Premier League · Gameweek 5 Verdict
Gameweek 5 Verdict: AI gets 5 from 10 as upsets strike
Published 21 September 2026 · FootyPulse AI Analysis
The AI called five winners correctly from ten Gameweek 5 fixtures, matching the hit rate but falling short of consistency. Season accuracy now sits at 38% (18 from 47), a figure that underlines the Premier League's stubborn refusal to behave predictably. From Brentford's shock demolition of Chelsea to Nottingham Forest's home collapse, the weekend served up chaos in equal measure to logic.
The scoreboard
Five correct calls from ten fixtures gives Gameweek 5 a 50% success rate—passable, but hardly cause for celebration. The season tally stands at 18 correct from 47 matches, a 38% accuracy that reflects the brutal variance of top-flight football. The model is learning, but the Premier League remains a classroom full of pop quizzes.
Biggest upset
Nottingham Forest vs Coventry City delivered the gameweek's most painful miss. The AI backed Forest with 68% confidence, only to watch Coventry leave the City Ground with a 0-1 victory. A Championship side dismantling a Premier League host at nearly 7-in-10 pre-match probability? That's the kind of result that makes you question everything—or at least recalibrate the away-team discount.
Best correct call
Brentford's 3-0 annihilation of Chelsea was the model's finest moment, despite entering the fixture as the underdog. The AI gave the Bees just 38% to win—less than a coin flip—yet they delivered a statement performance that exposed Chelsea's fragility. When your least confident correct call involves a three-goal margin, you take the win and ask questions later.
Every fixture
- Brentford 3-0 Chelsea: AI called Brentford (38%), correct ✓
- Tottenham Hotspur 2-3 Aston Villa: AI called Spurs (58%), incorrect ✗
- Brighton & Hove Albion 3-0 Arsenal: AI called Arsenal (48%), incorrect ✗
- Newcastle United 2-1 Hull City: AI called Newcastle (42%), correct ✓
- Everton 1-0 Ipswich Town: AI called Everton (58%), correct ✓
- Nottingham Forest 0-1 Coventry City: AI called Forest (68%), incorrect ✗
- Leeds United 0-0 Crystal Palace: AI called Leeds (64%), incorrect ✗
- Manchester City 5-3 Sunderland: AI called City (68%), correct ✓
- Bournemouth 0-1 Liverpool: AI called Liverpool (50%), correct ✓
- Fulham 1-1 Manchester United: AI called United (41%), incorrect ✗
Closing thoughts
Gameweek 5 was a reminder that narrow probability margins mean everything. The model's most confident calls—Forest and Leeds, both at 64%+ —failed to land, while the 38% Brentford punt paid off spectacularly. Football doesn't reward certainty; it rewards getting the variance right. As the season progresses, these lessons accumulate. See the pre-match preview to understand where the model's head was before kick-off.
Half-time re-predictions this gameweek
At half-time the AI re-reads each match and re-predicts. Here is how those calls did against the pre-match calls, on the same set of Gameweek 5 matches.
Pre-match
5/10
50% correct
At half-time
6/10
60% correct
Where half-time changed the call
Brentford 3–0 Chelsea
Pre-match: Brentford ✓ · Half-time: Chelsea (1–2) ✗
Why the call changed
- Shots on target 2 v 2
- Possession 46% v 54%
Tottenham Hotspur 2–3 Aston Villa
Pre-match: Tottenham Hotspur ✗ · Half-time: Aston Villa (1–2) ✓
Why the call changed
- Shots on target 4 v 3
- Possession 56% v 44%
Brighton & Hove Albion 3–0 Arsenal
Pre-match: Arsenal ✗ · Half-time: Brighton & Hove Albion (3–1) ✓
Why the call changed
- Shots on target 3 v 1
- Possession 47% v 53%
This is a single gameweek — far too small a sample to read anything into, in either direction. See the running pre-match vs half-time comparison on our accuracy page for the season picture.
Share this verdict:
Prediction figures in this article reflect the data available at the time of publication and may have shifted since. See the live Gameweek 5 fixtures for the current numbers on each match.
See how the AI called it before kickoff:
Read the Gameweek 5 preview →