Predict RavenPortugal to win in 90 min, ~43%
The evidence
Portugal rate higher on form and results — 1989 vs 1912.
Portugal create more — 1.61 vs 1.30 expected goals a game.
Portugal give up less — 0.74 vs 1.31 expected goals against a game.
Portugal keep the ball more — 52.7% vs 49.5% of the game.
Croatia press higher up the pitch more often — 5.7% vs 4.3% of play.
Dead level — Portugal and Croatia match up here.
Portugal sit deeper to defend more — 20.1% vs 19.4% of play.
Croatia cover more ground at high speed — 6.1 km vs 5.5 km.
How the nine models read it
Blending all 8 models and correcting for known group-stage biases, Portugal is the call here at about 43%.
Models’ consensus+17.6 pts → Portugal
Portugal are a slight favourite to beat Croatia (54% to win), though a tight low-scoring draw is very much in play (30.8%).
Attack vs defence+38.7 pts → Portugal
Portugal are favoured to beat Croatia, winning roughly 47 times out of 100 to Croatia's 28.
Chance-creation form+6.8 pts → Portugal
Portugal and Croatia look evenly matched, with no side meaningfully on top.
Attack vs defence+26.2 pts → Portugal
Portugal is a slight favourite here with about a 45% chance to win, partly because the other side ran a heavier physical load and is likelier to tire.
Physical load+0.3 pts → Portugal
Portugal are clear favourites to beat Croatia, mostly on cleaner attacking play.
Breaking through the defence
Portugal is the clear favourite (68%) — its passing play is the harder one to break down.
Passing structure
Pulling every angle together, Portugal is narrowly the side to back here, at about a 48% chance to win.
Model agreement+24.1 pts → Portugal
Portugal and Croatia look evenly matched, with Portugal a hair ahead on the day.
Overall quality+7.4 pts → Portugal
Method
Market-blind: equal-weight linear (arithmetic) opinion pool of all 8 model views (seven base models + the stacked ensemble), then MCBoost multicalibration (alpha=0.15, 5 correction round(s)) learned on the group stage across overlapping style/strength subgroups, replayed on this fixture and bounded to within 12pp per outcome of the consensus, then a knockout 90-minute draw calibration (target ~31% draws, the 2022 Qatar knockout rate) lifting the draw on even ties. No betting or market-implied probabilities were used at any stage.
Forecast locked 2026-06-29 — before any Round-of-32 match kicked off.
Market-blind: the eight model views are pooled into one consensus, then a light bias correction learned on the group stage is applied. No betting or market-implied probabilities were used at any stage.