Every ranking on SportsBiff comes from the formula on this page. We publish it because a score you can't check is just another claim — and this market has enough of those.
Version 1.0 · Last changed 27 August 2026
One thing: how much a Biff returns for every pound staked.
Every settled tip is scored at a flat one unit, whatever stake the Biff advised. A winner returns the decimal odds minus one. A loser returns minus one. A void or a push returns nothing and is excluded entirely — it carries no information about skill either way. Half-won and half-lost settle at half those amounts.
We use flat stakes rather than the stake a Biff advises, and that choice is deliberate. Staking is the easiest part of a record to inflate: pile ten units on the bets you feel good about, one unit on the rest, and a mediocre record starts to look sharp. Flat stakes remove that lever entirely.
The ROI shown on a Biff's profile uses the stakes they actually advised, because that's what a subscriber following them would have made. Biff Score uses flat stakes, because that's the fairer comparison between one Biff and another. The two numbers measure different things and it is normal for them to disagree.
Win rate on its own tells you nothing, because it says nothing about price. Winning nine bets out of ten at odds of 1.05 returns 9 × 0.05 − 1 = −0.55 units. Winning three out of ten at odds of 6.00 returns 3 × 5 − 7 = +8 units. The first record has a 90% win rate and loses money. The second has 30% and makes a fortune.
Whatever information win rate contains is already inside the return figure, alongside the odds it needs to be meaningful. Adding it separately would count the same thing twice and reward Biffs who pad their record with short prices. We display it, because people expect to see it. We don't rank on it.
Total profit is return per pound multiplied by how many bets you placed. Volume is a decision a Biff makes, not a measure of how good they are. Rank on total profit and the fastest way up the board is to post forty tips a day.
Volume is still rewarded here — just correctly. More settled tips means less uncertainty, which means less shrinkage, which means a higher score. It earns its way in through the evidence term rather than being counted directly.
This is the part that makes Biff Score different from every other tipster ranking, and it is the whole reason the page exists.
A Biff with six tips and a 200% return has not proved anything. A Biff with six hundred tips and a 6% return has. Rank the two on return alone and the first one wins, which is exactly how every tipster leaderboard ends up promoting whoever got luckiest most recently.
So we calculate how uncertain each Biff's number is, and pull thin records back toward 50 in proportion. The amount left over is the confidence figure shown beside every score.
| Settled tips | Confidence |
|---|---|
| 10 | 2% |
| 25 | 6% |
| 50 | 11% |
| 100 | 20% |
| 194 | 32% |
| 400 | 49% |
| 800 | 66% |
Calculated at average odds of 2.03. Higher odds mean more uncertainty, so confidence rises more slowly — see the next section.
Confidence never reaches 100%, and it isn't supposed to. No number of bets makes a record certain. It tells you how much weight the score is putting on what a Biff has actually done, versus how much it is still holding back.
None of this hides anything from you. Return, win rate, profit and tip count are all on the leaderboard, all sortable, across every window from 7 days to all time. Biff Score decides the default order — you decide the rest.
A one-unit bet at fair odds o has a variance of exactly o − 1. A tip at 9.00 therefore swings eight times as hard as one at 2.00. Two Biffs can post the same number of tips and the same return, and one of those records is far more likely to be luck.
Because that variance comes straight from the odds, we know it from a Biff's first tip. No sample needed. It's also what makes the combined leaderboard fair: a horse racing Biff at big prices and a football Biff at short ones are compared on the same footing rather than being flattered or punished by the market they happen to bet in.
Here is what it means in practice. To show a genuine 10% edge — enough that the result sits two standard errors clear of zero — a Biff needs:
| Average odds | Variance per tip | Settled tips needed |
|---|---|---|
| 1.50 | 0.5 | 200 |
| 2.00 | 1.0 | 400 |
| 2.50 | 1.5 | 600 |
| 3.00 | 2.0 | 800 |
| 5.00 | 4.0 | 1,600 |
Almost no tipster anywhere has posted that many verified tips. That is not a criticism of tipsters — it is the reason a ranking that ignores sample size is close to meaningless, and the reason a lot of Biffs sit near 50 early on.
It is also why the leaderboard does more than one thing. Biff Score is the default order because it is the honest one. But every column sorts, and every window from 7 days to all time is there. Chasing form is a perfectly reasonable way to bet — just do it knowing that is what you are doing, on numbers nobody was able to edit.
Nothing is hidden here. Every constant we use is listed underneath.
| Constant | Value | What it is |
|---|---|---|
| τ (tau) | 0.05 | Assumed spread of genuine skill across all Biffs |
| Half-life | 90 days | How fast a tip loses weight |
| Odds cap | 26.00 | Scoring only — displayed odds are untouched |
| Variance floor | 0.10 | Stops very short prices claiming certainty |
| Scale | 0.03 | Spreads the 0–100 range sensibly |
The ranking barely depends on τ. We swept it from 0.02 to 0.12 and the order of the leaderboard stayed between 0.98 and 1.00 correlated with itself throughout. It is a sanity bound, not a dial we can tune to move anyone up or down — and that is the point.
The full calculation for one Biff, every step shown. Recency weighting is left out here so the arithmetic stays checkable by hand; with it, recent tips would simply count for more.
Read that last step carefully, because it is the whole idea. BetKing's raw return is +15.1%. The score does not treat that as their true edge — it treats it as evidence, and 194 tips is only enough evidence to act on about a third of it. The remaining two thirds is held back until more tips settle.
For context, this record sits just over two standard errors from zero, which is roughly the conventional threshold for calling a result meaningful. It is a good record. It is not yet a proven one, and the score says so.
We built a population of 3,000 simulated tipsters where we set each one's true skill in advance and therefore knew, unlike in real life, exactly who was good. Most bled the bookmaker's margin, some broke roughly even, and a small minority had a genuine edge — the mix you would expect in the real market. Each was given a volume and an odds profile, their tips were simulated, and the resulting leaderboard was ranked four different ways.
Twelve independent runs. This is what ended up in the top 25:
| Ranked by | Genuinely profitable | True average edge |
|---|---|---|
| Biff Score | 60.7% | +1.7% |
| Total profit | 38.3% | −0.6% |
| Win rate | 43.7% | 0.0% |
| Return percentage | 22.3% | −2.4% |
Be clear about what this is. It is a simulation we ran ourselves, not a study of real tipsters — nobody has a dataset of real tipsters with their true skill labelled, which is precisely why this problem is hard. What it shows is that when skill is knowable, this method finds it far more often than the alternatives. It does not prove the same margin holds in the real world.
Any ranking that only tells you its strengths is selling something. Here are the real limits.
The formula on this page takes settled tips, their odds, and their dates. Nothing else is an input, and nothing else ever will be.
Recalculate any Biff from the formula above and their public tip history. If your number disagrees with ours, tell us — contact us. If we've made an error we'll fix it and say so here, with the date.