FPL GW3 Review: The Sell Was Right, The Buy Was Not

Gameweek 3 was a poor round. The game averaged 51 points, down from 81 in Gameweek 2, and most of the damage came from one decision that half the game got half right.

The sell was Bruno Fernandes. That was correct. He is £12.0m, 47% owned, and he returned 2 points.

The buy, for most people, was Saka. He returned 2 as well.

So the round did not punish you for selling the expensive midfielder. It punished you for what you bought with the money.

The receipts

The Algorithm's Gameweek 3 XI was picked before the deadline and belongs to nobody, which is why we can print it. It scored 64 against the game's average of 51.

It also projected 67.8, so the miss was downwards. That is the honest half: the XI beat the round by 13 and still fell 3.8 short of its own number.

The eleven, in a 3-4-3, were Donnarumma in goal, Gabriel, O'Reilly and Gvardiol at the back, Semenyo, Saka, Szoboszlai and Cherki in midfield, and Haaland, Isak and Thiago up front. Haaland took the armband.

Set against Gameweek 2, when the same XI scored 105 against a projection of 62.7, the pattern is a projection that is closer in a low-scoring round than in a high-scoring one. Two rounds is not a trend. It is worth watching.

On transfers, comparing the recommended move against the move the manager actually made in the same squad in the same week, hits included: the recommendation won 38% of the time, lost 18% and finished level in 44%. The average edge was 1.33 points a manager. Across the whole recommendation rather than the single move, the plan beat the manager 54% of the time and the manager beat the plan 21%, worth 3.45 points a head.

Move by move, the shares tell the story of the round:

Four of the six moves involved selling the same player. The one that paid was the least popular version of it: Semenyo returned 6 where Fernandes returned 2. The one that cost was Saka for Tzolis, because Tzolis returned 5.

Among members, who are engaged managers who have linked a real team, just under three in five finished the round above the game's average, the best of them inside the top 10,000. Fewer than two in five sit inside the top million overall. That is a description of the people, not a claim about the product.

The captaincy verdict

Nine in ten members agreed with the armband call, which is less impressive than it sounds: the crowd had largely arrived at Haaland anyway. Haaland accounted for 83% of the recommendation and returned 9 points. Semenyo accounted for 16% and returned 6.

Where the captain call and the manager's own pick differed, the manager was right 54% of the time. The call was right 40% of the time. That is a loss and it goes on the record.

Where we were wrong

Two things.

The XI undershot its projection, and the shortfall sat in defence. Gabriel returned 2, Calafiori returned 2, and O'Reilly played 150 minutes across the season so far and has 4 points.

Second, Saka. He is a legitimate hold at 11.3% owned with 1.85 expected goal involvement in 246 minutes, and he reads 5.7 for Gameweek 4 and 42.1 across the eight ahead. But he was the most-recommended incoming player in a round where he blanked, and the cheaper Semenyo route was worth four more points to the plans that took it.

What to buy and sell for Gameweek 4

The deadline is 13:30 UK on Saturday 12 September. That is a matchday deadline, so Friday's press conferences are the last team news you get.

Buy: Palmer (£9.6m), Chelsea. He reads 6.3 for Gameweek 4, the highest single-gameweek projection on our board, and Chelsea host Hull with a 56.6% clean-sheet chance and 2.447 expected goals for, the best in the round at both ends. Palmer is 21.4% owned against João Pedro's 72.1%. He is the underowned route into the same fixture.

The armband disagrees with the template. Haaland reads 5.3 for Gameweek 4 at Old Trafford, a point behind Palmer, and he is 71.2% owned. He rebounds to 7.1 in Gameweek 5 against Sunderland at home and totals 50.4 across the eight ahead, the highest of any player. This is a one-week armband switch, not a sale.

Sell: the Hull budget block. Tzolakis (£4.6m, 8.9% owned) has 26 points and Ajayi (£4.1m, 10.8% owned) has 25, which is why they are in so many teams. Hull travel to Chelsea with an 8.7% clean-sheet chance and 2.447 expected goals against. The cheap points were real. This is the week they stop.

Differential: Tavernier (£6.0m), Bournemouth, 3.0% owned. He has 21 points and 1.84 expected goal involvement from 268 minutes, and reads 4.3 for Gameweek 4 with 31.6 over the eight ahead. The fixture is not the argument, since Bournemouth host Brentford with a 21.4% clean-sheet chance. The underlying is, and at 3.0% owned you are paid for being early. Our guide to picking a differential that actually pays sets out the test.

Availability risk that matters: O'Reilly is 15.8% owned and carries a 75% chance of playing off a back injury. Price minutes before points.

All projections are as of 7 September 2026 and every player's number is on our free predicted points page, updated daily.

FAQ

Was selling Bruno Fernandes a mistake?

Not on the week. He returned 2 points from £12.0m. It becomes a mistake if you sold him for Saka rather than Semenyo, and it may still turn: Fernandes reads 4.5 for Gameweek 4 but 42.1 across the eight ahead, level with Saka.

Should I captain Palmer over Haaland in Gameweek 4?

Our numbers say yes by a point, 6.3 to 5.3, and only for this round. Haaland is back on top from Gameweek 5.

Is a 3.8-point miss on the XI a bad projection?

It is a small one, and it was downwards, which is the direction you would rather miss in. The fixture picture moved more than the players did.

The Algorithm plans the next eight gameweeks for your own squad, transfers, captain and chips. Score your team free before Saturday.