Variance in Betfair Trading: Why a Good Edge Still Loses for Weeks

This page contains affiliate links — if you open an account through them we may earn a commission at no cost to you. It never changes our verdict.

This is a cluster sub of our betting exchange concepts pillar. The pillar defines the core ideas every exchange trader needs; this page takes the most psychologically important of them — variance — and gives it the room it deserves, because misunderstanding variance is behind more abandoned good strategies and more blown banks than almost anything else.

What Variance Is

Variance is the spread of actual results around their expected value. If your edge means you should make, on average, £2 per trade, variance is the fact that any given trade returns +£30, −£25 or nothing — and any given week can land well above or well below the £2-per-trade average. It is not a flaw in your strategy; it is the inescapable randomness of outcomes around a true average, and it exists in every form of betting and trading.

The single most important thing to internalise is that variance is symmetric and temporary, while edge is directional and permanent. Over a short run, variance can completely swamp your edge — bury it under good or bad luck so you cannot see it. Over a long enough run, variance averages toward zero and your edge is all that is left. The trader's job is to survive the short run, where luck rules, long enough to reach the long run, where skill rules.

Why Exchange Traders Feel It So Hard

Exchange traders feel variance acutely because they place a high volume of trades and watch every one settle in real time. A horse-racing scalper might make a hundred trades in an afternoon, each a small win or loss flickering on the ladder — that is a hundred little hits of randomness per session, and the emotional weight of a losing run of nine in a row is heavy even when the maths says it is routine. The constant, visible settlement makes variance viscerally present in a way it is not for someone who places one bet a week.

This is why the same trader who calmly accepts that a coin can land heads nine times in a row will panic at nine losing trades. The maths is identical; the money and the live screen make it feel different. Recognising that the feeling is not evidence — that a bad run feels like a broken strategy but usually is not — is the core skill, and it is the same emotional discipline our piece on fear and greed is built around.

Expected Value vs What Actually Happens

Expected value (EV) is what your edge earns on average; realised results are EV plus or minus variance. The trap is judging a strategy by its realised results over a short window, when over that window variance dominates and tells you almost nothing about the EV. A +EV strategy that runs badly looks identical to a −EV strategy over fifty trades — you literally cannot tell them apart from the P&L alone.

This is why serious traders judge a strategy on its process and its long-run sample, not on last week. If the entries were correct, the stops respected, the edge real and well-evidenced, then a losing week is a variance event to be ignored, not a signal to change. The concept underpinning all of this is covered in our expected value explainer — EV is the thing variance scatters around, and you have to know your EV (or at least believe in it on real evidence) before you can dismiss a bad run as noise. Converting prices to implied probability is how you check whether an edge plausibly exists at all.

Sample Size: The Whole Story

Variance and sample size are two sides of one coin: the smaller the sample, the more variance dominates; the larger the sample, the more edge shows through. Over ten trades, results are almost pure luck. Over a hundred, your edge starts to be visible but luck can still hide or fake it. Over a thousand or more, the edge is clearly expressed and variance has largely washed out. Any judgement about whether a strategy works has to be anchored to where you are on that curve.

The practical implication is humbling: most traders simply do not have enough data to know whether they have an edge. A trader fifty trades into a new strategy who declares it “working” or “broken” is reading noise. This is why the trading journal matters so much — only a logged sample of hundreds of trades lets you separate signal from variance, and only then can you make an evidence-based call rather than an emotional one.

A Feel for Standard Deviation

You do not need the formulae, but you do need the intuition: results cluster around the average, and the spread of that cluster is the standard deviation. Most outcomes land within one standard deviation of your expected result, but a meaningful minority land two or even three away — and those tail events, the brilliant week and the brutal one, are the ones that mess with your head. A run two standard deviations below expectation feels like catastrophe but happens routinely given enough trading.

The useful mental model is that your equity curve will be jagged even if your edge is real and constant — up-slopes, flat stretches, and frightening dips, all around a gently rising trend if you genuinely have an edge. When you can picture that jagged-but-rising line, a bad fortnight stops looking like the end of the trend and starts looking like one of the dips the model always predicted. The job is to keep your stake small enough that even a two-or-three-deviation dip cannot end you — which is where bankroll comes in.

Bankroll as a Variance Buffer

Your bankroll is, more than anything, a buffer that lets you survive variance long enough for your edge to express itself. The deeper the buffer relative to your stakes, the larger a bad run it can absorb without ruin. This reframes bankroll management from “how much can I bet” to “how big a losing streak must I be able to survive” — and the answer, given how routine long streaks are, is “a much bigger one than feels necessary.”

This is exactly why staking a small, fixed fraction of your bank — the discipline in our bankroll management and bankroll and risk guides — is non-negotiable. Bet too large a fraction and a perfectly normal variance dip wipes you out before your edge ever gets the sample size it needs. The Kelly criterion work is fundamentally about this trade-off: sizing to grow as fast as possible while keeping the chance of a variance-driven ruin acceptably low. Almost everyone should err toward smaller, precisely because they underestimate variance.

From the Desk: A Winning System in a Losing Month

How to Survive Variance

Surviving variance comes down to three habits. One: stake small enough that a long losing streak — far longer than you think likely — cannot end you. Two: judge your strategy on a large logged sample and on process quality, never on last week's P&L. Three: separate your emotional reaction from your decision-making, so a bad run produces a calm review of the evidence rather than a panicked change or a chase.

The fourth, quieter habit is patience with sample size: accepting that you will not know whether something works for a long time, and being comfortable trading a believed-but-unproven edge at small size while the data accumulates. Most blow-ups come from impatience with this — from needing the edge to prove itself this week and sizing up or chasing when it does not. Variance punishes impatience and rewards the boring discipline of small stakes and large samples, the same discipline running through the whole concepts pillar. When a run does damage your bank, the losing-streak recovery process takes over.

The Honest Verdict

Variance is the single concept that, properly understood, prevents the most common ways traders destroy themselves: abandoning good strategies in normal bad runs, doubling down on lucky ones, and sizing too big to survive the inevitable swings. It is not exciting, it cannot be eliminated, and it will make a real edge feel broken on a regular basis. Accepting that is the price of trading anything with a probabilistic edge.

My honest verdict: respect variance more than feels necessary, because almost everyone underestimates how long and how deep a normal losing streak can be. Stake small, log everything, judge on the long run, and keep your emotions out of the short run. Do that and variance becomes a manageable feature of the landscape rather than the thing that quietly ends your trading — and the rising trend your edge produces gets the time it needs to show up.

How Long Can a Normal Downswing Last?

The most useful thing I can give you about variance is a feel for how long a perfectly normal losing run can last, because almost everyone wildly underestimates it and quits a good strategy far too early. The intuition people carry — “if I have an edge, I should be roughly up most weeks” — is simply wrong, and acting on it is how real edges get abandoned at the worst possible moment.

Consider a strategy with a genuine 55% per-trade win rate on roughly even-money trades. That is a real, decent edge. Yet a run of six or seven losers in a row from such a strategy is completely routine — it will happen many times over a few thousand trades, just as a fair coin throws long streaks of heads. Scale that up and a profitable approach can have a losing week, or even a losing month, with nothing wrong at all; the example from the desk above, a +EV system finishing a 200-trade month at −£40, is not bad luck so much as ordinary luck doing what it does. The deeper your sample, the more such stretches you will live through, not fewer.

The practical takeaways are concrete. First, your bankroll has to be sized for a downswing far longer than feels plausible — if a six- or seven-trade losing streak can dent you meaningfully, your stakes are too big relative to your bank, and the Kelly and bankroll rules exist precisely to prevent that. Second, you cannot use a short losing run as evidence your edge has gone; only a large logged sample, reviewed against your process, can tell you that. Third, the emotional discipline to keep executing cleanly through a normal downswing — the hardest thing in trading — is what actually separates winners from the people who had an edge and threw it away. Expecting downswings, sizing for them, and trading through them calmly is not a detail of managing variance; it is the entire job.

FAQ

Related Reading

Stay in the concepts cluster: exchange concepts pillar, expected value, implied probability. Money: Kelly criterion, losing-streak recovery, bankroll management, trading journal.