Gameweek 3's Biggest Fantasy Surprises
Two of this week's five biggest surprises came from the exact same match — one from each side. The first surprise post built entirely on our updated model.
Model v2 — all five predictions revised: the same same-day scoring-bug fix and model updates described in the GW4 post apply retroactively to every gameweek with real current-season minutes behind it. Ranking is unchanged — only the underlying numbers moved.
Same method as Gameweeks 1 and 2 — every player below played at least 30 minutes, ranked by actual points minus PitchMetric's predicted xP for that fixture. This is the first list built entirely on the updated model — though as it turns out, the update's headline piece (leaning on price when there's not much track record yet) still hasn't had real players to work with this early in the season; it needs a few matches' worth of established players before it can learn anything, and nobody's quite there yet. What's below instead reflects the model's fixture-level read, or the same minutes-not-starts fix already visible in last week's Cherki case. One honest caveat: Gameweek 3 hadn't been officially finalized by FPL at the time of writing (bonus points can shift slightly in the day or two afterward), so treat the numbers below as very likely final rather than certainly final.
Ranked among 228 players who played at least 30 minutes in gameweek 3, by actual points minus predicted xP. Average surprise across that pool: +1.22 (std. dev. 2.93).
Photo: Timmy96, CC0, via Wikimedia Commons
.jpg){kind=link}
Box score: 1 goal, clean sheet, 3 bonus · xG 0.90, xA 0.02 · ICT not yet finalized by FPL
The single biggest surprise of the round, and one that isn't really about Bogle personally at all — it's a fixture-level miss. Leeds went away to a Brighton side that had just put four past Aston Villa and beaten Manchester United, and the model's underlying read gave a Leeds defender very little credit for keeping anything out that day. Bogle, an established first-choice right-back rather than an unknown quantity, broke that read on his own: he scored, and personally kept a clean sheet for the 60 minutes he was on the pitch, before Brighton's equalizer arrived after he'd already been substituted, in a match that finished 1-1. The model had a real opinion here — it was just wrong.
Photo: Ardfern, CC BY-SA 4.0, via Wikimedia Commons
.jpg){kind=link}
Box score: 2 goals, 2 bonus · xG 0.66, xA 0.01 · ICT not yet finalized by FPL
An established regular at left-back for Palace, not someone the model had no history on — the surprise here is the kind of return, not the player. Palace won 3-2 away at Fulham with both sides trading goals throughout, and Mitchell personally scored twice at the other end of a game his own team conceded two in — no clean-sheet bonus propping this one up, just pure goal threat from a full-back. Attacking returns from a defender, however established, are inherently harder for a model built on clean-sheet probability and a player's usual scoring rate to see coming than defensive ones are.
Photo: Pottercomuneo, CC BY-SA 4.0, via Wikimedia Commons
.jpg){kind=link}
Box score: 1 goal, 1 assist, 2 bonus · xG 0.11, xA 0.03 · ICT not yet finalized by FPL
An established Newcastle winger with a real track record, who simply had an excellent day at exactly the moment the model wasn't banking on one: a goal and an assist at home in a 2-2 draw with Bournemouth, worth 12 points against a 1.53 prediction. Nothing structurally unusual about this miss — just good, in-form football beating a modest, evidence-based baseline on the day.
Photo: Bryan Berlin, CC BY-SA 4.0, via Wikimedia Commons
Box score: 1 goal, 2 bonus · xG 0.19, xA 0.01 · ICT not yet finalized by FPL
The same Brighton-Leeds match as Bogle above, seen from the other side. Vušković is a brand-new arrival at Brighton — a club-record transfer worth up to £50m from Tottenham this summer, after a breakout loan season at Hamburg that won him the Bundesliga's Rookie of the Year award — so there's a real, defensible reason the model had comparatively little of his own to go on, on top of how tough it read the fixture for either side that day. His goal proved to be Brighton's equalizer, canceling out Bogle's earlier one for Leeds in a match that finished 1-1. Two of this week's five biggest surprises, from a combined 150 minutes of the same game.
Photo: Mikael Hervestad, CC BY-SA 4.0, via Wikimedia Commons
.jpg){kind=link}
Box score: 2 goals, clean sheet, 3 bonus · xG 0.21, xA 0.05 · ICT not yet finalized by FPL
Isak needs no introduction as a talent — Liverpool paid a British transfer record (£125m) for him roughly a year before this season kicked off — but reputation alone was never what fed this prediction. What mattered is that after a blank in Gameweek 1, he'd already opened his account with a goal in Gameweek 2, and that real, if thin, current-season record is what the model was actually working from: 3.77 points predicted, well above a total unknown's floor. He doubled up with a brace away at Ipswich Town in a 2-0 win, worth 13 points. At just 0.21 combined expected goals for two actual finishes, this was clinical finishing well ahead of the underlying chance quality — the smallest, most ordinary kind of surprise on this list: a proven forward simply having a very good day.
Brighton's home draw with Leeds produced two of these five on its own, from opposite ends of the same ninety minutes — proof that “surprise” doesn't mean the model was clueless about the match, just wrong about a specific name within it. Compare that to Isak's case: the least surprising kind of surprise, a proven talent simply having a very good day against a prediction that already knew he was decent, just not decent enough to see two goals coming.
Gameweek 3'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 0.00 · ICT 2.0
Booked and hooked at half-time of a match Brentford could only draw 1-1 at home — a below-par Brentford performance overall (four of our GW3 worst five wore the same shirt), with Yarmoliuk taking the earliest exit of the group.
Box score: — · xG 0.00, xA 0.00 · ICT 6.8
Played the full match but a home draw against newly-promoted Sunderland, with a goal conceded and no attacking return, was a long way short of the price-implied ceiling the model had on him.
Box score: — · xG 0.00, xA 0.00 · ICT 5.2
An own goal in a 2-2 home draw with Bournemouth turned what looked like a solid underlying shift into a negative score — the same single-event problem as Lindelöf's GW1 own goal: one moment overwhelming ninety minutes of otherwise ordinary defending.
Box score: — · xG 0.00, xA 0.00 · ICT 4.9
The other side of the same Newcastle-Bournemouth draw — a yellow card and two goals conceded away from home wiped out any clean-sheet hope from the opening whistle.
Photo: Werner100359, CC BY 4.0, via Wikimedia Commons
.jpg){kind=link}
Box score: — · xG 0.00, xA 0.00 · ICT not yet finalized by FPL
A booking and a half-time withdrawal in the same underwhelming Brentford home draw — the third of four Brentford players in this week's worst five, all from the same flat afternoon.