FPL GW6 review: Henderson tops the scorers on 18

The FPL GW6 review starts with a surprise at the top. Henderson scored 18 points, the highest in the gameweek, from two goals, an assist and three bonus. Our model had him at 0.5. The FPL average score was 58 and the highest score was 131.

Who were the highest scorers in FPL GW6?

De Cuyper was second on 15 points, with a goal and three bonus in 67 minutes. Dewsbury-Hall followed on 13, then João Pedro, Doku and Cherif on 12 each. O'Shea had 11, and McGinn, Cherki and Cunha all made 10.

Two of those were models of efficiency. Doku managed 12 points in just 34 minutes, and Cherif collected 12 in 77. Bonus did a lot of the work across the list, with most of the ten picking up three bonus points.

Where was the model right, and where was it badly wrong?

Our forecasts are projections, and gameweek-level football is noisy. Across 315 players graded, the mean absolute error was 1.95 points. That is a fair account of a typical miss. The tails are where the interesting cases sit.

Beat the forecast most:

• Henderson: forecast 0.5, scored 18
• Cherif: forecast 0.4, scored 12
• Doku: forecast 0.7, scored 12
• De Cuyper: forecast 6.1, scored 15
• Dewsbury-Hall: forecast 4.1, scored 13

Fell short most:

• B.Fernandes: forecast 7.4, scored 1
• Hall: forecast 5.7, scored 0
• Bentancur: forecast 3.0, scored -2
• Dalot: forecast 4.5, scored 0
• Struijk: forecast 2.5, scored -2

The first list is mostly low-minutes or low-expectation players, which is where any projection will struggle: Henderson, Cherif and Doku were all forecast under one point, so the model did not see those returns coming. De Cuyper is a different case. We had him at 6.1, a strong projection, and he still beat it by a wide margin.

The second list is the one to learn from. These were players we expected to deliver and who did not, and the biggest miss was our own captain pick.

How did the model captain do?

Our model's captain was B.Fernandes. He returned 1 point, which doubled to 2. With a forecast of 7.4 he was the largest single shortfall of the week, so the captaincy cost us as well as anyone who followed it. Captain choice is the highest-variance decision in the game, and a good process will still produce weeks like this.

How did we do against the FPL Experts?

Our team scored 84 in the gameweek. The FPL Experts scored 38, our model XI scored 64 and the FPL average was 58. Our team came out on top this week. Over the season, our team has beaten the FPL Experts in four of six gameweeks, 17 points ahead on total, and 46 points above the FPL average. The full comparison is on the us v the FPL Experts page, and the gameweek 6 archive has every score.

How we worked this out

Forecasts are our pre-deadline expected points for each player, and points are the official FPL scores once the gameweek finished. Every graded player is compared against what we published before the deadline. The running record of how accurate the model is lives on the model accuracy page.