How the numbers are made

What it is

The result of one bet to level stakes of one point. A back of 1pt at 4.0 that wins is +3.00; a loser is −1.00. A lay of 1pt at 3.8 that wins is +1.00; a loser is −2.80, because laying 1pt at 3.8 puts 2.80 on the line to win 1. P&L is the same number as a percentage of the stake — +300% and −280% — for anyone who thinks in returns.

Why it matters

Points is how every betting record in the world is read. It makes a strategy comparable to a tipster’s sheet, and it makes a lay’s liability visible: −280% looks like a mistake until you see it is 2.8 points staked to win 1.

How it is worked out

Back: win = (price − 1) × stake, lose = −stake. Lay: win = +stake, lose = −(price − 1) × stake. Then commission comes off the winners only — see the next section.

On the example

In the League One example the average lay price was 3.81. A winning lay pays +1.00 less commission = +0.98; a losing one at 3.81 costs −2.81.

What it is

Betfair takes 2% of what you win on a market, and nothing when you lose. Gross is the return before that; Net is after it. Every figure on the simulator that says “net” has the 2% already taken.

Why it matters

Commission is the hurdle every edge has to clear, and it is charged on the winners, so a strategy that wins often pays more of it than one that wins rarely at big prices. Two strategies with the same gross can have different nets.

How it is worked out

For each bet: if it won, keep 98% of the profit; if it lost, keep the loss as it is. Net is the average of those per £1 staked. The premium charge, if you are on it, comes off the period’s profit once rather than bet by bet.

On the example

League One: gross +5.04% a bet, net +3.54%. The gap is 1.5 points a bet, on a strategy that wins 74.7% of the time — the winners are many and small, so commission bites.

What it is

Won is how often the bet came off. Implied is how often the price you took said it would. Edge is the gap between them.

Why it matters

A price is a probability written the other way up: 4.0 means “one in four”. If the thing you back lands more often than the price said, you are getting paid too much — that is what an edge is. The gap is a signpost, not proof: the rest of this page is about whether the gap is real.

How it is worked out

Implied = 1 ÷ price, averaged over the bets. For a lay it is turned round — laying the draw at 3.81 is betting on the 73.8% chance it does not happen. Edge = Won − Implied. Only read it on bets held to the result; a trade closed early wins when the price moves, which is a different event.

On the example

League One: the draw did not happen 74.7% of the time; the prices said 73.3%. Edge +1.4 points. Small, and as you will see below, not yet shown to be real.

What it is

The close is the last price matched on Betfair before the market went in-play. Alongside it we read the same line three more ways over the final minutes before the off — the middle price (median), the average (mean) and the most common(mode). The window is 5 or 10 minutes, set in section 06.

Why it matters

One price is one moment. Four readings show the run-up as well as the close itself, so you can see whether the market had settled on a number or was still moving when the whistle went.

How it is worked out

We hold one price a minute for every runner of every game. The close is the last minute before in-play with a trade in it. The other three are the median, mean and mode of the prices in the final 5 or 10 minutes. The mode is exact because Betfair prices sit on a ladder of 350 fixed steps.

On the example

League One CLV against the close: +0.39%. Against the middle of the last ten minutes +0.22%, the average +0.25%, the most common +0.21%. All four agree the price taken was a shade better than the market’s last word — and all four agree it was only a shade.

What it is

CLV — closing line value — is how much better your price was than the close, on average. Beat close is the share of bets that got a better price than the close at all.

Why it matters

A strategy can lose money for a season and still be right, and CLV is what tells you before the money does. If you keep beating the close, you keep taking prices the market later agreed were too big — that edge shows up in the P&L eventually. If you never beat it, a profitable run is luck. This is why the site puts CLV above P&L.

How it is worked out

For a back: (your price − close) ÷ close. Backing at 4.0 when it closed at 3.6 is +11.1%. For a lay the sum runs the other way — you want to have laid at a shorter price than the close. Averaged over the bets struck before kick-off; a bet struck in-play has no close and is left out rather than counted as zero.

On the example

League One: CLV +0.39%, beat close 46.2%. Under half the bets beat the close, and the average gain was under half a percent. That is not the signature of a real edge — a real one beats the close more often than not, by more than that.

Worth knowing

Beat close counts a bet only when the price was strictly better than the close. Getting exactly the closing price counts as not beating it, which is why the figure can sit under 50% while CLV is positive.

What it is

Expected is what the closing prices say the bets should have returned. Luck is what they actually returned, less that.

Why it matters

This is the cleanest way to split skill from fortune. If the closing prices say a set of bets should have lost 1% and they made 3.5%, the 4.5% between is luck until the CLV says otherwise. Run the other way, a strategy that should have made 3% and made nothing is a real edge on a bad run — the thing you ride out rather than kill.

How it is worked out

Take the close as the true chance (1 ÷ close). For each bet, work out the return that chance implies at the price you took, commission included: for a back, chance × (price − 1) × 0.98 − (1 − chance). Average it. Luck = actual net on those bets − expected.

On the example

League One: expected -1.08% — the closing prices say laying the draw here should lose about 1% a bet, which is roughly the commission. Actual net +3.54%. Luck +4.63%. Nearly all of the profit is above what the market’s own prices justify.

What it is

The standard deviation of the result per £1 bet — how far a typical bet lands from the average. Big for long prices, small for short ones.

Why it matters

It is the number behind everything else in this part. Two strategies with the same return but different spreads are not equally proven: the noisier one needs many more bets to say the same thing.

How it is worked out

Take each bet’s net result per £1, subtract the average, square it, average the squares, take the square root. The same sum as Excel’s STDEV.

On the example

League One: spread 164.2%. Every bet lands, on average, about 1.6 points from the +0.035 mean — the winners are +0.98, the losers around −2.8. That is a very noisy series, and it is why a 3.5% average over 3,727 bets still is not proof.

What it is

How many standard errors the average sits from zero. Under 2 either way is noise. Negative means reliably losing, not merely weak.

Why it matters

It is the classic test for “could this be chance”, and it scales properly: doubling the bets makes the same average about 40% more convincing, not twice.

How it is worked out

t = average ÷ (spread ÷ √bets). The bottom half is the standard error — how much the average would wobble if you ran the same number of bets again.

On the example

League One: 3.54% ÷ (164.2% ÷ √3,727) = 3.54 ÷ 2.69 = 1.32. Short of 2. The whole record, every league: -1.76 — reliably a loser.

What it is

The chance of a result this far from zero if there were no edge at all. It is t, said as a probability.

Why it matters

“2.3 standard errors” means nothing to most people. “A 2% chance this is luck” means something. Under 5% is the usual bar for calling something significant; under 1% is strong. It needs to be below the bar, not near it.

How it is worked out

The area in both tails of the normal curve beyond ±t. The same sum as Excel’s 2 × (1 − NORM.S.DIST(ABS(t), TRUE)).

On the example

League One: t of 1.32 gives p = 18.8%. Nearly one in five. That is a result you would see by chance one time in five with no edge at all — well short of the 5% bar.

What it is

Instead of one figure for the return, the band the true return most likely sits in. If the low end is below zero, the edge has not been shown. The confidence — 90, 95 or 99% — is set in section 06.

Why it matters

This is the honest version of every return figure. “+3.5%” sounds like a fact. “Somewhere between −1.7% and +8.8%” is the truth, and it makes clear the strategy could be a loser.

How it is worked out

Average ± z × standard error, where z is 1.645 at 90%, 1.96 at 95%, 2.576 at 99%. Excel’s CONFIDENCE.NORM gives the ± part.

On the example

League One at 95%: 3.54 ± 1.96 × 2.69 = -1.7% to +8.8%. Zero is inside it. At 90% the band is −0.9 to +8.0; at 99% it is −3.4 to +10.5. Whichever you choose, the low end is below zero.

What it is

How many bets it would take, at this return and this spread, for the result to clear the bar. Blank when the return is not positive — there is nothing to prove.

Why it matters

It turns “not enough data” into a number. If a row shows a profit on 400 bets and the answer is 2,000, you know how far off you are — and whether it is worth waiting.

How it is worked out

Bets needed = (z × spread ÷ average)². It is the t formula solved for the number of bets at which t would reach z.

On the example

League One: (1.96 × 164.2 ÷ 3.54)² ≈ 8,259 bets at 95%. It has 3,727 — under half. At this rate it would take another four or five seasons to know.

What it is

The chance the strategy is genuinely profitable, as a percentage. An estimate, and labelled as one.

Why it matters

A straight answer to the question everyone actually asks. It starts sceptical — assume there is no edge — and lets the bets pull it from there. A big return on few bets is pulled hard back towards even money; the same return on thousands barely moves.

How it is worked out

Bayesian. The starting assumption is no edge, with an edge of ±2% a bet treated as about as large as real ones get. That assumption and the evidence are combined in proportion to how sure each is, and the result is the chance the combined estimate sits above zero.

On the example

League One: 78%. Promising, not proven. The whole record across every league: 5%.

What it is

Up is how many seasons made money, out of the seasons with enough bets to say. Less best is the whole record with its best season taken out.

Why it matters

A record built on one enormous season and four flat ones is not the same thing as one that made money four years out of five, and the headline cannot tell them apart. If Less best is far below Net, one season is the strategy.

How it is worked out

Split the bets by season, count the ones in profit. Then find the season that made the most money — by money, not by rate, because a good rate on a thin season proves nothing — and recompute the net without it.

On the example

League One: 6 of 8 seasons up. Less best +2.32% against a net of +3.54% — so the best season carries about a third of it, which is normal rather than alarming. Six of eight is the more encouraging figure on the row.

What it is

Drawdown is the worst fall from a previous high, in points. Dry is the most bets in a row without beating the previous best. Losing run is the most losers in a row.

Why it matters

A strategy that returns 59% a bet and spends two years below its previous high is the same strategy as one that returns 59% a bet — but only one of those descriptions gets it abandoned in month nine.

How it is worked out

Run the bets in date order keeping a running total. Track the highest the total has been. Drawdown is the biggest gap below that high; dry spell is the longest stretch below it; losing run is the longest string of negatives.

On the example

League One: worst drawdown 58.6 points — at £100 a bet, £5,860 down from a high before it recovered. Longest losing run 5.

What it is

The longest losing run that is normal at this strike rate over this many bets. Read against Losing run.

Why it matters

This is the heart of the platform: telling a normal losing run of a real edge from an edge that has died. A run inside this number is the ride — sit tight. A run well past it is the thing to worry about.

How it is worked out

ln(bets) ÷ −ln(1 − strike rate). Backing 4/1 shots over a thousand bets makes a run of thirty normal; laying 1.2 shots makes a run of three a surprise.

On the example

League One: strike rate 74.7%, 3,727 bets → run to expect 6. Actual longest run 5. Inside it. Nothing here says the edge has died — the other measures say it was never shown to be alive.

What it is

Every form rule — “the home side has scored in its last five” — is built only from games played before the fixture being judged.

Why it matters

Using anything from the same day or later is looking at the answer before giving it. A backtest that does it will show an edge that vanishes the moment you try to trade it.

How it is worked out

For each fixture, the form window is cut at the day before. What the side did that afternoon, or the week after, is not in it.

On the example

A rule reading “last 5 home games” on a fixture dated 14 March uses the five home games before 14 March, and nothing from 14 March itself.

What it is

Section 05 lets you hold back a stretch of games — by date, season or year — that nothing above it is allowed to see. The strategy is built on the rest, then run on the held-back half.

Why it matters

Any strategy can be tuned until it fits the past. The only test that means anything is games it never saw. If it works on those, it might work next season; if it only works on the games it was tuned on, it will not.

How it is worked out

The held-back games are reported on their own and never folded into the headline — a figure that quietly includes them is not a test of anything.

On the example

Build on 2018–2023, hold back 2024 onwards. If the training half shows +3% and the held-back half shows +2.5%, that is a real candidate. If the held-back half shows −1%, the +3% was fitted, not found.

What it is

A filter on which games you bet. Each row is one condition: this market, this runner, a price between these bounds, read at this moment. A game only counts if the condition holds. It never changes when you bet — that is the minute or trigger above it.

Why it matters

Most strategies are not “lay every draw” but “lay the draw when the home side is a modest favourite”. The boundary is how you say that. The runner does not have to be the one you are betting — laying the draw only when the home side is under 1.8 is a condition on the home price while you trade the draw, and that is exactly the point.

How it is worked out

Each bound has its own comparison, the way Excel would ask it — ≥ or > on the low side, ≤ or < on the high side. On a price ladder the difference is one rung, and it matters: “home side ≤ 2.0” and “home side < 2.0” are different sets of games when 2.0 is a popular price. Then when it is read: at the opening minute (the same minute your position opens — the normal case); at a minute of its own (“only if the home side was under 1.8 three hours out”); or between two minutes, either held at any point in the window or held the whole time. A game with no price at the moment asked is left out — a gap in the data must not read as a pass.

On the example

Lay the draw at −60, only where the home side is ≤ 2.0 at the opening minute: 14,689 games. Change ≤ to <: 14,310 games — the 379 games where the home side was exactly 2.0 are the difference. “Match Odds · Home · ≥ 1 and ≤ 5” lets nearly everything through, because no price is under 1.01 and 5.0 covers most home sides — a boundary that wide is not filtering anything.

Worth knowing

A window here is not a trigger. Every qualifying game is still struck at the minute set above; the window only decides whether the game qualifies. The next section is the difference.

What it is

Section 02 decides when you get in and out — a minute, or a trigger that fires the first moment a condition holds. The price boundaries decide which games count. They never touch the entry.

Why it matters

The two pick a different set of games and a different set of prices, so swapping one for the other silently changes the answer. “Only when the draw is above 3.5” is a boundary; “get in the moment the draw goes above 3.5” is a trigger. Same words, different bet.

How it is worked out

A boundary reads the price at the opening minute (or a minute of its own, or across a window) and lets the game through or not. A trigger searches a window for the first minute the condition holds and opens there; a game where it never fires gets no bet, counted apart from the losers.

On the example

Lay the draw at −60 only if the draw is above 3.5: one bet per qualifying game, all struck at −60. Lay the draw when it first goes above 3.5 between −180 and kick-off: a bet at a different minute in every game, and none in the games where it never happened.

What it is

Minute 0 is the moment the market went in-play. Minutes before are negative; minutes after run on the wall clock, half time included.

Why it matters

Betfair is the only source of truth for prices, and its prices carry real timestamps. Pre-match, that is exactly what you want. In-play it means +70 is seventy real minutes after kick-off — about the 55th minute of football — because the fifteen-minute break is inside it.

How it is worked out

One price per runner per minute, from six hours out to the final whistle, keyed to the in-play tick. A translation to match minutes is planned; until then, in-play minutes are wall-clock minutes.

On the example

“Close at +45” closes forty-five real minutes after kick-off — roughly the half-time whistle plus stoppage time, not the 45th minute on the referee’s watch.