We publish our misses. Gameweek 3's picks went out on 3 September with 272 projections attached. Here is how they did.
- The captain was right. We said Haaland at 7.9. He scored 9.
- The rest was not. Our eleven projected 60.8 and scored 31.
- The average manager scored 51.
How did our picks do?
| Pos | Player | Projected | Actual |
|---|---|---|---|
| GKP | Caoimhín Kelleher | 4.3 | 2 |
| DEF | Marc Guéhi | 6.2 | 8 |
| DEF | Gabriel | 5.7 | 2 |
| DEF | Malick Thiaw | 4.6 | −1 |
| DEF | James Tarkowski | 4.3 | 3 |
| MID | Bruno Fernandes (C) | 7.1 | 2 |
| MID | Iliman Ndiaye | 6.9 | 3 |
| MID | Phil Foden | 6.8 | 1 |
| MID | Bryan Mbeumo | 5.3 | 8 |
| FWD | Igor Thiago | 5.4 | 2 |
| FWD | Dominic Calvert-Lewin | 4.2 | 1 |
60.8 projected. 31 scored. 33 with our own captain doubled. Two of the eleven beat their number. The best score in the game that week was 119.
One caveat, and it was in the original piece: that squad was built for Gameweeks 3 to 10, not for Gameweek 3 alone. It still had to play the week.
Was the captain call right?
Yes. Haaland scored, kept a clean sheet and took three bonus in a 1-0 win over Coventry. Of our top 20 he beat his number by the most. He was also the most-captained player in the game, so we cannot claim much credit for it.
But the squad printed under that advice captained Bruno Fernandes at 7.1, and he scored 2.
That is the solver, not a contradiction. Haaland is £15.5m, which is 15.5% of the budget, and across eight weeks it would rather spend that elsewhere. We wrote "if you already own Haaland, keep him and captain him" directly above the table.
So one article gave an 18-point captain and a 4-point captain. Which one you got depended on whether you read the line or copied the table. That is our fault, and it is a writing fault.
What happened to the Man City bet?
We led on it: six of our top eight were City players, and you are only allowed three.
The fixture came in. City won 1-0 and kept a clean sheet. The six still returned 30 points against 40.4 projected.
| Player | Projected | Actual |
|---|---|---|
| Erling Haaland | 7.9 | 9 |
| Iliman Ndiaye | 6.9 | 3 |
| Phil Foden | 6.8 | 1 |
| Antoine Semenyo | 6.3 | 6 |
| Enzo Fernández | 6.3 | 3 |
| Marc Guéhi | 6.2 | 8 |
The fixture read was right. A 1-0 win pays one scorer, one clean sheet and three bonus, and nothing to the other five. We can pick which team will control a game. We cannot pick which of their eleven scores.
Three per club is the rule. It saved us here.
Where were we most wrong?
| Player | Projected | Actual | Minutes |
|---|---|---|---|
| Phil Foden | 6.8 | 1 | 24 |
| Malick Thiaw | 4.6 | −1 | 90 |
| Bruno Fernandes | 7.1 | 2 | 90 |
| Nico O'Reilly | 5.0 | 0 | 0 |
Three of those four are minutes, not points. Foden played 24 of 90. O'Reilly, owned by 19% of the game, never came on. We rated what they would do if they played, and then they didn't. Nothing sees a team sheet before it is published.
Thiaw is the other kind: 90 minutes, an own goal, minus one. Nothing to fix there.
Across all 272 players our average error was 2.02 points, and we project 0.18 high — the same direction as the two weeks before.
Who did we miss?
| Player | Owned | Projected | Actual |
|---|---|---|---|
| Tyrick Mitchell | 4.8% | 2.9 | 15 |
| Jayden Bogle | 0.5% | 1.7 | 14 |
| Alexander Isak | 16.6% | 2.9 | 13 |
| Harvey Barnes | 1.1% | 3.7 | 12 |
| Cody Gakpo | 8.1% | 5.3 | 11 |
| Tyrique George | 0.3% | 2.1 | 10 |
Two of the top four are defenders who scored. Mitchell got two goals in a 3-2 win at Fulham. Bogle scored and kept a clean sheet. We rate a full-back on clean sheets plus a small chance of a goal, and when that chance lands it pays four or five times the number.
Eighteen players under 10% owned scored 8 or more. We projected fifteen of them below 4. Our differentials are players we think are cheap for what they do. They are not players we expect to haul.
The scorecard
All 272 players, no filters. The measure is how well our order matched the real order.
| Ranking by… | GW1 | GW2 | GW3 |
|---|---|---|---|
| Our model | +0.413 | +0.376 | +0.426 |
| Ownership — what the crowd picked | +0.360 | +0.394 | +0.414 |
| Points per game so far | +0.396 | +0.218 | +0.230 |
| Price | +0.306 | +0.286 | +0.192 |
| Average error, points per player | 2.05 | 1.99 | 2.02 |
| Hauls (6+) in our projected top 20 | 6/20 | 9/20 | 7/20 |
This was our best week of the season and we still only beat the crowd by 0.012. We called Gameweek 2's 0.018 the other way a loss, in bold. This gets the same treatment.
Read the third row instead. Ranking players by points per game so far scores +0.230. We score +0.426. Beating the crowd by 0.012 is noise. Beating the obvious method by 0.196, three weeks running, is the point.
Seven of our top 20 hauled. Seven blanked. One didn't play. We do not find certainties — we move the odds a little, and this is what a little looks like on a bad week.
Tell us where this is wrong
- Did you captain Haaland, or copy the table? If you copied the table you got 4 points from your armband. We would rather know.
- Which of the eighteen cheap hauls did you own? Mitchell, Bogle, Barnes and George were all under 5%. Tell us what you saw.
- Build Gameweek 4 yourself on the Team Planner. Gameweek 4's picks are up, and the model has stopped captaining Haaland.
Every number here comes from the file we published on 3 September, before the deadline, scored against the finished gameweek.