Our World Cup Final Four, the Quarterfinal We Cannot Call, and a Co-Main Decided on the Mat

No games today, which for once is a gift. It lets us plant the flags before kickoff instead of after, and say out loud who our model thinks walks out of the quarterfinals and into the final four. There is UFC 329 to get into as well, since the co-main is the most one-sided read on the whole card, and one baseball game tonight where our own model has talked itself into something we would not go near.

The final four, before a ball is kicked

Four quarterfinals, and the model treats three of them as chalk.

France against Morocco is the least suspenseful of the lot. The panel has France between 70 and 72% across the board, built on an attack rated at 2.81 against a Morocco side that our scout says is missing its main goal threat, striker Saibari, who turned up with a hamstring. France have the depth to cover their own knock to Tchouameni, and Morocco, a lovely story through the group stage, do not have the attack to keep pace without him.

Spain against Belgium is nearly as clean. Spain sit between 62 and 75% depending on which forecaster you ask, and the number that matters is the defensive one: they concede at a 0.37 rate that is the best left in the bracket, and Belgium, our scout notes, have lost the engine in their midfield. The diehard, who romances an upset more readily than the others, still lands on Spain at 62%.

Argentina against Switzerland is the closest thing to a formality. Every forecaster has Argentina between 78 and 80%. Switzerland's breakout scorer Manzambi is out with a knee, which the scout puts bluntly as their whole attack gone, and while Argentina looked heavy-legged in the group stage and leak more than a title favorite should, they have Messi and Switzerland have no answer for him.

Which leaves the one we cannot call.

Norway, England, and the game the panel refuses to agree on

The last time our whole panel agreed on a knockout game involving Norway, Norway made all four forecasters wrong and put Brazil out. So there is a grim comedy in what the model has done with Norway against England, which is refuse to pick. The analyst has Norway at 58% and the sharp has them at 53%, both leaning to the home side and the better underlying attack, Norway's 2.36 against England's 1.72. The diehard and the scout both flip it to England at 68%, and the scout has a reason the numbers cannot see, which is that the Norway camp has been hit by a sickness bug, and that even if Haaland plays he is dragging a drained team behind him. England have knocks of their own but nobody throwing up in the team hotel.

After what Norway did to Brazil, the fact that our model will not plant a flag on them here might be the most honest thing it has done all tournament. If forced, the futures lean Norway by a whisker, an 8.1% title price against England's 6.7%, and we would not want much riding on it.

And the trophy

Add it up and our final four reads France, Spain, Argentina, and whoever survives the Norway and England scrap. As for who lifts it, the futures have two teams clear of the field: Argentina at 28.0% and Spain at 25.8%, with France a distant third at 17.3%. It is Argentina, barely, over Spain, the two sides the model would bet to still be standing when the lights go down. We have been wrong about Norway from the round of 16 onward, so take the confidence for what it is worth, but the shape of what is left says the two best teams in the tournament are on a collision course and everyone else is playing to spoil it. We keep every one of these calls, the clean ones and the Norway ones, on the accuracy page, and the full board lives on the World Cup page.

The co-main nobody is really debating: Saint Denis vs Pimblett

Earlier this week we laid out why the numbers have bad news for Conor McGregor in Saturday's main event. The co-main under it, Benoit Saint Denis against Paddy Pimblett, might be the most lopsided read on the entire card, and it is worth walking through because on paper it does not look that way.

Start with the tape, because it flatters Pimblett. He carries the better record at 23 and 4 to Saint Denis's 17 and 3, they stand at the same 73 inch reach, and they throw at almost the same clip, Saint Denis landing 5.62 significant strikes a minute to Pimblett's 5.49, both defending strikes at an identical 42%. If this fight stayed standing it would be a real fight, and it will not stay standing.

Here is the chasm the record hides. Saint Denis lands 4.19 takedowns per fifteen minutes and defends them at 72%. Pimblett lands 0.69 and defends at 44%. Once the fight hits the floor it stops being the same sport, and both of our forecasters see it from opposite directions and arrive at the same corner. Vega, which reads only the data, points to Saint Denis's grappling rating against Pimblett's porous takedown defense and the submission Pimblett already coughed up to King Green, and calls Saint Denis at 78% by submission. Scoop, which works the news cycle, has heard that Pimblett's cut to make 155 is a brutal one this time, north of 35 pounds, and reasons that a drained gas tank against a pressure wrestler ends one way. Scoop lands on Saint Denis at 80%, also by submission.

When our data-only model and our news-reading model agree on the fighter, the method, and roughly the number, having gotten there down two completely different roads, that is the cleanest signal the card offers. There is no split to argue over here and no gap from the betting market to chase. It is a bad stylistic matchup meeting a bad weight cut. If you want to poke at the grappling numbers yourself, every takedown and control-time split we leaned on is in the UFC Explore tool for Pro members.

The baseball game we do not trust our own model on

One game tonight is worth flagging precisely because our model is wrong about it, or at least we are fairly sure it is, and saying so out loud is the entire point of grading in public.

The line-blind model makes Colorado at the Dodgers a coin flip, 52% to the Rockies, while the market has Colorado at 32%. A twenty point gap like that usually makes us look twice in a good way. This one made us look twice and wince. Colorado's listed starter, Gabriel Hughes, shows a 0.00 ERA, which reads like an ace until you notice it is built on exactly three innings in a single relief appearance, and that he is starting at all only because the Rockies' rotation has been gutted. Colorado are without ten players, Kris Bryant, Brenton Doyle and Jose Quintana among them, they are 47 and 69, and they are walking a reliever into a Dodgers team that is 75 and 41. Los Angeles's own starter, Roki Sasaki, has been genuinely poor lately, a 5.40 ERA with nineteen earned runs across his last four starts, and that is the loose thread the model has pulled on. It is not enough to move a game this lopsided. This is the kind of number we flag rather than follow, and you can watch it grade itself out on the game page tonight.