Gameweek 6 scored: right about the flags, wrong about the captain again

The locked file was written at 06:38 UTC on Saturday, 3.4 hours before the deadline. I said two hours in every preview, so I should explain that. When the deadline is in the morning, the daily backup job runs before the two hour mark and gets there first. That is how it is set up, and the file says its own margin so nobody has to take my word for it.

All ten games are finished and the bonus points are in. FPL had not marked the week as final when I pulled the numbers on Tuesday, so a point could still move here and there. I do not expect anything that changes the conclusions.

The list beat the baseline again, by about the same

The twenty players averaged 3.65 points. The 163 players the model expects to play an hour or more averaged 3.35. That is an edge of 0.30 points a player, about 9%. In Gameweek 4 it was 0.36. Two scored weeks, two small edges of almost exactly the same size.

There is a harder way to look at it. The 206 players who actually played 60 minutes or more averaged 3.61, and against that the list is level. Nobody knew on Saturday morning who would play 60, so it is not a fair test. But it does tell you how much of a projection is just picking players who start, and that is most of it.

One more thing I do not love. The twenty were projected 4.89 on average and got 3.65. Any top twenty will come in under its projection, because the players at the top of a list are partly there because the model was a bit too kind to them. But 1.2 points a player is a big gap, and it was 1.0 in Gameweek 4. That is one to keep an eye on.

The captain lost to the crowd again

Top of the locked list was Saka at 5.58. Arsenal beat Leeds 2-1 and he played all 90 minutes. He had 0.41 xG and scored 2 points. Calafiori and Bruno G. got the goals.

Haaland, the most captained player in the game again, scored at Anfield in a 3-0 win and got 6. Doubled, that is 4 against 12. In Gameweek 4 it was Isak 2 against Haaland 9. So the model's top captain pick is now two from two against the crowd, and 4 points against 15 in total.

It was not close on the rest of the captain list either. The five names were Saka 2, Haaland 6, Bruno 1, Tarkowski 3 and Mendy 3. Haaland was the best of them.

To be fair to the preview, what I actually wrote was to captain Haaland if you own him, because the gap was 0.11 and that is rounding. Anyone who followed that was fine. But the scorecard scores the table, not my paragraphs, and the table said Saka. I also wrote that Anfield was one of Haaland's tougher games. City won 3-0 there.

Every pick, scored

Six of the twenty got six points or more. Ten got two or fewer, which is worse than Gameweek 4, when it was eight. Hall played 90 minutes and got nothing, as Newcastle lost 4-1 at Coventry. Everyone on the list played, so there was no Nketiah this time.

Six players got 12 or more this week, and João Pedro was the only one on the list. Henderson got 18 for Chelsea, De Cuyper 15, Dewsbury-Hall 13, and Doku and Cherif 12 each. Of those five, De Cuyper is the only one I even mentioned, and I will get to him.

What I got right

The flags. I ended the preview by saying the week would be decided by the flags, not the captaincy, and told people not to sell before the press conferences. João Pedro was sold by 363,000 managers during the week while he sat on 75% with a knee. He was fit, scored twice against Bournemouth and got 12. Palmer had two assists and got 8. Both flags came off before the lock, so the model had them at their full numbers, and Chelsea's three players gave the list 29 of its 73 points.

That is the one I am happiest with, because it was not really a projection. It was just "wait", and waiting was right.

The floor on defensive contributions. Hull against Everton finished 2-2, so the clean sheets I was hoping for never came. But Tarkowski made 17 clearances, blocks, interceptions and tackles, Mendy 13 and Branthwaite 10. All three got the 2 points for it, which is why they ended up on 3 and not 1. Thomas made 22 for Coventry and got 4. The thing I have been banging on about since the fixture run piece did show up. It just was not enough on its own.

Mbeumo over his teammate. Same game as Bruno, eighth on the list, scored and got 8.

What I got wrong

Bruno. I said twice that I did not see why people were selling him, and that this was the wrong week to do it. He got 1 point at home to Spurs. One week does not make the sellers right about the rest of the season, but they were right about this one.

The cheap defenders. This is the one that hurts. I said Egan, Thomas and Davis were all worth starting ahead of a City defender at Anfield, because the model gave City a 14% chance of a clean sheet. City kept a clean sheet. Guéhi got 6. Gvardiol only played 24 minutes and still got 4 with an assist. Egan got 1, Davis got 1 and Thomas got 4. If you swapped a City defender out for one of mine, you probably lost points on it.

Mendy and Egan. I wrote a whole section on why Mendy's 5.17 was too high, mostly because he had never played more than 70 minutes. He played 90. He got 3 and Egan, the one I told you to start instead, got 1. The 5.17 was still too high. I was still wrong about which Hull defender to pick.

De Cuyper. The model had him at 3.53, 96th for the week, and I put him in the table of players the cheap defenders should start ahead of. Brighton won 2-0 at Sunderland, he scored and got 15. In fairness, the midweek piece said GW6 and GW7 were the two good Brighton fixtures and the problem comes after that. That part gets tested in GW8 and GW9. But 3.53 for a player who scored 15 is a miss, and I am not going to dress it up.

Calafiori against Ampadu. The locked file names one player it left out despite the crowd, and one it backed against it. Calafiori, in 51% of teams, was left out at 4.20. He scored and got 9. Ampadu, in 0.9%, was backed at 4.59. He got 1.

A few smaller ones. I said Konsa should not be in your eleven at 2.68. He played 90 and got 4. Davis was my one week start and got 1, even though Ipswich won. Hall was the defender I liked most over the next five, and he started with a zero. Isak was the flagged player I kept coming back to, and he did not play. Neither did Gakpo. I told people to think twice before selling Szoboszlai, and he did not get on the pitch either.

The record so far

Same caveat as last time: "likely starters" is filtered on what the model expects now, not what it expected before the week. It is a fair benchmark, not a perfect one.

The honest summary

The list keeps beating the baseline by a small amount, and the captain pick keeps losing to Haaland. Both of those are now two weeks long instead of one, which is still not much. On the writing side, I was right when the advice was to wait, and mostly wrong when I told you to go against the crowd on defenders.

If there is one thing I take from this week, it is that I spent a whole article arguing about which £4.0m defender to start, and the answer was the City defender everyone already had.

Gameweek 7's deadline is Saturday 17 October at 11:00 UK time. The numbers lock before that, and get scored the same way next week.