Deadline: Friday 28 August, 18:30 BST. Crystal Palace host Man City ninety minutes later.
Last week we projected 8.3 points for our captain. He scored 2. We went through the wreckage on Tuesday; this piece does the other half of the promise — a measured scorecard first, then the model re-picks without flinching. A projection you quietly revise after the fact isn't a projection.
How did the model actually do in Gameweek 1?
We scored every one of the 342 players we published a number for on 18 August against what they went on to score. No filtering, no excuses.
| Ranking by… | Correlation with actual GW1 points |
|---|---|
| Our model | +0.413 |
| Last season's points per game | +0.396 |
| Ownership — what the crowd picked | +0.360 |
| Price | +0.306 |
So the model ranked players slightly better than last season's form, and better than the crowd. "Slightly" is doing real work in that sentence — 0.413 against 0.396 over one gameweek is not a gap you can bank.
The rest of the card is less flattering:
- Average error: 2.05 points per player. In line with the 2.11 our 2025/26 backtest recorded over 31 gameweeks.
- Six of our projected top 20 returned 6+ points. Ten of them scored 2 or fewer. Half our best names blanked.
- Two of the top 20 didn't play at all — Bruno Guimarães and Michael Keane.
And one number that matters more than the headline: over the 231 players who actually got on the pitch, the correlation drops to +0.250 and the error rises to 2.38. The prettier figure is partly the model correctly projecting nearly nothing for players who were never going to feature. Both numbers are true. We're publishing both.
Where it went wrong, by name
| Under-projected | Club | We said | He scored |
|---|---|---|---|
| Maxim De Cuyper | BHA | 2.9 | 17 |
| Jack Hinshelwood | BHA | 4.2 | 16 |
| Anton Stach | LEE | 2.6 | 13 |
| Michael Kayode | BRE | 3.8 | 13 |
| Cole Palmer | CHE | 4.6 | 13 |
| Over-projected | Club | We said | He scored |
|---|---|---|---|
| Bruno Guimarães | ARS | 6.7 | 0 |
| Bruno Fernandes | MUN | 8.3 | 2 |
| Igor Thiago | BRE | 5.9 | 0 |
| Michael Keane | EVE | 5.0 | 0 |
| Erling Haaland | MCI | 6.6 | 2 |
Two of those five aren't football at all. Guimarães and Keane never got on the pitch, and neither did Viktor Gyökeres, our sixth-worst miss. All three were projected on last season's final five appearances, three months out of date. That is exactly the input this week's run repairs — see what changed below. You can watch this gap update daily on Over/underperformers.
The Gameweek 2 fixtures
Difficulty is FPL's own rating, 1 (easiest) to 5 (hardest), from each side's point of view.
| Kick-off (BST) | Home | Away | ||
|---|---|---|---|---|
| Fri 28, 20:00 | Crystal Palace | 4 | Man City | 3 |
| Sat 29, 12:30 | Liverpool | 3 | Nott'm Forest | 4 |
| Sat 29, 15:00 | Bournemouth | 3 | Everton | 3 |
| Sat 29, 15:00 | Coventry | 2 | Hull City | 2 |
| Sat 29, 17:30 | Spurs | 2 | Newcastle | 3 |
| Sun 30, 14:00 | Chelsea | 2 | Brighton | 4 |
| Sun 30, 14:00 | Leeds | 3 | Brentford | 3 |
| Sun 30, 14:00 | Sunderland | 2 | Fulham | 3 |
| Sun 30, 16:30 | Man Utd | 2 | Ipswich | 4 |
| Mon 31, 20:00 | Aston Villa | 4 | Arsenal | 4 |
Two things to plan around. Arsenal don't play until Monday night, so an Arsenal captain can't be reacted to. And Coventry v Hull is the promoted derby — the one fixture on the board our model can barely see at all, for reasons we'll get to.
The full grid, all season, is on Fixtures.
The captain call
Bruno Fernandes. Again. At 9.4 points, which is higher than the 8.3 we gave him last week.
We know how that reads. He blanked, over 139,000 managers sold him net, and the model has come back and asked for more. But nothing about GW1 argued he'd stopped being good — he played all 90 at Hull, Man Utd lost 2-0, and one bad afternoon against a promoted side doesn't re-rate a player who takes the penalties for the third-best attack in our team ratings. What changed is the fixture: Man Utd host Ipswich at home, rated a 2, and that is a bigger swing than last week's trip to Hull.
| Player | Pos | Club | Price | Owned | GW2 projection |
|---|---|---|---|---|---|
| Bruno Fernandes | MID | MUN | £12.0m | 48.8% | 9.4 |
| Bryan Mbeumo | MID | MUN | £8.0m | 36.2% | 7.0 |
| Cole Palmer | MID | CHE | £9.5m | 14.7% | 6.0 |
| Erling Haaland | FWD | MCI | £15.5m | 68.1% | 5.3 |
| Marcus Tavernier | MID | BOU | £6.0m | 2.2% | 5.3 |
| Harry Maguire | DEF | MUN | £5.0m | 17.2% | 5.2 |
The gap to second is 2.4 points, which is larger than the model's own average error. That is about as clear as this thing gets.
Haaland is fourth again. City go to Selhurst Park, rated a 3, and Palace's defence rates eighth in our team model, ahead of Man Utd's. He is still the second-highest projected player in the game over the next eight weeks. This is a call about Friday, not about August.
Highest projected in each position
Not a squad — that's further down. This is who the model wants in your eleven this week.
Goalkeepers
| Player | Club | Price | Owned | GW2 |
|---|---|---|---|---|
| Senne Lammens | MUN | £5.0m | 15.7% | 4.4 |
| Robin Roefs | SUN | £5.0m | 4.0% | 4.2 |
| Caoimhín Kelleher | BRE | £5.0m | 6.4% | 4.0 |
| Đorđe Petrović | BOU | £4.5m | 3.3% | 3.9 |
| Bernd Leno | FUL | £4.5m | 3.0% | 3.8 |
Defenders
| Player | Club | Price | Owned | GW2 |
|---|---|---|---|---|
| Harry Maguire | MUN | £5.0m | 17.2% | 5.2 |
| Gabriel | ARS | £8.0m | 29.3% | 5.2 |
| Virgil van Dijk | LIV | £6.5m | 19.6% | 5.1 |
| Marcos Senesi | TOT | £6.0m | 7.7% | 4.9 |
| Marc Guéhi | MCI | £6.0m | 19.2% | 4.8 |
Midfielders
| Player | Club | Price | Owned | GW2 |
|---|---|---|---|---|
| Bruno Fernandes | MUN | £12.0m | 48.8% | 9.4 |
| Bryan Mbeumo | MUN | £8.0m | 36.2% | 7.0 |
| Cole Palmer | CHE | £9.5m | 14.7% | 6.0 |
| Marcus Tavernier | BOU | £6.0m | 2.2% | 5.3 |
| Dominik Szoboszlai | LIV | £7.0m | 43.0% | 5.3 |
Forwards
| Player | Club | Price | Owned | GW2 |
|---|---|---|---|---|
| Erling Haaland | MCI | £15.5m | 68.1% | 5.3 |
| Igor Thiago | BRE | £8.0m | 16.8% | 5.2 |
| Dominic Calvert-Lewin | LEE | £6.0m | 29.1% | 4.8 |
| João Pedro | CHE | £7.6m | 67.2% | 4.5 |
| Richarlison | TOT | £6.0m | 2.9% | 4.1 |
Man Utd take four of the twenty places above. If Ipswich keep it tight on Sunday afternoon, this week's article ages badly in one go — that's the honest shape of the risk. Sort any of it yourself on Players.
Best value on the board
Projected GW2 points per £1m, for anyone the model has at 3.0 or better.
| Player | Pos | Club | Price | Owned | GW2 | Per £m |
|---|---|---|---|---|---|---|
| Harry Maguire | DEF | MUN | £5.0m | 17.2% | 5.2 | 1.05 |
| Luke Shaw | DEF | MUN | £4.5m | 19.8% | 4.2 | 0.94 |
| Marcus Tavernier | MID | BOU | £6.0m | 2.2% | 5.3 | 0.88 |
| Bryan Mbeumo | MID | MUN | £8.0m | 36.2% | 7.0 | 0.88 |
| Malick Thiaw | DEF | NEW | £5.0m | 1.7% | 4.4 | 0.87 |
| Senne Lammens | GKP | MUN | £5.0m | 15.7% | 4.4 | 0.87 |
| James Justin | DEF | LEE | £4.5m | 1.1% | 3.8 | 0.85 |
Maguire at £5.0m is the standout, and he is the cheap end of the same bet as the captain: Man Utd, at home, to a promoted side.
Differentials
Under 5% owned, ranked over the next eight gameweeks rather than one. A differential bought for a single week is a coin toss.
| Player | Pos | Club | Price | Owned | GW2 | GW2–9 |
|---|---|---|---|---|---|---|
| Marcus Tavernier | MID | BOU | £6.0m | 2.2% | 5.3 | 41.2 |
| Enzo Fernández | MID | CHE | £7.0m | 4.7% | 4.9 | 39.7 |
| Malick Thiaw | DEF | NEW | £5.0m | 1.7% | 4.4 | 38.5 |
| Phil Foden | MID | MCI | £7.0m | 4.1% | 4.2 | 37.2 |
| Dango Ouattara | MID | BRE | £6.5m | 1.1% | 4.3 | 36.0 |
| Sven Botman | DEF | NEW | £5.0m | 0.5% | 4.1 | 36.0 |
| Kevin Schade | MID | BRE | £6.0m | 3.5% | 4.1 | 35.1 |
Tavernier has now topped this table two weeks running, still under 3% owned. Two Newcastle defenders make it on fixtures alone — see who else walks into a kind run on Fixture Swings. More differentials, updated daily, on Differentials, and the spread of outcomes behind each one on Point Potential.
The full £100m squad
This is the model's answer to "if you were starting from scratch on Friday, what would you own?" It's a solver working the real rules — fifteen players, 2/5/5/3, £100.0m, maximum three per club — picking the squad with the best starting eleven across Gameweeks 2 to 9, captain doubled.
£100.0m exactly. 70.9 projected points for the GW2 eleven, captain included. Formation 3-5-2.
| Pos | Player | Club | Price | Owned | GW2 | GW2–9 |
|---|---|---|---|---|---|---|
| GKP | Caoimhín Kelleher | BRE | £5.0m | 6.4% | 4.0 | 32.3 |
| DEF | Gabriel | ARS | £8.0m | 29.3% | 5.2 | 44.1 |
| DEF | Marcos Senesi | TOT | £6.0m | 7.7% | 4.9 | 38.2 |
| DEF | Marc Guéhi | MCI | £6.0m | 19.2% | 4.8 | 40.4 |
| MID | Bruno Fernandes (C) | MUN | £12.0m | 48.8% | 9.4 | 59.8 |
| MID | Bryan Mbeumo (V) | MUN | £8.0m | 36.2% | 7.0 | 44.1 |
| MID | Cole Palmer | CHE | £9.5m | 14.7% | 6.0 | 48.7 |
| MID | Marcus Tavernier | BOU | £6.0m | 2.2% | 5.3 | 41.2 |
| MID | Enzo Fernández | CHE | £7.0m | 4.7% | 4.9 | 39.7 |
| FWD | Igor Thiago | BRE | £8.0m | 16.8% | 5.2 | 43.8 |
| FWD | Dominic Calvert-Lewin | LEE | £6.0m | 29.1% | 4.8 | 35.1 |
Bench, in order: Malick Thiaw (NEW, £5.0m), Luke O'Nien (SUN, £4.0m), Liam Delap (CHE, £5.5m), and Martin Dúbravka (TOT, £4.0m) as the second keeper.
Thiaw projects 4.4 — higher than the goalkeeper — and still starts on the bench. That's the formation doing it, not a mistake: with five midfielders and two forwards there is only room for three defenders, and dropping Calvert-Lewin for him loses points.
Seven players in that fifteen take penalties, free-kicks or corners for their club. Set-piece duty is a large slice of why the solver likes them, and you can check every taker on Set Piece Takers before you commit.
Does the horizon change the answer?
Ask the same solver for the best squad for Gameweek 2 alone and it shares 10 of 15 with the eight-week squad. The five it swaps in are Roefs, Maguire, van Dijk, Szoboszlai and a cheaper third forward — all one-week fixture plays.
Ten of fifteen surviving a change of planning horizon is the useful signal here. Those ten are the part of this squad we'd defend. The other five are inside the noise, and we've written before about how far ahead planning actually pays.
Compared with the squad we published for Gameweek 1, eight of fifteen are unchanged. Of the seven players that article named as its defensible core, five survive — Guimarães is out on an injury flag and Roefs lost the goalkeeper slot. Build your own version on the Team Planner.
What changed in the model this week
Three things, and one of them was a bug we found while writing this.
1. The model can finally see this season's minutes. Every player's expected minutes now include Gameweek 1, not just last season's final five games. This was the single weakest input a week ago and it's the main reason the squad moved.
2. A non-appearance is data, and we weren't recording it. The stats feed only publishes a row for players who got on the pitch. So a player dropped from the squad entirely had no row — and kept last season's minutes untouched. The players whose minutes most needed revising were precisely the ones the update couldn't reach. First run of the week put a third-choice goalkeeper at 0.0% ownership in the projected top two. We now write a zero-minute row for every player whose club played and who didn't feature — 304 of them in GW1 — and he dropped straight out.
3. Team ratings barely moved, on purpose. One round of matches now feeds the team model alongside last season's 380, but weighted so a single game is worth about a thirty-ninth of a season. Hull beating Man Utd was the biggest single move in the league and it lifted Hull's defence rating by 0.04. The promoted three still rate 18th, 19th and 20th. If ten matches could reorder the league, the model would be chasing noise every week.
All three are recorded in the model ledger with the expectation we wrote before running it — including the one part of that expectation that turned out wrong. If you want the architecture rather than the week's diff, here is how the model works.
Who the model can't see
This is the honest hole, and it's a big one. Our projections need a real per-90 record — three full matches' worth — before we'll publish a number. Invent one and you're just guessing with extra steps. But that means four of Gameweek 1's double-figure scorers are invisible to us:
| Player | Club | Pos | Price | Owned | GW1 pts |
|---|---|---|---|---|---|
| Nobel Mendy | HUL | DEF | £4.0m | 3.1% | 15 |
| Semi Ajayi | HUL | DEF | £4.0m | 1.6% | 14 |
| Mamadou Sangaré | BRE | MID | £5.6m | 7.6% | 14 |
| Konstantinos Tzolakis | HUL | GKP | £4.5m | 2.4% | 10 |
So is Christos Tzolis at Arsenal, owned by more than a quarter of the game.
It hits the promoted clubs hardest, because most of their squads have never played in the division. A handful of their players do have a record and are projected normally — Issa Diop is on our own bench — but Saturday's Coventry v Hull is largely a fixture we have nothing useful to say about. It fixes itself as the season plays: Sangaré qualifies after three more starts. Until then, treat silence from us about a promoted club as ignorance, not a verdict.
Who's flagged
Fifteen players owned by 1% or more carry a flag going into the deadline. These are the seven that most managers own.
| Player | Club | Owned | Status |
|---|---|---|---|
| Morgan Gibbs-White | NFO | 9.2% | Knee — 75% |
| Moisés Caicedo | CHE | 3.2% | Unspecified — 75% |
| Bruno Guimarães | ARS | 3.0% | Thigh — 75% |
| Ismaïla Sarr | CRY | 3.0% | Groin — out |
| Dan Burn | NEW | 1.8% | Ankle — back 14 Sep |
| Jérémy Doku | MCI | 1.4% | Calf — back 5 Sep |
| Will Osula | NEW | 1.1% | Foot — 25% |
Three notes the flag itself can't give you, all from Thursday's press conferences. Arsenal described Guimarães as "potentially available" after managing his load through the week, so that 75% could move either way. Ezri Konsa is fit and training despite being one of the week's most-sold defenders. And Hull said the players who came off in Gameweek 1 had cramp, not injuries — which sits awkwardly against Semi Ajayi carrying a 75% hamstring flag in the API right now. When the press conference and the flag disagree, the flag is usually the older of the two. Source: Fantasy Football Scout's Thursday team news.
What this is, and what it isn't
The same caveat as last week, updated once and not repeated. This is the model's own machinery run at a point in the season where its normal backtest can't run — that gate needs eight played gameweeks and there is one. It has not passed our evaluation gate. The published accuracy record is still the 2025/26 walk-forward: a rank correlation of 0.359 against 0.268 for a points-per-game baseline, beating that baseline in 31 gameweeks out of 31 tested.
One gameweek is one gameweek. The scorecard at the top of this page is a data point, not a trend, and we'll have thirty-seven more of them before it means anything.
Next: Gameweek 3 picks, and the GW2 scorecard that goes with them. In the meantime — Players for the full database, Fixtures for the difficulty grid, Team Planner to build your fifteen, and My Team to check your own score against this.