Variance: what a losing month proves

Almost every argument about whether someone is any good is really an argument about sample size. Here is what the numbers say.

Almost every argument about whether a bettor is any good is really an argument about sample size. Betting results are noisy enough that a season can tell you very little, and understanding how little is what stops you from firing a good process or trusting a bad one.

How wide the swings really are

Take a bettor with a genuine 2% edge, betting one unit at a time at standard -110 juice. That is a real, respectable edge. Over 500 bets, their expected profit is about 10 units — and the ordinary range of outcomes around it is roughly plus or minus 45 units.

Expected after 500 bets: +10u
Ordinary range: -35u to +55u
Chance of being down: ~1 in 3

A bettor with a genuine edge is losing money after 500 bets roughly one time in three. Not because they did anything wrong — because 500 bets is a small number.

This is variance: the natural spread of results around what the odds predict. It is not bad luck, it is not a slump, and it is not something that averages out on any timescale a person actually experiences.

How long is a real sample?

Roughly what a record can tell you, by size
Bets	What it distinguishes	Honest reading
50	Nothing at all	Entertainment
500	A huge edge from a huge deficit	Still mostly noise
2,000	A real edge from a real loss	Suggestive
10,000	Roughly how big the edge is	Meaningful

Ten thousand bets is several years for most people. That is the actual timescale on which a win-loss record becomes evidence, which is why the field leans so hard on closing line value instead — it is measurable per bet, and it does not need the result.

The two mistakes this causes

Abandoning a good process during a normal drawdown. A 15-unit drawdown feels like proof that something is broken. Usually it is proof that 500 is a small number.
Trusting a hot streak. A 60% month over 40 bets happens to bettors with no edge at all, regularly. Regression is not a punishment; it is what the average was the whole time.

This is also how pick sellers work

Run enough tipsters and some will show a great record by chance alone. The winning one is then advertised and the rest are quietly dropped. It is why a screenshot of a good month is worth nothing, and why graded clv over a real sample is the only claim worth taking seriously.

What to watch instead of profit

Closing line value, which is available on every bet immediately.
Whether your process was followed — bet sizes, prices taken, markets stayed in.
Drawdown depth against your unit size, which tells you whether you are betting too large regardless of how the results look.

VigWatch publishes its model's losing months as well as its winning ones for exactly this reason. A track record with no bad months in it has either been running for a very short time or is not showing you everything.

VigWatch publishes the model's losing months too — this lesson is why that matters. See a real record, losses included.