Tools & Analytics · Updated
The Prognosis Engine
Pick any fixture from the next round in any of the four English divisions. Our model prices the whole match: match result, correct scores, total goals, both teams to score, Asian handicaps, cards and corners: off the expected goals it works out itself.
Showing Ipswich v Liverpool · Premier League matchweek 3
The interactive engine needs JavaScript. The model's numbers for the coming Premier League matchweek are below: expected goals, the most likely scoreline and the chance of each result, worked out from this season so far, blended with last season in the division.
| Fixture | Model xG | Home | Draw | Away | Likely score |
|---|---|---|---|---|---|
| Ipswich v Liverpool | 1.87 – 1.28 | 51% | 23% | 26% | 1-1 |
| Newcastle v Bournemouth | 2.08 – 1.69 | 47% | 21% | 32% | 2-1 |
| Brentford v Sunderland | 1.76 – 0.94 | 57% | 23% | 20% | 1-0 |
| Brighton v Leeds | 1.88 – 1 | 58% | 22% | 20% | 1-1 |
| Fulham v Crystal Palace | 1.63 – 1.1 | 50% | 25% | 25% | 1-1 |
| Man City v Coventry | 2.17 – 1.05 | 63% | 20% | 17% | 2-1 |
| Nott'm Forest v Spurs | 1.22 – 1.28 | 35% | 27% | 38% | 1-1 |
| Hull v Aston Villa | 1.57 – 1.39 | 42% | 24% | 34% | 1-1 |
| Everton v Man Utd | 1.43 – 1.54 | 35% | 24% | 41% | 1-1 |
| Arsenal v Chelsea | 1.98 – 1.02 | 60% | 21% | 19% | 1-1 |
The full board for every division is on our football predictions page.
Where These Numbers Come From
Every price on this page starts with expected goals, and we work those out rather than ask you to. Each club is rated on the matches it has actually played: goals scored and conceded, split between home and away, measured against its division's average. Before a season starts that means last season's record; once it is under way the current campaign takes over, carrying about half the weight by a dozen games in and most of it by the spring.
That last detail is what lets all four divisions carry a real model. A club that has been promoted or relegated keeps its measured record and is adjusted by one division step, so a side coming down from the Premier League is priced as the strong Championship team it probably is, not as a generic newcomer.
The two figures then go through a Poisson goal model with a low-scoring correction, which gives the probability of every scoreline. Every market on the page is built from that grid. Cards and corners use each club's real per-game averages from the same season, and where a feed genuinely has no corner data: clubs promoted out of the National League, the corners section is left out rather than filled with a guess.
How to Find Value
Value is simple to define and hard to find. It exists when the bookmaker's price is longer than the fair price. Price a match here, then check the real odds. If the bookmaker offers 2.10 where the engine says 1.90, that gap is your edge.
The fair odds on this page carry no margin, which is why they look generous next to a betting slip. A bookmaker builds a cut into every market, so most prices you meet will be shorter than the model's. The ones that are not are the only bets worth writing down.
Work through a full card that way and the value bets surface on their own. This is exactly how our tips are built: a fair price and a market price, compared, with the working shown.
What the Model Does Not Know
A model is a starting point, not an oracle, and being honest about the gaps is the difference between a tool and a fairground trick. It has no idea who is injured, who is being rested before a cup tie, or which manager is about to be sacked. It reads last season's goals, nothing else.
It also assumes goals arrive independently, which is not quite true — teams chase games and sit on leads. The low-scoring correction pulls some of that back, and it is the reason the model is steadier on totals than on exact scorelines. Treat the fair price as the floor of your research, then add what you know that the numbers cannot.
That is also why the expected goals are fixed here. If the figures could be typed over, the fair odds would only be as good as the guess behind them, and the page would tell you nothing about what we actually think. These are our numbers, and we publish them where they can be checked against results.
Frequently Asked Questions
Which matches can I price with it?
Every fixture of the next matchweek in all four English divisions: 46 matches across the Premier League, Championship, League One and League Two. Choose a division, then a match.
The same model drives our football predictions board.
Where do the expected goals come from?
Each club is measured on the matches it has actually played: goals scored and conceded, home and away, against its division's average. Right now that means this season so far, blended with last season; a club promoted or relegated between seasons keeps its real record adjusted by one division step, and that adjustment fades as this season's results replace it.
The full method is on our how we rate page.
Why can I not change the numbers myself?
Because then it would price your guess, not ours. The expected goals are the model's output, so the fair odds you see are the ones we stand behind.
If you want to test how variance treats a price over hundreds of bets, use the Bankroll Simulator.
How do I spot value with it?
Compare the engine's fair odds to the price on offer. If the bookmaker is longer than the fair price, that is value.
Check the live prices at the best football betting sites.
Is it a guaranteed winning system?
No. A model finds edges, it does not remove risk. It has no idea who is injured, who is rotating before a cup tie, or how the wind is blowing.
Bet within limits you have set in advance. See responsible gambling.