Italy–Germany

1–1

HT 0–0

Kick-off: Saturday, 4 June 2022 at 18:45
Status: Match Finished
Venue: Stadio Renato Dall'Ara, Bologna

What the model said, and what happened

Ended level — we gave the draw 27%double chance: correct

This verdict goes into the public record unchanged; we do not filter out the ones we get wrong. The model's track record

Referee

Srdan Jovanovic, Serbia→

There's no card data for this referee from before this match yet. We don't invent an average; it will appear here once the data exists.

Model prediction

Why this prediction?

The model puts the most likely outcome at 45.9%: “Italy to win”. Second comes “a draw” at 27.4%.
The highest double chance value is 73.3% (1X — Italy win or draw).
These values were computed from 2,805 past matches.

This is a prediction, not a guarantee; a probability is not a certainty.

These sentences are built by a fixed template; every number in them comes straight from the model's computation, none invented while the text was written.

Italy to win45.9%

Draw27.4%

Germany to win26.7%

When the model has said 45–50% in the past, 2,111 of 4,488 predictions were correct (47.0%). Accuracy record →

Double chance

1X — Italy win or draw

What these probabilities rest on

Expected goals comes out of a team's attack weighed against its opponent's defence. A strong attack is damped when it meets a strong defence — which is why comparing the ratings on their own is misleading.

Italy

Attack: Italy score above the league average (43% more).

Opponent's defence: Germany concede 33% below the league average — a defence that seriously brakes even a strong attack.

Germany

Attack: Germany score 100% more than the league's average team — one of the sharpest attacks in the league.

Opponent's defence: Italy concede 51% below the league average — a defence that seriously brakes even a strong attack.

Week 1 of the season. This prediction was computed from previous seasons. Summer transfers, budget changes and squad rebuilds aren't in the ratings yet — the model can only see that a team has been rebuilt once new results come in.

We measured this and we're not hiding it: in the opening week of past seasons the model came in 3.3 points below the accuracy it claimed (n=220) — for the rest of the season it beats its own claim. The tendency points this way, though it is not proof on its own: in the opening week it slightly over-calls home wins (said 42.7%, got 39.5%) and slightly under-calls draws (said 26.1%, got 28.2%).

This sample is small and not statistically established, which is why we do not correct the model for it. The correction we tried (pulling probabilities toward the league average) didn't win on holdout, so it wasn't added. Read opening-week predictions with a little more caution.