How accurate are football predictions? We explain how to judge them fairly, what baseline to beat and how we publish every call, hit or miss.
How Accurate Are Football Predictions? Our Public Record
Quick facts
- Short answer: judge predictions over hundreds of games, not a weekend
- Baseline to beat: always backing the home side
- Our record: every call logged before kick-off, hits and misses shown
- Started: 31 October 2026
- No claims: nothing here is guaranteed
How accurate are football predictions? Honestly, less than most people want, and the only fair way to answer is to measure them over many games against a simple baseline. This page explains how to judge any prediction, including ours, and shows the upcoming games below with our pick for each one. Further down, you can see our public record of every call we have made.
Upcoming games and our predictions
Premier LeagueSun 1 Nov, 16:30
Liverpool
ArsenalArsenal to win
Premier LeagueMon 2 Nov, 20:00
Newcastle
EvertonNewcastle to win
ChampionshipTue 3 Nov, 19:45
Watford
LincolnLincoln to win
ChampionshipTue 3 Nov, 19:45
Preston
Sheffield UtdSheffield Utd to win
ChampionshipTue 3 Nov, 19:45
Blackburn
Stoke CityStoke City to win
ChampionshipTue 3 Nov, 19:45
Derby
NorwichNorwich to win
ChampionshipTue 3 Nov, 19:45
QPR
BurnleyQPR to win
ChampionshipTue 3 Nov, 19:45
Charlton
BoltonBolton to win
ChampionshipTue 3 Nov, 19:45
Portsmouth
SwanseaSwansea to win
ChampionshipTue 3 Nov, 20:00
Birmingham
MillwallBirmingham to win
What does accuracy mean for football predictions?
There are two ways to judge a prediction. The first is the hit rate: how often did the most likely outcome actually happen? The second is calibration: when a model says 60%, does that outcome really happen about 60 times in 100?
However, hit rate hides a lot. A model can be right 50% of the time and still be badly calibrated if it keeps calling games at 80% that only come in at 50%. In the long run, a good model does well on both measures, and it publishes enough games for the numbers to settle.
What is the baseline a prediction has to beat?
Football has three results, so a pure guess is right about one time in three. A better, equally simple method is to back the home side every week. In the Premier League, home sides have historically won somewhere between four in ten and just under half of games, with draws around a quarter and away wins making up the rest.
As a result, a model that cannot beat “always pick the home team” adds nothing. This is the standard we hold ourselves to, and it is the first question to ask of any prediction site: compared with what?
How many games do you need to judge a prediction model?
More than you would think. A short run of results can look impressive or terrible purely through luck. As a rough guide, the margin of error on a hit rate near 50% is about 16 percentage points after 10 games, about 5 points after 100 games and about 2.5 points after a full 380-game Premier League season.
For that reason, when a site boasts about a good weekend or a lucky streak, the right reaction is caution. Sample size matters more than any single result, and it is the reason we show our running total instead of picking our best days.
How do we measure our own predictions?
Before each match on our football predictions today page, our pick, whether home win, draw or away win, is saved. We never edit it afterwards. When the game finishes, we compare it with the final score and add it to the tally.
One detail is worth knowing. Some games are almost level, and the model still has to choose a side. We count that pick like any other, so if the game ends in a draw, it is a miss. That makes the test stricter, not easier.
Our prediction record
We started logging our prediction for every match on this page on 31 October 2026. Results will appear here as games finish, including the misses. Pending: 11.
| Date | Match | Our prediction | Result | |
|---|---|---|---|---|
| 3 Nov | Birmingham v Millwall | Birmingham to win | – | Pending |
| 3 Nov | West Brom v Southampton | Southampton to win | – | Pending |
| 3 Nov | Watford v Lincoln | Lincoln to win | – | Pending |
| 3 Nov | Preston v Sheffield Utd | Sheffield Utd to win | – | Pending |
| 3 Nov | Blackburn v Stoke City | Stoke City to win | – | Pending |
| 3 Nov | Derby v Norwich | Norwich to win | – | Pending |
| 3 Nov | QPR v Burnley | QPR to win | – | Pending |
| 3 Nov | Charlton v Bolton | Bolton to win | – | Pending |
| 3 Nov | Portsmouth v Swansea | Swansea to win | – | Pending |
| 2 Nov | Newcastle v Everton | Newcastle to win | – | Pending |
Our record began on 31 October 2026, so the early numbers are small and will bounce around. We will keep every result on the page, including the misses, and we will say plainly if we fall short of the home-side baseline.
Why are football results so hard to predict?
Because goals are rare, one deflection, red card or goalkeeping error can undo a well-supported prediction. Injuries and suspensions land close to kick-off, and the data often lags behind. Early in a season there is little form to read, and after an international break squads return tired or unsettled.
In practice, even the best models are only moderately better than simple rules. That is not a flaw to hide. It is the nature of a low-scoring game, and the right response is honesty about the odds, not louder claims.
How can you spot a prediction site you should not trust?
Look for four warning signs. First, claims of certainty, such as “100% sure”, “never lose” or “guaranteed”, which no honest site can make. Second, no public record, or a record that only shows wins. Third, old predictions that quietly disappear. Fourth, pressure to pay or sign up for “locked” picks.
A trustworthy source explains its method, shows its misses and does not promise outcomes. It also tells you what it is not: our predictions are for information and entertainment, not advice to bet.
Where can you check the numbers yourself?
If you want to test a prediction against the facts, our Premier League form table shows who is in form, the clean sheets table shows the tightest defences, and our league tables show the wider picture. You can read about our methods on the about page and our editorial policy.
Gamble responsibly
This page is for information and entertainment, for people aged 18 and over. Predictions are not guarantees and are not advice to bet. If you choose to gamble, set a limit you can afford, never chase losses and take breaks. Free, confidential support is available from BeGambleAware at begambleaware.org or on 0808 8020 133, and you can read our responsible gambling page for more. We do not take bets and we do not link to bookmakers from this page.
How accurate are football predictions: frequently asked questions
How accurate are football predictions?
It depends on how you measure. A fair test is whether predictions beat simple baselines, such as always backing the home side, over a large enough number of games. Any single weekend tells you very little.
Are football predictions reliable?
They are useful as a guide to likelihood, not as a promise. A 55% chance of a win still fails to arrive in 45 comparable games out of 100. Reliability shows up over hundreds of games, not a handful.
How many games do you need to judge a prediction model?
At least 100, and preferably several hundred. After 10 games, luck alone can move a hit rate by 15 percentage points either way. After 380 games, the same luck moves it by about 2.5.
Can any site predict football results 100% correctly?
No. Any site claiming certain, guaranteed or never-lose predictions is making a claim that cannot be true. Football has too much randomness, and a site with a real record would publish its misses.
How does English Matchday measure its predictions?
We log the model’s most likely outcome before kick-off, never edit it afterwards, and compare it with the final score once the game finishes. The full list of hits and misses is on our predictions page.
Where does the data come from?
The predictions come from a licensed football data feed that combines recent form, attack and defence strength, goals and head-to-head history. We do not change the numbers by hand.
