A controlled away win in Salento
Juventus left Lecce with a 2-0 win in Serie A, and the result matched what the form sheet suggested. Lecce arrived on the back of a 3-0 win at Udinese, a result that [1] reported with goals for Kiolonda Gaspar, Konan N’Dri and Jamil Siebert. That win had ended a sequence of defeats at Lecce, who had lost 1-2 at home to Bologna the week before. Juventus came in after a 0-3 home defeat to Lazio, which was the sort of result that put pressure on Luciano Spalletti’s side to answer.
They did. Lecce failed to score for the fourth time in their last eight league outings, and Juventus kept a clean sheet after conceding three in their previous game. The head-to-head record gave a hint of how tight these sides can be: Juventus won 1-0 at Lecce in May 2026, and the January meeting in Turin finished 1-1. This time the margin was two goals, and the visitors never needed to chase the game.
Lecce also had selection problems. The input lists Medon Berisha and Nikola Stulic as injured, Ivan Ilic as hurt during the Bologna match, and Willem Geubbels and Paco Esteban as not fully fit. That is a lot of attacking and midfield depth to lose for a side that had scored only nine league goals before this game.
How the models graded out
Every model picked the away win, so the separation came from scorelines and confidence. Claude Keeper predicted 0-2, the exact score, and finished with a Brier score of 0.0608, the best of the group. Kimi Cool and Gemini Thunder both went for 0-1, which got the outcome right but missed the score. Their Brier scores were 0.1784 and 0.245. GPT Striker and DeepThinker both predicted 1-2, with Brier scores of 0.305 and 0.245. The baseline also predicted 0-1, but its Brier score of 0.4936 was the worst, which shows how little weight it put on the correct outcome.
The pattern is familiar. A correct outcome earns credit, but the Brier score punishes hedging, and the models that held back probability from the away win paid for it. GPT Striker’s 1-2 implied a Lecce goal that never arrived, and its probabilities left it with the weakest score among the five models that made a call beyond the baseline.
Value picks: one clean hit, one miss
Three models also logged value picks. Claude Keeper backed the away side on the head-to-head market at odds of 1.46, and it landed for a profit of 0.46. GPT Striker took the same selection at the same price and made the same 0.46. Gemini Thunder went against its own predicted outcome and picked the draw at 4.3, which missed and lost 1.00. Kimi Cool, DeepThinker and the baseline made no value pick.
The contrast is instructive. The short-priced favourite delivered a modest return, while the longer-priced draw cost a full unit. Gemini Thunder’s draw pick also sat awkwardly beside its 0-1 score prediction, since it forecast an away win while recommending a bet on a draw.
What it means
For Juventus, the 2-0 win lifts them to 7th in Serie A with 14 points from eight games (W4 D2 L2, 11 scored, 8 conceded). The week before they had lost 0-3 to Lazio, so this was a clear response. The form line now reads W, L, D, W, L over the last five, with the next test away at Genoa on 28 October. After an away win, the question is whether they can repeat it on the road in midweek.
For Lecce, the picture is tougher. They stay 12th with 9 points from eight games (W3 D0 L5, 9 scored, 14 conceded). The win at Udinese looked like a turning point, but the defeat here brought them back down. Their next fixture is away at Frosinone on 29 October, and with several injuries to work around they need goals from a squad that has produced very few. The site’s forecast for both fixtures locks 24 hours before kick-off, and the grades above show how it will be judged.
