bets are included has no edge, and has been benefiting from selection bias. This is the single most informative check available and it costs nothing but the discipline of recording. Our the journal guide covers how to structure the record.
Why Confidence and Edge Are Different Things
A bettor is routinely confident in a bet that has no edge, and the two are genuinely different quantities.
Confidence is a feeling about a prediction, and it is produced by a coherent story as often as by an accurate estimate. Edge is the difference between the true probability and the implied probability, and it is a number.
A coherent, well-argued, emotionally compelling view of a football match will frequently be wrong by more than the price, and the bettor will be entirely certain throughout. Our the psychology guide covers this failure mode directly.
The Honest Arithmetic of Being Right
There is a calculation that is worth doing honestly, because it produces an uncomfortable answer for most bettors.
Take a bettor who is right sixty percent of the time on even-money bets. That sounds like an edge. But even money pays one to one, and the break-even rate is fifty percent, so the ten points of accuracy is a genuine ten percent advantage on the winning bets and a ten percent overpayment on the losing ones.
The arithmetic works out to a small positive expectation — but a small one, which returns us to variance and to the fact that a small edge is invisible over a realistic sample. Our the bankroll guide covers what makes the difference survivable.
