We publish our misses. Gameweek 6's picks went out on 7 October with 277 projections attached. Here is how they did.
- Our captain was wrong. Bruno Fernandes returned 1 point, doubled to 2. Erling Haaland returned 6. With the armband on Haaland our total would have been 52, not 47.
- The squad scored 47 with the captain doubled. The average manager scored 58. That's 11 below, after 12 and 10 below in the two weeks before. The best score in the game was 131.
- The ranking beat the crowd again. Ranking players by our model scores +0.4359 against the finished gameweek. Ranking them by ownership scores +0.2331. A 0.203 gap, behind only Gameweek 5's 0.227.
How did our picks do?
| Pos | Player | Projected | Actual |
|---|---|---|---|
| GKP | James Trafford | 4.3 | 2 |
| DEF | Marc Guéhi | 5.0 | 6 |
| DEF | Maxim De Cuyper | 4.7 | 15 |
| DEF | Silva | 4.6 | 0 |
| DEF | Leif Davis | 4.6 | 1 |
| DEF | Bobby Thomas | 4.4 | 4 |
| MID | Bruno Fernandes (C) | 8.5 | 1 |
| MID | Bryan Mbeumo | 7.5 | 8 |
| MID | Bukayo Saka | 7.4 | 2 |
| MID | Marcus Tavernier | 4.9 | 1 |
| FWD | Erling Haaland (V) | 8.3 | 6 |
72.6 projected. 46 scored by the eleven, 47 with the captain doubled. We finished 25.6 points short of our own number. Three of the eleven beat their projection: De Cuyper, Mbeumo and Guéhi.
De Cuyper did most of the saving. He scored against Sunderland, kept a clean sheet and took 3 bonus points, all in 67 minutes: 15. Nobody else in the eleven scored more than 8.
Our Gameweek 6 Wildcard draft carried the same eleven and the same captain, so it scored the same 47.
Was the captain call right?
No. Bruno Fernandes played 90 minutes in Man Utd's 1-1 draw with Spurs, picked up a yellow card and returned 1.
We called the 0.2-point gap between him and Haaland noise before the deadline, and captained him anyway. The noise went against us. Haaland scored in Man City's 3-0 win at Liverpool, took no bonus and returned 6.
The thing we said we couldn't see was his minutes, after he asked to come off for Norway. He played 79.
One correction to Gameweek 5. That review said captaining Haaland over Fernandes "was worth 10 points on the armband alone." That was wrong. The armband adds the captain's score a second time, so swapping a player who scored 2 for one who scored 6 is worth 4. The same sum this week gives 5: Haaland's 6 against Fernandes's 1.
That means the armband explains 5 of our 11-point gap to the average manager. The other 6 sit in the squad.
Why did two clubs cost us again?
Two of our eleven came from Bournemouth, who lost 5-1 at Chelsea. Silva scored 0 and Marcus Tavernier scored 1. We projected them at 9.5 between them.
Two more came from Man Utd. Fernandes scored 1 and Bryan Mbeumo scored 8. The match's biggest United score went to Matheus Cunha, who wasn't in our squad: 10, from an assist, 3 bonus points and a defensive contribution.
Cunha, Fernandes and Mbeumo together were projected at 21.2 and scored 19. We had the club right and the split wrong.
Gameweek 5 was the same shape: two United picks, 4 points between them. Two gameweeks is not a pattern, and we haven't tested whether it is. It's the first place we'll look.
Where were we most wrong?
| Player | Projected | Actual | Minutes |
|---|---|---|---|
| Bruno Fernandes | 8.5 | 1 | 90 |
| Bukayo Saka | 7.4 | 2 | 90 |
| Malick Thiaw | 5.1 | 0 | 90 |
| Gabriel | 6.6 | 2 | 90 |
All four played the full match. Arsenal beat Leeds 2-1 and Saka had no goal or assist, while Gabriel's clean sheet went when Leeds scored. Newcastle lost 4-1 at Coventry City, and Thiaw's two points for playing were wiped out by four goals conceded.
Across all 277 players we projected, our average error was 1.94 points. That is the lowest of the six weeks. The average bias was +0.19, so we leaned optimistic again. The lean sat in goalkeepers and defenders, 126 players projected 0.6 points too high on average. Midfielders went the other way, 123 of them projected 0.25 too low.
The two clubs we underrated most were Man City and Brighton. City's 14 projected players scored 69 against our 49.4. Brighton's 10 scored 65 against our 35.3, in a 2-0 win at Sunderland.
Who did we miss?
| Player | Owned | Projected | Actual |
|---|---|---|---|
| Jordan Henderson | 0.0% | 0.7 | 18 |
| Kiernan Dewsbury-Hall | 5.9% | 4.4 | 13 |
| Jérémy Doku | 0.7% | 0.9 | 12 |
| Dara O'Shea | 2.7% | 3.4 | 11 |
Jordan Henderson was the top scorer in the game. He scored twice and set up a third in Chelsea's 5-1, and we ranked him 244th of 277. We're not claiming these four as differentials we called. None was above 6% owned, and we rated all four below where the gameweek put them, by between 7.6 and 17.3 points. Doku did it in 34 minutes.
Then there are the players our model doesn't project at all. João Pedro scored 12 in 71 minutes, with two goals, and Cole Palmer scored 8 from two assists in 87. Both were flagged at 75% on Tuesday 6 October, so both were among the 108 players we removed from our numbers. Pedro was owned by 64.8% of the game. Both played.
Sidiki Cherif scored 12 for Coventry City and Abdoul Ouattara scored 10 for Ipswich Town. Neither has enough minutes for a number. Twelve players we didn't project scored 6 or more, against 40 that we did.
Treat silence from us as ignorance, not a verdict.
The scorecard
All 277 players, no filters. The measure is how well our order matched the real order.
| Ranking by… | GW1 | GW2 | GW3 | GW4 | GW5 | GW6 |
|---|---|---|---|---|---|---|
| Our model | +0.413 | +0.376 | +0.426 | +0.476 | +0.557 | +0.436 |
| Ownership — what the crowd picked | +0.360 | +0.394 | +0.414 | +0.414 | +0.329 | +0.233 |
| Last season's points per game | +0.396 | +0.218 | +0.230 | +0.270 | +0.198 | +0.182 |
| Price | +0.306 | +0.286 | +0.192 | +0.276 | +0.174 | +0.234 |
| Average error, points per player | 2.05 | 1.99 | 2.02 | 2.19 | 2.00 | 1.94 |
| Hauls (6+) in our projected top 20 | 6/20 | 9/20 | 7/20 | 10/20 | 4/20 | 10/20 |
Our earlier reviews labelled the points-per-game row "points per game so far". It has always been last season's. The label is fixed here.
Both correlations, because one alone flatters. Over all 277 players our model scores +0.436 and ownership +0.233. Over the 229 who actually played it's +0.286 and +0.126. The gap holds, and both numbers fall once the 48 who didn't play are taken out.
The ranking fell back from Gameweek 5's +0.557, and it still led the crowd by 0.203. Ownership had its worst week of the six. Ten of our top 20 hauled, tied with Gameweek 4 for our best, and eight of the twenty scored 2 or fewer.
None of that reached the squad. 11 below the average manager sits between Gameweek 5's 12 and Gameweek 4's 10, and it's the third week in a row below. Gameweek 4 called this a split between ranking and outcome, Gameweek 5 called it a trend of two, and this makes three. 5 of the 11 points are the armband. The other 6 we haven't explained yet.
Tell us where this is wrong
- Which armband did you play, and why? Haaland was the most-captained player in the game again, and owned by 75.0%. Tell us what tipped it, and whether our 0.2 changed anything.
- Did you own Henderson, Dewsbury-Hall or Doku? Tell us how you found them. We ranked all three below where they finished, and we'd like to know what we missed.
- Show us your Gameweek 7 draft and why. The deadline is Saturday 17 October, 11:00 BST. Build it on the Team Planner, share the link, and name the one pick you'd defend against the model.
Send it through the feedback form for anything specific, or join the Discord if it needs a conversation. Both get read every day.
Every number here comes from the file our model wrote on Tuesday 6 October at 15:48 BST, before the deadline, scored against the Gameweek 6 results after FPL confirmed bonus points on Tuesday 13 October.
Next: the Gameweek 7 picks. 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.