The Premier League has paused for the international break after five matchweeks, and returns on Saturday 10 October 2026 with Matchweek 6. Manchester City have a perfect record, Tottenham are bottom without a win, and Liverpool v Manchester City at Anfield is the pick of the fixtures when the league resumes.
It's the ideal moment to look at how an AI prediction model actually arrives at a correct score, and why five matches of data can be dangerously misleading. Every figure below comes from the official 2026-27 and 2025-26 Premier League results; the calculations are shown so you can redo them yourself.
The Premier League table after Matchweek 5
| Pos | Team | P | W | D | L | GF | GA | GD | Pts |
|---|---|---|---|---|---|---|---|---|---|
| 1 | Manchester City | 5 | 5 | 0 | 0 | 13 | 5 | +8 | 15 |
| 2 | Arsenal | 5 | 4 | 0 | 1 | 8 | 4 | +4 | 12 |
| 3 | Brighton | 5 | 3 | 1 | 1 | 16 | 5 | +11 | 10 |
| 4 | Brentford | 5 | 2 | 3 | 0 | 10 | 4 | +6 | 9 |
| 5 | Leeds | 5 | 2 | 3 | 0 | 7 | 3 | +4 | 9 |
| 6 | Liverpool | 5 | 2 | 3 | 0 | 7 | 4 | +3 | 9 |
| 7 | Everton | 5 | 2 | 3 | 0 | 6 | 3 | +3 | 9 |
| 8 | Hull City | 5 | 2 | 2 | 1 | 6 | 4 | +2 | 8 |
| 9 | Newcastle | 5 | 2 | 2 | 1 | 9 | 9 | 0 | 8 |
| 10 | Chelsea | 5 | 2 | 1 | 2 | 10 | 12 | −2 | 7 |
| 11 | Ipswich | 5 | 2 | 0 | 3 | 7 | 11 | −4 | 6 |
| 12 | Manchester United | 5 | 1 | 2 | 2 | 8 | 8 | 0 | 5 |
| 13 | Nottingham Forest | 5 | 1 | 2 | 2 | 4 | 5 | −1 | 5 |
| 14 | Sunderland | 5 | 1 | 1 | 3 | 6 | 10 | −4 | 4 |
| 15 | Crystal Palace | 5 | 1 | 1 | 3 | 6 | 11 | −5 | 4 |
| 16 | Aston Villa | 5 | 1 | 1 | 3 | 4 | 9 | −5 | 4 |
| 17 | Bournemouth | 5 | 0 | 3 | 2 | 6 | 8 | −2 | 3 |
| 18 | Coventry | 5 | 1 | 0 | 4 | 1 | 10 | −9 | 3 |
| 19 | Fulham | 5 | 0 | 2 | 3 | 5 | 8 | −3 | 2 |
| 20 | Tottenham | 5 | 0 | 2 | 3 | 2 | 8 | −6 | 2 |
What stands out:
- Manchester City: five wins out of five, including 1-0 at Old Trafford in the derby and 5-3 against Sunderland in Matchweek 5.
- Arsenal, champions last season with 85 points, won their first four before losing 3-0 at Brighton.
- Brighton have the best attack (16 goals), helped by a 5-0 win at Coventry.
- Leeds and Everton have the joint-best defence (3 conceded) and, like Liverpool and Brentford, are still unbeaten.
- Tottenham have scored only 2 goals in 5 matches and are yet to win.
- The promoted sides are split: Hull City are 8th, Ipswich 11th, while Coventry have scored just once.
Across the 50 matches played so far, there have been 141 goals (2.82 per match), 18 home wins, 16 draws and 16 away wins. The most frequent scorelines are 0-1 and 2-2 (six times each), then 1-1 and 0-0 (five times each). Keep that last point in mind: draws and tight scores are everywhere.
Matchweek 6 fixtures (10-12 October 2026)
UK time (BST) is the same as Lagos time; add 2 hours for Nairobi, Kampala or Dar es Salaam, and 1 hour for Johannesburg. Accra is one hour behind the UK. Check kick-off times before each match, as broadcast picks can move them.
| Date | UK / Lagos | Nairobi | Fixture |
|---|---|---|---|
| Sat 10 Oct | 12:30 | 14:30 | Arsenal v Leeds |
| Sat 10 Oct | 15:00 | 17:00 | Chelsea v Bournemouth |
| Sat 10 Oct | 15:00 | 17:00 | Ipswich v Fulham |
| Sat 10 Oct | 15:00 | 17:00 | Aston Villa v Brentford |
| Sat 10 Oct | 15:00 | 17:00 | Sunderland v Brighton |
| Sat 10 Oct | 17:30 | 19:30 | Manchester United v Tottenham |
| Sun 11 Oct | 14:00 | 16:00 | Crystal Palace v Nottingham Forest |
| Sun 11 Oct | 14:00 | 16:00 | Hull City v Everton |
| Sun 11 Oct | 16:30 | 18:30 | Liverpool v Manchester City |
| Mon 12 Oct | 20:00 | 22:00 | Coventry v Newcastle |
How AI calculates a correct score, step by step
A serious prediction model doesn't "guess" a score. It estimates how many goals each team should score, then works out the probability of every possible scoreline. Here's the logic, applied to Liverpool v Manchester City, using only this season's goals.
Step 1: measure each team against the league average
So far this season, a Premier League team scores 1.41 goals per match on average (141 goals ÷ 100 team-matches). Home sides average 1.52, away sides 1.30.
- Liverpool: 7 scored and 4 conceded in 5 matches, i.e. 1.40 and 0.80 per game. Attack index = 1.40 ÷ 1.41 ≈ 0.99; defence index = 0.80 ÷ 1.41 ≈ 0.57 (lower is better).
- Manchester City: 13 scored and 5 conceded, i.e. 2.60 and 1.00 per game. Attack index ≈ 1.84; defence index ≈ 0.71.
Step 2: work out the match's expected goals
Cross each attack with the opposing defence, then apply the home or away average:
- Liverpool = 0.99 × 0.71 × 1.52 ≈ 1.07 expected goals
- Manchester City = 1.84 × 0.57 × 1.30 ≈ 1.36 expected goals
Step 3: turn expected goals into scorelines
The Poisson distribution converts each expected-goals figure into the probability of scoring 0, 1, 2, 3 goals and so on. Multiply the two teams' probabilities and you get every scoreline. Adding them up by outcome gives the 1X2:
- Liverpool win 29.5%, draw 27.1%, City win 43.4%.
- Most likely scores: 1-1 (12.8%), 0-1 (12.0%), 1-0 (9.4%).
That's the core of it. The full method, with a complete scoreline grid, is in our guide on how to predict a correct score.
Why five matches aren't enough
Run the same raw calculation on other Matchweek 6 fixtures and it starts producing nonsense. Five matches is a tiny sample: one big win or one bad afternoon distorts everything.
A standard fix is to blend this season's numbers with a longer history. In the table below, each team's goals per game this season are mixed with its 2025-26 Premier League averages, with the 38 old matches counting as the equivalent of ten. It's a simple illustration, not our production model.
| Fixture | Raw: 5 games only (H / D / A) | Blended with 2025-26 (H / D / A) | Most likely score (blended) |
|---|---|---|---|
| Liverpool v Man City | 29.5% / 27.1% / 43.4% | 25.0% / 23.7% / 51.3% | 1-1 (11.2%) |
| Sunderland v Brighton | 3.6% / 6.4% / 90.0% | 18.6% / 21.6% / 59.8% | 0-1 (10.3%) |
| Aston Villa v Brentford | 6.0% / 14.7% / 79.3% | 27.5% / 25.2% / 47.3% | 1-1 (12.0%) |
The raw version gives Brighton a 90% chance of winning at Sunderland, and its most likely score is 0-4. No sensible model would say that about a Premier League away game: Brighton's 16 goals include a 5-0 at Coventry, and five matches can't separate talent from a hot streak. Blend in last season and Brighton remain clear favourites, at about 60%, with a much more ordinary scoreline on top. The same happens with Aston Villa, fourth last season but 16th after five games: the raw numbers write them off at home, the blend puts the match back in balance.
Notice too that after blending, the most likely score in all three games carries only 10 to 12%. That's the normal range for a football match, and the reason no honest tool sells a "sure" correct score (more on that in sure correct score predictions: do they really exist?).
What a real prediction model adds
The calculation above uses goals only. Elofoot's model starts from the same principle but feeds it better inputs. For each match, it first estimates each team's expected goals from its xG, adjusted for recent form, league position, home advantage and the strength of its division. Those two averages feed a Poisson distribution that gives the probability of every scoreline, then the 1X2, over/under and both-teams-to-score markets. Working from xG rather than raw goals already smooths out a lot of luck, and expected goals are kept within realistic bounds.
Then comes the part people call "AI": a language model writes the scenario of the match in plain English, based on those numbers. It never produces the numbers itself. If a data point is missing, it isn't invented. The whole process is explained in our methodology and on AI football match analysis.
Model predictions for every Matchweek 6 fixture are published on Premier League predictions ahead of kick-off.
Five things to watch in Matchweek 6
These are facts from the table, not predictions:
- Liverpool v Manchester City (Sunday, Anfield): City's perfect start meets a Liverpool side that is unbeaten but has drawn three of five.
- Arsenal v Leeds (Saturday lunchtime): Arsenal's first game since the 3-0 defeat at Brighton, against one of the two meanest defences in the league.
- Manchester United v Tottenham (Saturday evening): Spurs are bottom, winless and have scored twice all season.
- Sunderland v Brighton: the league's top scorers travel to a side that has conceded 10.
- Coventry v Newcastle (Monday night): Coventry have scored once in five matches; Newcastle have conceded 9.
For a full breakdown of any fixture (probabilities, most likely score, likely scorers, scenario), your first full prediction is free, no card needed. Today's matches from every competition are on football predictions today.
Responsible gambling. Elofoot's predictions are statistical estimates, not certainties or betting advice. Betting is for adults only (18+ in the UK and most countries, older in some) and carries real risks: debt, dependency, isolation. Never stake money you need and never chase losses. If gambling stops being fun, talk to someone you trust or a local support service. In Great Britain, the National Gambling Helpline (run by GamCare) is free and open 24/7 on 0808 8020 133.
Frequently asked questions
When does the Premier League return after the international break?
On Saturday 10 October 2026, with Arsenal v Leeds at 12:30 UK time. Matchweek 6 ends on Monday 12 October with Coventry v Newcastle.
Who is top of the Premier League?
After five matchweeks, Manchester City lead with 15 points from five wins, three points ahead of Arsenal. Brighton are third on 10 points.
How does AI predict Premier League correct scores?
A statistical model estimates each team's expected goals for the match (from xG, form, league position and home advantage), then uses the Poisson distribution to calculate the probability of every scoreline. The most likely correct score is the one with the highest probability, typically between 10 and 15% in a Premier League match.
Why are early-season predictions less reliable?
Because five or six matches is a very small sample. One big win can make a team look far stronger than it is. Good models blend current-season data with a longer history and keep extreme values in check.
Where can I find free Premier League predictions?
Every fixture has a free public prediction (1X2 probabilities and most likely score) on Elofoot's Premier League predictions page.
