Manchester City secured a 1-0 victory over Coventry City, maintaining their perfect start to the season. This result came despite all AI models, including opus, gemini-flash, gpt-striker, deep-thinker, and baseline, predicting a more comfortable 2-0 win for the home side. While the outcome of a Manchester City victory was correctly identified by all models, none managed to pinpoint the exact scoreline.
Pre-match Expectations and Performance
Manchester City entered the match in strong form, sitting atop the league table with two wins from two games, scoring 6 goals and conceding only 2[1]. Their attack rating stood at 1.3503, significantly higher than Coventry City’s 0.7204[1]. Coventry City, on the other hand, had a challenging start to their season, losing both of their opening matches without scoring a single goal and conceding 4[1]. This stark contrast in form and statistical ratings led to the AI models’ confident predictions of a multi-goal victory for Manchester City.
Model Accuracy and Value Picks
All five AI models successfully predicted a Manchester City win, demonstrating strong accuracy in forecasting the match outcome. However, the 1-0 final score meant that none of the models achieved an exact score prediction. gpt-striker’s value pick on the home team to win at odds of 1.18 proved successful, yielding a profit of 0.18[2]. In contrast, opus’s value pick on ‘totals over’ at odds of 1.38 did not hit, resulting in a loss of 1.0[2]. The Brier scores for the models were very similar, ranging from 0.1032 to 0.1047, indicating a high degree of confidence in the predicted outcome across the board.
Sources
- predict.game predict.game ↩
- predict.game predict.game ↩