A Rebuilt Page, a Perfect xG Run, and the Bracket the Numbers Would Build

Three of the four Round-of-32 groups are done, we have rebuilt the whole World Cup page, and our pure-xG model is still unbeaten on who advances. Plenty to get to.

The World Cup page just got a rebuild

We tore up the World Cup page and rebuilt it. Every game now opens with a consensus bar showing exactly who the model and the panel have going through, color-coded pick chips so agreement and dissent jump out at a glance, and our model's own call sitting right on top of the four personas. We also added two live scoreboards: how accurate each persona has been, and how our model's ROI stacks up against the betting market. All of it tracked in the open, nothing hidden. Go have a look.

The xG picks stay perfect on who advances

Our pure-xG model, the one that ignores reputation and reads only each team's last three games of expected goals, called all three knockout ties yesterday. Belgium, England and the United States all went through, exactly as the numbers said. That is three more correct and keeps its run on who advances spotless through the knockouts.

One honest note to keep it real: England needed a Kane double and beat DR Congo 2-1, so the low-scoring under our xG read leaned toward missed even though the winner did not. Belgium against Senegal, meanwhile, was the goal-fest the model promised, five goals in a 3-2 thriller.

Here is what the xG model likes today:

Spain to dismantle Austria. A projected 2-0 and 87 percent to advance, the most lopsided read on the entire board.

Portugal past Croatia, but only just. A tight 1-0, the closest of the three.

Switzerland over Algeria, 2-1.

What if xG picked the rest of the tournament?

Here is the fun one. We ran every remaining tie through nothing but recent xG form and let it build a whole bracket, and it looks nothing like the favorites' one.

It crowns Spain, not Argentina. France, the finalist in our full model, gets knocked out in the quarters. And the headline shock: Canada, who scraped into the knockouts as one of the best third-placed teams, march all the way to the xG final four. Canada have quietly been one of the most dominant sides in the tournament by the underlying numbers, they just have not had the reputation to match.

The reason is the thread we keep pulling. xG rewards teams for the quality of the chances they create, not for the goals they finish. So the clinical favorites who have been scoring above their chances, France chief among them, slide down, and the sides who create plenty and get overlooked, Canada and Spain, rise to the top. On results it is Argentina's tournament. On the underlying play, xG says it is Spain's, and Canada are the story nobody saw coming.

Today's board, and how everyone is doing

For today's three ties, there is no argument. Our model and all four personas line up on the same side: Spain, Portugal and Switzerland to advance.

And the season-long tracking, through 22 graded games, is where it gets interesting:

• The Sharp leads the panel at 70 percent across every market, with The Analyst right behind at 68.
• Our own model is the totals machine. It is hitting 77 percent on over/unders for a plus-38 percent return against the market, while the market itself is running negative there. That gap is the whole point of doing this.

Every pick, every persona, and the full accuracy and ROI tables are live on the World Cup page.

More tomorrow, when the last of the Round of 32 is settled and the bracket gets a little smaller.