Deadline: Saturday 12 September, 13:30 BST. Five games kick off ninety minutes later. Nothing is played before it, so there's no Friday night game to watch first.
The short version:
- Captain Cole Palmer. We project 7.29. Chelsea are home to Hull City, the worst defence on the board.
- Erling Haaland projects 6.85, and 71.2% of managers own him. The gap is 0.44 points. Our average error is 2.1, so either pick is fine.
- Our two squads disagree about the captain for the first time this season. They share 7 players out of 15, so this week is about fixtures rather than form.
Every projection here comes from our model run on Monday 8 September, 16:30 BST, looking at Gameweeks 4 to 11.
Who should I captain in Gameweek 4?
Cole Palmer, 7.29 projected. That's the highest single-week number we have for anyone in Gameweek 4. Chelsea are home to Hull City, who rate 18th of 20 in our team model, and the game is in the Saturday 15:00 block.
Erling Haaland projects 6.85. Manchester City are away at Manchester United, rated difficulty 4 for both sides. It's the hardest fixture of the week.
The gap is 0.44 points. Our average error is 2.1 points per player per gameweek. The gap is inside our own noise. Our model has stopped captaining Haaland, but it hasn't made a case worth taking a hit over.
So: if you own Haaland, captain him. At 71.2% ownership that's the safe pick whatever our number says. If you don't own him, take Palmer.
Manchester United v Manchester City kicks off Sunday 13 September, 16:30 BST. Captain anyone in that game and you're locked in for a day and a half after the deadline.
What is the best £100m squad for Gameweek 4?
Fifteen players, real rules: 2/5/5/3, £100.0m, three per club maximum. Best starting eleven across Gameweeks 4 to 11.
£100.0m exactly. 64.5 projected points for the Gameweek 4 eleven, captain doubled. 3-5-2.
| Pos | Player | Club | Price | Owned | GW4 projection | GW4–11 projection |
|---|---|---|---|---|---|---|
| GKP | Konstantinos Tzolakis | HUL | £4.6m | 9.1% | 4.2 | 35.2 |
| DEF | Maxim De Cuyper | BHA | £4.8m | 19.5% | 4.9 | 32.7 |
| DEF | Riccardo Calafiori | ARS | £5.7m | 47.5% | 4.7 | 38.8 |
| DEF | Silva | BOU | £5.0m | 0.4% | 4.4 | 34.9 |
| MID | Bruno Fernandes (V) | MUN | £12.0m | 46.6% | 6.6 | 61.3 |
| MID | Marcus Tavernier | BOU | £6.0m | 3.1% | 5.5 | 43.3 |
| MID | Bryan Mbeumo | MUN | £7.9m | 25.7% | 5.5 | 51.4 |
| MID | Iliman Ndiaye | MCI | £5.9m | 10.3% | 5.0 | 43.9 |
| MID | Elliot Anderson | MCI | £6.3m | 4.7% | 4.8 | 40.7 |
| FWD | Erling Haaland (C) | MCI | £15.5m | 71.2% | 6.8 | 59.7 |
| FWD | Igor Thiago | BRE | £7.9m | 12.1% | 5.1 | 42.9 |
Bench in order: Martin Dubravka (TOT, £4.0m), Malick Thiaw (NEW, £5.0m), Dara O'Shea (IPS, £4.0m), Joshua Zirkzee (MUN, £5.4m).
Konstantinos Tzolakis at £4.6m is our highest-projected keeper, and he plays Chelsea away on Saturday. He's also the third-highest scorer in the game with 26 points. The model is buying his next eight fixtures, not this one. It's the pick most likely to look silly by Saturday evening.
Two Manchester United players sit in a squad that captains a Manchester City player. Bruno Fernandes, Bryan Mbeumo and Erling Haaland are all in the same game. With the captain doubled, that one game carries 25.8 of the eleven's 64.5 points. Forty per cent of the week rides on Sunday afternoon.
Build the same squad for Gameweek 4 alone and only 7 of the 15 players stay. That's the lowest overlap of the season, and it means the week is driven by fixtures rather than form. The one-week squad captains Palmer, which is where the disagreement at the top of this piece comes from.
Build your own fifteen on the Team Planner.
What do the Gameweek 4 fixtures look like?
Difficulty is FPL's own rating, 1 (easiest) to 5 (hardest), from each side's point of view.
| Kick-off (BST) | Home | Away | ||
|---|---|---|---|---|
| Sat 12, 15:00 | Bournemouth | 3 | Brentford | 3 |
| Sat 12, 15:00 | Chelsea | 2 | Hull City | 4 |
| Sat 12, 15:00 | Crystal Palace | 2 | Ipswich Town | 3 |
| Sat 12, 15:00 | Liverpool | 2 | Fulham | 4 |
| Sat 12, 15:00 | Aston Villa | 3 | Nott'm Forest | 4 |
| Sat 12, 17:30 | Spurs | 3 | Everton | 3 |
| Sat 12, 20:00 | Sunderland | 4 | Arsenal | 3 |
| Sun 13, 14:00 | Coventry City | 2 | Brighton | 2 |
| Sun 13, 16:30 | Man Utd | 4 | Man City | 4 |
| Mon 14, 20:00 | Leeds | 2 | Newcastle | 3 |
The deadline is Saturday lunchtime and the week ends Monday night. The two games most likely to decide it land last: Manchester United v Manchester City, then Leeds v Newcastle.
The full season grid is on Fixtures.
Which players can our model not see yet?
289 players have a projection from us. 295 don't, and 100 of those have already played this season.
We need 270 minutes of Premier League football, about three games, before we publish a number on anyone. Make one up and you're just guessing. These are the highest scorers we're leaving blank:
| Player | Club | Pos | Price | Owned | Minutes | Points so far |
|---|---|---|---|---|---|---|
| Semi Ajayi | HUL | DEF | £4.1m | 11.0% | 243 | 25 |
| Nobel Mendy | HUL | DEF | £4.0m | 7.0% | 191 | 24 |
| Ryan Giles | HUL | DEF | £4.0m | 1.3% | 264 | 20 |
| Mamadou Sangaré | BRE | MID | £5.7m | 14.2% | 210 | 18 |
| Bobby Thomas | COV | DEF | £4.0m | 8.7% | 257 | 15 |
Three of those five are Hull City defenders, and our highest-projected keeper is the Hull City keeper. We can see Tzolakis and not the back four in front of him, only because he's past 270 minutes and they aren't.
It sorts itself out as the season goes. Giles qualifies after one more start. If we're quiet about a promoted club, that means we don't know yet, not that the player is bad.
How did the model do in Gameweek 3?
We beat the crowd, by too little to celebrate.
Here is every one of the 272 players we published a number for before the Gameweek 3 deadline, with no filters:
| Ranking by… | GW1 | GW2 | GW3 |
|---|---|---|---|
| Our model | +0.413 | +0.376 | +0.426 |
| Ownership — what the crowd picked | +0.360 | +0.394 | +0.414 |
| Average error, points per player | 2.05 | 1.99 | 2.02 |
| Hauls (6+ points) in our projected top 20 | 6/20 | 9/20 | 7/20 |
Three gameweeks in, we lead the crowd 2-1, by an average of 0.02. We called Gameweek 2's 0.018 gap a loss in bold. Gameweek 3's 0.012 gap our way gets the same treatment, which is almost none.
Our Gameweek 3 captain call was Haaland at 7.9, and he returned 9. That was the best top-20 call of the week. Our worst miss was Phil Foden, 6.8 projected, 1 scored. His ownership has since dropped to 4.8%, and he is still the third-best differential in our Gameweek 4–11 table. So we and the crowd now disagree about him in public, and Sunday settles it. Biggest under-projection: Jayden Bogle, 1.7 projected, 14 scored.
Tell us where this is wrong
- Show us your Gameweek 4 draft and why. Build it on the Team Planner, share the link, and name the one pick you'd defend against the model. If you're captaining Haaland into Old Trafford, tell us what you can see that a 0.44-point gap can't.
- Tell us when a number looks wrong. Not "the model is bad". This player, this figure, here's why. That's the most valuable message we can get, and we found a bug in our own pipeline that proves it.
- Tell us what you'd want to watch. The Watch button on every player page saves a player's price and news, then tells you what changed next time you look. If it should flag a fixture swing or a set-piece change instead, say so.
Send it through the feedback form for anything specific, or join the Discord if it needs a conversation. Both get read every day.
Next: Gameweek 5 picks and the Gameweek 4 scorecard. Until then: Players for every projection, Fixtures for the difficulty grid, Team Planner to build your fifteen, and My Team to score your squad against ours.