The Championship’s second weekend delivered the sort of early-season jolt that makes the division impossible to predict, with West Ham’s 2-1 home defeat by Charlton the standout shock. The model entered the round confident about the Hammers but walked away with only three correct match results from the 12 games. A lot of forecasts that leaned on opening-day performances came unstuck as the round unfolded — though the win-or-draw calls did much of the heavy lifting, landing eight times.
How our forecasts did
The scoreboard tells a blunt story. The model found match results unforgiving: just three of the 12 calls proved correct, the lowest tally of any topic alongside the exact scoreline market, which drew a complete blank. Half-time and full-time fared little better, with only two correct from the dozen attempts.
In contrast, the win-or-draw calls provided a steadier hand, coming good in eight of the 12 fixtures. Both teams to score and first team to score each split the round down the middle with six correct, while total goals managed five.
Saturday 22 August
Birmingham 2-2 Bristol City
The model expected a tight Birmingham win, but a game that produced four goals — and saw both sides score — quickly undid the call for fewer than 2.5 goals and the clean-sheet expectation. The half-time and full-time forecast of Birmingham pulling away after the break was long gone by the final whistle.
Lincoln 1-3 Portsmouth
A low-confidence call for a Lincoln victory missed the mark entirely as Portsmouth scored first and ran out comfortable winners. The model did correctly flag the winner-either-way pattern and the likelihood of more than 2.5 goals and both teams scoring, but the exact scoreline and half-time calls were off.
Millwall 3-0 Norwich
Millwall’s dominant display at The Den justified the medium-confidence match result call, and the model correctly identified the hosts would score first and lead at the interval. The forecast for both teams to score was the only blemish in an otherwise sharp projection for this one.
Blackburn 2-1 Middlesbrough
Middlesbrough were backed to win on the road, but Blackburn overturned that gently confident call with a 2-1 victory. The first-team-to-score forecast landed when Boro opened the scoring, but the model’s expectation of fewer than 2.5 goals and a clean sheet for at least one side were washed away by a match that turned lively.
Derby 2-2 Cardiff
Derby’s failure to see out a win damaged a cautious match result call, and the medium-confidence tip for fewer than 2.5 goals was undone by a four-goal share. The model had read recent tight meetings between these sides, but this one refused to follow the script.
Preston 1-3 Wolves
Everything clicked for the visitors here. The medium-confidence Wolves win landed, the first team to score and half-time lead calls were correct, and the match sailed past the 2.5-goal threshold. Only the exact scoreline (1-2) came up one goal short.
QPR 0-0 Bolton
A clean miss across the board. QPR were a medium-confidence home win pick, but a goalless stalemate at Loftus Road defied every forecast: no winner, no goals, no opening strike, no both-teams-to-score. The model had expected a much livelier afternoon.
Southampton 3-1 Stoke City
The model’s only high-confidence match result call of the round was emphatically correct. Southampton’s attacking sharpness, flagged in the analysis, was too much for a Stoke side that travelled with injuries and created little. The win, win-or-draw, both teams to score, and first team to score calls all hit, though the 3-1 scoreline meant the exact score and half-time forecasts missed the mark.
Swansea 0-0 Sheffield Utd
A cautious Swansea win call slipped away in a goalless draw, but the model read the game’s texture well: fewer than 2.5 goals and a clean sheet for at least one side were both correct. The prediction that Swansea would score first, however, never materialised.
West Ham 1-2 Charlton
The round’s biggest upset saw West Ham fall at home despite being a high-confidence pick. Every call attached to the fixture ended up incorrect — Charlton scored first, both teams found the net, and only two goals were scored when the model had anticipated more. The analysis had noted West Ham’s late fragility on the opening day, and Charlton exploited it.
Wrexham 1-1 Watford
A low-confidence Wrexham win was undone by a draw that the model had rated the least likely outcome. The both-teams-to-score call was the lone success on a fixture where the winner-take-all expectation never took shape.
Sunday 23 August
West Brom 3-1 Burnley
A draw was the low-confidence call here, but West Brom’s 3-1 victory put paid to that and to the fewer-than-2.5-goals projection. The hosts did score first and the game delivered both teams on the scoresheet — two wins for the model’s more modest forecasts — but the scoreline and half-time picture were way off.
What we got wrong
The West Ham forecast will sting the most. Charlton’s 2-1 win at the London Stadium overturned a high-confidence home victory call, and the model’s reading of the Hammers’ late fragility proved more significant than anticipated. The match also missed the goal-laden scenario the model had projected, leaving every one of the seven attached calls incorrect.
Derby-Cardiff highlighted the risk of leaning too heavily on historical tightness. The model saw a run of low-scoring meetings and backed fewer than 2.5 goals with medium confidence, only for the match to deliver four goals before full-time. Similarly, QPR’s 0-0 draw with Bolton defied a raft of calls that assumed an open game with a home winner; the model’s expectations of a high-scoring contest were completely wide of the mark.
Across the round, the model’s difficulty with precise outcomes was laid bare: not a single exact scoreline landed, and only two half-time and full-time calls came through. When the margin for error is that fine, one unexpected goal tends to unravel everything.
After two weeks of the Championship season, the table is already littered with evidence that early numbers don’t always tell the full story. The model will bank the win-or-draw return as a foundation, but the match result misfires in a league famous for its unpredictability are a reminder that data, however well digested, never quite tames this division.
