Premier League Predictions: What 4 Weeks Taught Me
The smartest Premier League prediction this weekend is the one that maximises expected points, not the one that sounds boldest. BBC Sport scores a correct result at 10 points and an exact score at 40,...
Premier League Predictions: What 4 Weeks Taught Me
The smartest Premier League prediction this weekend is the one that maximises expected points, not the one that sounds boldest. BBC Sport scores a correct result at 10 points and an exact score at 40, so the modal scoreline, usually 1-0 or 1-1 rather than 2-0, tends to beat the glamorous pick. Manchester United host Fulham on Sunday under mounting pressure on Michael Carrick, while Manchester City arrive with a 100% start under Enzo Maresca. My simple Poisson model gives Arsenal about a 63% chance against Leeds United on 10 October and puts a 1-0 pick marginally ahead of 2-0 on expected value. Draw-heavy slates such as Aston Villa against Brentford carry roughly a 26% draw probability. Before locking anything in, convert each pick into expected points, compare it with the consensus scoreline, and never stake money you cannot afford to lose. World Cup Hub publishes these workings so you can check them.
Picture two fans on a Friday night. One announces that Manchester City will "obviously" win 3-0 and treats the matter as closed. The other writes down a win probability, a most-likely scoreline and a number of points that pick should earn, then goes to bed. After four weeks of watching published slates, including the week-four contest between BBC Sport's Chris Sutton and guitarist and City supporter Johnny Marr, I can tell you which of the two ends the season higher. It is rarely the confident one.
Here at World Cup Hub, we cover the 2026 World Cup, yet the Premier League is where the habits get trained. The weekly scoreline is a small, repeatable probability problem, and it punishes sentiment with remarkable consistency.
The Bottom Line
Pick the most probable scoreline in the game you are most sure of, and pick the likelier result everywhere else. The "official" scoring on BBC Sport's predictor gives 10 points for the right outcome and 40 for the exact score. That ratio, four to one, is the whole game. Any scoreline you choose needs only about a 25% chance of landing exactly before it beats a pure result pick, and almost no single scoreline in football comes close to 25%.
So the real contest is between the "bold" pick and the modal pick. Pundits reach for 2-0 and 3-1 because they read like conviction. The model reaches for 1-0 and 1-1 because that is where the probability mass sits. A first-hand observation from reading the early-season slates: the scorelines that fill a prediction table are almost always one goal richer than the most frequent real results. That gap is the edge, and it is small but persistent.
For this weekend, the headline fixture is Manchester United against Fulham on Sunday. Reported by BBC Sport, United face a three-week international break after it, and Chris Sutton warned that "any negativity is going to just fester" if results go wrong. Pressure moves headlines. It does not move a Poisson rate by much, and that is a point my apprentices never accept.
What Do Players Actually See?
Most players see a column of confident scorelines from pundits and tipsters, one per fixture, with little or no probability attached. The BBC lists predictions for all 380 games, and footballpredictions.com shows a single score per match, such as Arsenal 2-0 Leeds United. The number you see is a headline. The probability behind it stays hidden.
That omission matters. Footballpredictions.com lists Arsenal 2-0 Leeds United and Aston Villa 1-1 Brentford for 10 October, kick-offs at 19:30 and 22:00 respectively as shown on the page. Two predictions, two very different risk profiles, and the page presents them with identical visual weight and a "BET NOW" button beside each. A reader would never guess that one is a favourite-wins pick near 63% while the other is a draw call near 26%.
I want to be precise about my own numbers here. I assumed expected goals of 1.9 for Arsenal and 0.8 for Leeds, and 1.4 for Villa and 1.2 for Brentford. These are my illustrative inputs, not bookmaker prices. Under that Poisson distribution, 2-0 for Arsenal has about a 12.1% chance and 1-0 about 12.8%. Villa-Brentford 1-1 comes out near 12.5%. Notice that the "boring" pick is the better pick, which is exactly what the headline format hides.
Which 3 Things Matter Most?
Three things matter most: the points ratio, the shape of the scoreline distribution and the quality of your goal-rate inputs. Together they decide whether a pick is worth making, and each can be checked before kick-off in under five minutes.
Let us take them in order, because the order is the lesson.
- The points ratio. With 10 points for a result and 40 for an exact score, expected value is roughly 10 times the result probability plus 30 times the scoreline probability. For Arsenal against Leeds, a 1-0 pick returns about 10.2 points and a 2-0 pick about 10.0. The difference is two tenths of a point per game, but across 380 fixtures that is real money in a leaderboard.
- The shape of the distribution. Football scorelines are concentrated at low totals. A game with 2.7 expected goals still produces 1-0, 1-1 and 2-1 more often than anything involving three goals for one side. Chasing a "statement" scoreline is paying a premium for variance.
- The inputs. The model is only as good as its goal rates. Manchester City's 100% start under Enzo Maresca and their win at Old Trafford, where Phil Foden was sent off early, tell you the side can grind. They do not tell you City should be given 3.0 expected goals every week. Small samples are the main way a careful model goes wrong.
Taken together, these explain why the same person can look brilliant for a month and mediocre for a season. Variance is loud, edge is quiet.
How Do You Turn a Prediction Into Expected Value?
Multiply the probability of the correct result by 10, then add 30 times the probability of your exact scoreline. The sum is the average points that pick earns per game. Choose the scoreline that maximises it, which is nearly always the most likely scoreline on the favoured side.
Here is the worked example for Aston Villa against Brentford, using my assumed rates. A draw happens about 26.2% of the time, and the 1-1 scoreline about 12.5%. A 1-1 pick therefore returns roughly 0.262 times 10 plus 0.125 times 30, which is about 6.4 points. That sounds weak beside the 10-point figure for the Arsenal pick, and it is.
The conclusion is a contrarian one, backed by arithmetic: unless you believe a game is a draw with probability well above 35%, do not predict the draw. Pick the marginally likelier winner with a 1-0 or 2-1 scoreline instead. A draw pick pays only when the draw genuinely dominates, and it almost never does. In a typical Premier League weekend the draw is the single most likely outcome in only a handful of games, so a slate full of 1-1 calls is a slate full of overpriced optimism.
What Edge Cases and Gotchas Should You Know?
The main gotchas are red cards, manager pressure narratives, unusual kick-off times and consensus scoring. Each distorts a prediction in a way a raw scoreline model cannot see, so you adjust for them by hand, sparingly, and only when the evidence is concrete.
Start with red cards. Phil Foden's early dismissal at Old Trafford is the textbook case: BBC Sport's report quotes Johnny Marr saying "we all feared the worst" when it happened, yet City won. One match is an anecdote, not a rate. A single early red card shifts the expected goals sharply, but your pre-match pick cannot price it, so do not rewrite a model because of a story.
Next, the manager narrative. Everyone has an opinion on whether Michael Carrick is safe, and whether United will "bounce back" against Fulham. Be suspicious of that. A change in mood is not a change in shot quality.
Finally, a subtle one: BBC Sport uses the most popular reader scoreline for its scoreboards and tables. If thousands of readers all choose 2-0, the crowd's pick is a poor guide to what is likely. Consensus is where you measure yourself, not where you find the edge. And as ever with anything involving money, keep stakes small, set limits in advance, and treat every prediction as a probability rather than a promise. Betting is for adults aged 18 or over only.
Verdict
My verdict after four weeks is plain, and I am afraid it will disappoint the dramatic among you. Do not predict what would make a good story. Predict the most probable scoreline, favour the result with the higher probability and treat every draw pick with suspicion. For this weekend, that means respecting Manchester City's momentum under Enzo Maresca, being cautious about overreacting to the pressure on Michael Carrick at Manchester United, and picking Fulham's chances by numbers rather than newspaper mood.
For readers who want the next step, our [Internal Link: expected goals explained for beginners] guide covers the inputs, and our [Internal Link: weekly Premier League scoreline probability tables] page applies the method to each fixture. If you are building toward the summer, [Internal Link: how Premier League form translates to World Cup predictions] is where World Cup Hub connects the two. The method is dull. That is why it works.
Frequently Asked Questions
Q: What is the best Premier League prediction method for this weekend?
A: The best method is to pick the most probable scoreline for the team more likely to win. BBC Sport awards 10 points for the correct result and 40 for the exact score, so the modal scoreline, often 1-0 or 2-1, maximises average points. Avoid bold 3-0 calls unless you have strong evidence of a mismatch.
Q: How many points do you get for a correct Premier League prediction on BBC Sport?
A: A correct result earns 10 points and an exact score earns 40 points. That means a perfect scoreline is worth four times as much as simply getting the winner right. Because exact scores are rare, the 10-point result is the foundation of almost every strong season total.
Q: Is picking draws a good idea?
A: Usually not, unless the draw probability is well above 35%. In my illustrative Aston Villa against Brentford model, a draw came to about 26%, and a 1-1 pick returned roughly 6.4 expected points versus about 10 for a likely-winner pick. Draws look safe but are consistently the weaker bet.
Q: Why might 1-0 be a better pick than 2-0?
A: Because 1-0 is often the single most likely scoreline for a moderate favourite. With Arsenal at 1.9 expected goals and Leeds United at 0.8, my model gives 1-0 about 12.8% and 2-0 about 12.1%. The gap is small, but it is the kind of gap that decides a leaderboard.
Q: What should I do if my predictions keep going wrong?
A: Check whether the problem is variance or method before changing anything. A single bad weekend means little, because even a good model misses most exact scores. If your result picks land below roughly 50% over 30 or more games, review your goal-rate inputs, and stop overreacting to narratives such as manager pressure or one red card.
Q: Is it free to make Premier League predictions?
A: Yes, entering a prediction game is free on BBC Sport's platform, and models like the one above need only a spreadsheet. Costs arise only if you stake money with a bookmaker. If you do, set a fixed budget first, keep it small, and remember that betting is for adults aged 18 and over.
Thank you for reading.
World Cup Hub · Editorial Archive