FPL GW4 Review: The Armband Call We Got Wrong

Gameweek 4 is scored and checked, and the honest headline is not the score. It is the armband.

The Algorithm put three names on its captaincy shortlist. It preferred the one that returned least of the three.

What The Algorithm projected, and what actually happened

The XI is frozen before every deadline so that it can be marked afterwards in public. For Gameweek 4 it was picked before the deadline in a 3-4-3: Raya; Gabriel, James, Lacroix; Palmer, Rogers, Saka, Wirtz; Isak, João Pedro, Haaland. Palmer wore the armband.

It projected 65.5. It scored 77. The game's average was 69.

Eight clear of the average is a good week. Being eleven and a half above your own projection is still a miss, and a miss upwards is worth saying out loud: the floor on that XI was read too low. In a tighter round, reading a floor too low is how you bench the wrong player.

The receipts

Managers agreed with the captain call in 64% of plans. Where they differed, the manager's own pick won 43% of the time and The Algorithm's won 47%, so once you were off the top line the armband was close to a coin toss. On the shortlist itself it was not close. The preferred name returned least.

Transfers read better. Same squads, same week, hits included, the recommended move and the manager's own finished level in 63% of cases, the recommended move won in 22% and the manager's own won in 15%, worth 0.54 points a manager. The best was Tarkowski in for Maguire, a swing of 7. The worst was Tarkowski in for Gvardiol, a swing of minus 3, with Gvardiol returning 11.

Then the number that matters most. The full plan on members' own squads averaged 0.51 points a manager below what they actually scored. Managers beat the plan in 33% of cases and the plan beat the manager in 25%. If I were paying for this, that is the figure I would check first, so it sits in the same section as the win rather than a footnote.

For context on who those managers are: they are engaged players who linked a team, so their results prove nothing about the product on their own. In Gameweek 4, 86% of them finished above the game's average and the best sat inside the top 5,000.

The real winners and losers of GW4

Gvardiol returned 11, which is the round's quiet lesson. He is £5.7m, 23.0% owned, and has 33 points from 344 minutes. Cheap minutes in a good defence keep beating expensive names on paper.

Gabriel returned 9 at £8.0m and 23.5% owned. Tarkowski returned 8 at £6.1m, on all 360 available minutes.

On the other side, Bruno Fernandes returned 2. He is £12.0m and 41.7% owned, and he still has 29 points and 3.41 expected goal involvements, so this was a blank rather than a role problem. Maguire and Ndiaye returned 1 each.

The season's leading scorer remains Bogle at £4.5m and 3.1% owned, on 37 points. Our own board does not rate him for the round ahead, which is the kind of disagreement worth watching rather than acting on.

What to buy and sell for Gameweek 5

Tarkowski (£6.1m), Everton. Fixtures: home to Ipswich. He reads 5.3 for the round, sixth on the whole board, and Everton have a 37.3% chance of a clean sheet, third best in Gameweek 5. Every available minute so far, 29 points, and 13.2% owned. Our model already flagged this move in Gameweek 4 and it was the best-performing recommendation of the week. Bottom line: the cheapest route into a top-three defensive fixture, and at 13.2% owned you are still ahead of the room.

Gibbs-White (£7.9m), Nott'm Forest. Fixtures: home to Coventry. He reads 5.8, second only to Haaland for the round, with Forest expected to score 1.8 and a 35.4% clean-sheet chance. All 360 minutes, 26 points, 2.58 expected goal involvements, 11.9% owned. Bottom line: a one-week spike you can hold, and the ownership is still under 12%.

Hold, do not captain: João Pedro (£7.8m), Chelsea. Fixtures: away at Brentford. He is the most-owned player in the game at 75.3% with 33 points, but he misses our top 30 for Gameweek 5 and Chelsea are expected to score 1.6 at a Brentford side with a 20.6% clean-sheet chance. Bottom line: no reason to sell a 75.3% owned forward on one flat fixture, and no reason to give him the armband either.

All projections are as of 15 September 2026 and you can read every player's expected points for the next eight gameweeks for free.

FAQ

Was the Gameweek 4 captain pick wrong?

Yes. It returned 5 while the second name on the same shortlist returned 9. The wider split was nearly even, which is the useful part: treat a shortlist as three live options, not a ranking to follow blind.

Who is the Gameweek 5 captain?

Haaland, and not narrowly. He reads 6.8 against 5.8 for the next name, with Man City at home to Sunderland and the best attacking read in the round. That was also the call in our Gameweek 4 preview, where the argument was minutes over ceiling.

Should I take a hit to bring in Tarkowski?

No. A 5.3 read is strong for a defender but it does not clear minus 4. Move him in with a free transfer or wait.

Want the same treatment on your own squad? Get a free score on your team and see where The Algorithm disagrees with you before Friday's 18:30 deadline.