Gameweek 5's Biggest Fantasy Surprises — and Disappointments
A Brian Brobbey hat-trick in a 5-3 defeat, a Man City win that put one player in our top two and two defenders in our bottom five, and Bruno Fernandes's GW2 hat-trick still inflating a prediction he keeps missing.
Same method as Gameweeks 1–4 — every player below played at least 30 minutes and is ranked by actual points minus PitchMetric's predicted xP for that fixture, using the predictions our daily snapshot captured before the gameweek rather than a hindsight reconstruction. Two honest notes on this one. It went up late, after the September international break, so it's written with Gameweek 6 already played. And the GW5 snapshot was taken on the morning of Saturday 19 September — after Friday night's Brentford 3-0 Chelsea had already finished — so that match's players are left out of the ranking, because their numbers were computed with its result already known. That's why the pool is 207 players rather than the full gameweek.
Ranked among 207 players who played at least 30 minutes in gameweek 5, by actual points minus predicted xP. Average surprise across that pool: -0.06 (std. dev. 3.66).
Box score: 3 goals, 3 bonus · xG 2.35, xA 0.04 · ICT 20.4
Three goals and 3 bonus points — in a 5-3 defeat at Manchester City. Unlike most hat-tricks in these lists, his 2.35 xG says the chances really were there, so this reads less like a finishing fluke than a striker the model had almost nothing on: four gameweeks of 2, 2, 2 and 1 points, with no goal or assist, gave a £5.8m forward a 1.98 prediction, and a one-game burst this size was never going to fit inside it.
Box score: 2 goals, 1 assist, 2 bonus · xG 0.16, xA 0.17 · ICT 17.2
Two goals and an assist in the same 5-3 game, from just 0.16 xG — this one is finishing, not chance quality. It's also half of a split story: the result that made Semenyo our #2 put two City defenders in our bottom five (Matheus Nunes and Rúben Dias, below), because conceding three goals wipes out clean-sheet points for everyone behind the ball.
Box score: 1 goal, clean sheet, 3 bonus · xG 0.32, xA 0.04 · ICT 8.1
The winner, a clean sheet and 3 bonus points in Coventry's 1-0 win at Nottingham Forest — a £4.0 defender at a newly-promoted club, exactly the profile our price-based prior is built to under-rate. Goalkeeper Rushworth (11 points, +9.45) narrowly missed this list on the same result, and Forest's Gibbs-White is at the other end of it, below.
Box score: 1 goal, 1 assist, clean sheet, 2 bonus · xG 0.44, xA 0.16 · ICT 14.9
A goal, an assist and 3 bonus points in Aston Villa's 3-2 win at Spurs, with a clean sheet on top: Villa conceded nothing while he was on the pitch, because both Spurs goals came after he went off at 71 minutes. He'd played just 27 minutes all season before this, so a 1.34 prediction was the model correctly knowing almost nothing about him.
Photo: Bryan Berlin, CC BY-SA 4.0, via Wikimedia Commons
Box score: 1 goal, 1 assist, clean sheet, 3 bonus · xG 0.11, xA 0.25 · ICT 10.5
A goal, an assist and a clean sheet as Brighton beat Arsenal 3-0 at home — his third appearance in our top five after GW2 and GW4. He has now beaten his prediction by 9.6, 12.1 and 10.5 points in three different gameweeks, and this week's prediction was still only 3.55. Worth watching whether the model starts to learn him.
Six of the ten entries on this page come from just two matches: Manchester City 5-3 Sunderland (Brobbey and Semenyo up, Matheus Nunes and Rúben Dias down) and Coventry's 1-0 win at Nottingham Forest (Dasilva up, Gibbs-White down). A lot of goals at both ends pulls every player on the pitch in opposite directions at once, and that's not something a model built from each player's own history can see coming. The one entry that isn't about a single match is Pascal Groß: three appearances in our top five in five gameweeks is hard to call pure variance.
Gameweek 5's biggest disappointments
Every gameweek also has an equal and opposite story: players our model rated highly who delivered far less. Same method, same 30-minute cutoff, ranked the other way — the five biggest gaps between what we expected and what actually happened, on the low side.
Box score: — · xG 0.00, xA 0.01 · ICT 2.8
Predicted almost 8 points at home to Sunderland, off the back of a 9-point GW4, and got an appearance point: Manchester City won 5-3, but conceding three goals means no clean sheet and no bonus. The mirror image of Semenyo's #2 spot above — same match, opposite end of the pitch.
Box score: — · xG 0.15, xA 0.02 · ICT 4.8
An own goal in a 1-1 draw at Fulham turned an otherwise unremarkable 90 minutes into a zero: 2 points for playing, minus 2 for the own goal, nothing else. The same single-event problem as Lindelöf's GW1 and Thiaw's GW3 own goals in these lists — one moment outweighing ninety minutes of ordinary defending.
Photo: Bryan Berlin, CC BY-SA 4.0, via Wikimedia Commons
.jpg){kind=link}
Box score: — · xG 0.28, xA 0.19 · ICT 8.6
The same 1-1 draw, no return from any chance (0.28 xG, 0.19 xA). The more useful number is the prediction: 8.07 for a midfielder whose three other gameweeks before this were exactly 2, 2 and 2 points. His 23-point, three-goal GW2 (our #1 that week) is doing nearly all the work in that estimate — and he's in this list again in Gameweek 6.
Box score: — · xG 0.20, xA 0.00 · ICT 1.2
The second City defender below Matheus Nunes from the same 5-3 win: three goals conceded at home means no clean sheet and no bonus however well City won it. A defender's fantasy score depends far more on the scoreline at the other end of the pitch than on his own performance.
Photo: Sebalston, CC0, via Wikimedia Commons
Box score: — · xG 0.00, xA 0.40 · ICT 11.0
Nottingham Forest lost 1-0 at home to newly-promoted Coventry — the other side of Dasilva's hit above. Like Fernandes, he arrived off one huge week (13 points in GW2, where he was also in our top five) and a prediction that assumed it was repeatable; his 0.40 xA says he created something, but none of it became an assist.