The new Premier League season burst into life with Hull City providing the opening weekend’s seismic shock, beating Manchester United 2–0 and wrecking one of our model’s highest-confidence calls. Across ten matches, the round delivered drama, upsets and a mixed bag for our forecasts. The model correctly called seven of ten match results and eight of ten win-or-draw outcomes, but badly undercooked the surprises at the MKM Stadium, the City Ground and the Amex.
How our forecasts did
The win or draw market was the round’s standout, landing eight of ten calls correctly. Match result predictions were solid, with seven correct, though the misses came in the places that hurt most. Both teams to score and total goals each found the mark six times, a steady if unspectacular return. First team to score proved the trickiest of the mainstream calls, succeeding only four times, while the half-time and full-time market managed just three correct. The exact scoreline calls were always a speculative layer, and only one of ten came off — Manchester City’s 2–1 win over Bournemouth.
Friday
Arsenal 3–0 Coventry
Arsenal opened with a routine win at home to newly promoted Coventry, and the model read this one perfectly apart from the exact scoreline. High-confidence calls for an Arsenal win, Arsenal to win or draw, and Arsenal to score first all landed, as did the forecast of more than 2.5 goals and at least one side being kept off the scoresheet. Even the low-confidence half-time and full-time call of Arsenal ahead at the interval and winning at the whistle was correct. Only the 2–0 scoreline prediction missed, by a single goal.
Saturday
Hull City 2–0 Manchester United
The round’s headline result was also the model’s biggest error. A high-confidence call for a Manchester United win was obliterated by Hull’s clinical display. The only correct forecasts in this game were the medium-confidence under 2.5 goals call and the low-confidence at-least-one-side-to-blank prediction, which both landed almost by accident. Every other call — including a United win-or-draw, United to score first and a United half-time/full-time lead — failed.
Everton 2–0 Crystal Palace
Everton extended their remarkable home record over Palace to an eleventh league meeting without defeat, and the model’s medium-confidence call for an Everton win was spot on. The more cautious Everton win-or-draw also landed, as did the forecast of Everton scoring first and the half-time/full-time call of an Everton lead throughout. The low-confidence overs and both-teams-to-score calls failed, with the clean sheet proving the difference.
Ipswich 2–1 Sunderland
A tight newly-promoted affair swung Ipswich’s way, and the model’s low-confidence Ipswich win call was correct. The forecast of a winner either way and both teams to score also landed. Ipswich scoring first was correct, but the call for under 2.5 goals and a 1–0 scoreline were undone by Sunderland’s reply, while the half-time draw call was off.
Nottingham Forest 0–1 Leeds
This was the round’s clean sweep of misses. All seven forecasts for this game were incorrect. The medium-confidence Forest win call was wrong, as was the low-confidence Forest win-or-draw, and both teams to score failed along with the rest. Leeds’ lone goal settled a contest the model thought the hosts would edge.
Brentford 3–0 Tottenham
Brentford’s thrashing of a weakened Tottenham side was flagged by the model’s low-confidence Brentford win call, a sharp spot. The over 2.5 goals call also landed, but the medium-confidence both-teams-to-score forecast and the first-scorer pick of Tottenham and half-time/full-time Tottenham lead were all wrong as Spurs failed to score.
Sunday
Brighton 4–0 Aston Villa
The Amex witnessed a rout few saw coming. The model did lean toward a Brighton win on low confidence, and Brighton to win or draw landed, but almost everything else fell apart. The under 2.5 goals call was buried by a four-goal blitz, both teams to score missed because Villa drew a blank, and Villa to score first and a draw at half-time were equally off the mark.
Manchester City 2–1 Bournemouth
The champions were pushed but prevailed, and this was the model’s most complete performance. The high-confidence Manchester City win, win-or-draw, and City to score first all landed, as did the medium-confidence overs and both-teams-to-score calls. Even the exact scoreline 2–1 and the half-time/full-time City lead were correct, making this the only fixture where a scoreline call hit.
Newcastle 2–2 Liverpool
A thrilling draw on Tyneside stopped the model’s medium-confidence Liverpool win call, but the Liverpool win-or-draw forecast was correct. Overs and both teams to score landed too, though Liverpool to score first and to lead at the break and win — both calls — were wrong after Newcastle’s spirited comeback.
Monday
Fulham 2–3 Chelsea
A cracking London derby ended with Chelsea edging it, exactly as the medium-confidence Chelsea win call predicted. The low-confidence winner-either-way call, overs, both teams to score, and Chelsea to score first all landed. The half-time/full-time call of Chelsea ahead at the break and winning was correct too, leaving only the 2–1 scoreline prediction a goal off.
What we got wrong: Hull, Forest and a scattering of upsets
The model’s high-confidence Manchester United win was the weekend’s worst miss. The analysis pointed to United’s deeper squad and top-flight pedigree, but defensive doubts over de Ligt and Ugarte were not enough to shift the call. Hull’s clean, two-goal win exposed the gap between preseason assumptions and opening-day reality.
Nottingham Forest’s home loss to Leeds was a similar miscalibration. The model saw a narrow Forest edge, reasoning that Leeds’ long injury list — Gnonto and Joseph among those out — would leave them blunt. Instead Forest failed to score altogether, and every call missed. The Brighton rout of Villa also saw multiple calls misfire, though at least the low-confidence Brighton win was in the right direction. Brentford’s demolition of Spurs was another where the model correctly sensed an upset but still got several layers wrong.
Opening weekends are notoriously volatile, and this one was no different. The model’s baseline in match results and win-or-draw markets held up well, but the granular calls wilted when the upsets landed. The Premier League table already looks lopsided, and it’s a reminder that early-season forecasting is as much about humility as it is about data.
