Scottish Premiership — Post-Match Review
Dundee United 1-1 Rangers
Tannadice Park — Opening Fixture
The new Premiership season opened with a result our model rated as the least likely outcome: a draw. Lachlan Rose gave Dundee United the lead just before half-time, Thelo Aasgaard levelled with a thumping finish after the break, and neither side could find a winner despite clear chances late on.
So how did the pre-match analysis hold up? The headline call was wrong. Much of the underlying read was right. This review looks at both honestly.
Pre-Match Projection vs Outcome
- Dundee United win: 22.9% — did not happen
- Draw: 19.6% — the actual result
- Rangers win: 57.6% — did not happen
- Both teams to score: 67.2% — hit
- Model verdict: Away Win (Low confidence) — incorrect
What the Model Got Wrong
The model made Rangers clear favourites at 57.6%, and they did not win.
That deserves stating plainly. A near-six-in-ten favourite failing to win is not a catastrophic miss in probabilistic terms — the same projection implied Rangers would drop points in more than four of every ten stagings of this fixture — but the verdict was Away Win, and the result was a draw.
The draw, at 19.6%, was the outcome the model rated least likely. When the low-probability outcome lands, the correct response is not to explain it away, but to note it and examine whether the reasoning behind the probabilities still made sense. In this case, largely, it did.
What the Model Got Right
The fragile favourite
The pre-match analysis warned that a single converted high-quality chance from Dundee United would materially cut the modelled Rangers edge. That is close to exactly what happened. Rose's opener — a tap-in from Josh Randall's low cross after a Rangers defensive error — put the favourites behind and forced them to chase the game for the remainder of the night.
Both teams to score
The model's strongest secondary signal was a 67.2% probability that both teams would score. Both did.
Volume versus quality
The pre-match read identified a tension in the attacking data: Rangers generating high attacking volume without matching chance quality, United creating fewer but cleaner openings.
The match followed that script closely. Rangers produced sustained pressure after the interval — Youssef Chermiti was denied twice in quick succession before Aasgaard's equaliser — but the clearest opening of the entire game fell to United. Randall was played clean through on goal late on and sent his finish straight at Ivor Pandur.
Had that chance been converted, the "one high-quality moment" scenario the model flagged would have won United the match outright, and this review would be assessing a far bigger upset.
The lineup caveat
The model's confidence was rated Low partly because team news was unavailable when the prediction was produced. That caveat proved material. Derek McInnes named four summer signings in his first competitive Rangers starting XI and handed Lawrence Shankland the captaincy following James Tavernier's departure, while Jim Goodwin fielded six of his own summer arrivals. That is a level of squad turnover no pre-season model can fully price in.
The Wider Context
For Rangers, the frustration extends beyond one result: this was the fourth consecutive season they have failed to win their opening league fixture. In a division where last season's title race went to the final day, dropped points in round one are not trivial — the pre-match article made exactly that point, and it applies to Rangers first.
For Dundee United, a point against one of the league's biggest clubs — with a near-miss at all three — is an encouraging marker after last season's seventh-place finish.
United travel to Hearts on August 9, while Rangers turn their attention to European qualifying in Poland before their next league assignment.
The Scorecard
An honest summary of the prediction's performance:
- Result call: Wrong — the least likely outcome occurred
- Both teams to score: Right
- Chance-quality read: Right — United created the cleanest openings
- Volume-without-quality read on Rangers: Right — pressure, but few clear chances converted
- Low confidence flag: Justified
- Lineup uncertainty flag: Justified
Final Word
Probabilistic predictions should be judged over many matches, not one. A single draw does not invalidate a 57.6% projection — but nor should a wrong headline call be quietly buried.
The fairer conclusion is this: the model was wrong on the result and right about the shape of the game. It identified Rangers as a fragile favourite, flagged the precise mechanism by which the edge could evaporate, and attached low confidence for reasons that proved material on the night.
The season is one match old. The sample starts here.
See how these numbers translate to real fixtures on our live match predictions page, updated daily throughout the season.