10→ Football

Chances of Every Game

Home win, draw or away win: the chances of every game left according to our goals model. Below, the test of how good it is, with all the numbers.

Sunday, November 1

  • Belgrano vs. BarracasBeing played now

    Chances hidden while the game is on: they were computed before kick-off.

  • Newell's vs. UniónBeing played now

    Chances hidden while the game is on: they were computed before kick-off.

  • Lanús vs. Gimnasia (M)20:45

    • 46%Lanús wins
    • 28%Draw
    • 27%Gimnasia (M) wins

    Expected goals: Lanús 1.3–0.9 Gimnasia (M)

Monday, November 2

  • Platense vs. Central Córdoba19:00

    • 47%Platense wins
    • 25%Draw
    • 28%Central Córdoba wins

    Expected goals: Platense 1.4–1.1 Central Córdoba

  • Defensa vs. Independiente19:00

    • 46%Defensa wins
    • 27%Draw
    • 27%Independiente wins

    Expected goals: Defensa 1.3–0.9 Independiente

  • Atl. Tucumán vs. Argentinos21:15

    • 35%Atl. Tucumán wins
    • 28%Draw
    • 37%Argentinos wins

    Expected goals: Atl. Tucumán 1.0–1.1 Argentinos

  • Ind. Rivadavia vs. Banfield21:15

    • 39%Ind. Rivadavia wins
    • 25%Draw
    • 35%Banfield wins

    Expected goals: Ind. Rivadavia 1.3–1.2 Banfield

The weekend of November 8 (day and time to be confirmed)

  • Aldosivi vs. RiverTBC

    • 31%Aldosivi wins
    • 27%Draw
    • 42%River wins

    Expected goals: Aldosivi 1.1–1.3 River

  • Unión vs. VélezTBC

    • 35%Unión wins
    • 24%Draw
    • 41%Vélez wins

    Expected goals: Unión 1.3–1.5 Vélez

  • Talleres vs. Estudiantes (LP)TBC

    • 40%Talleres wins
    • 25%Draw
    • 35%Estudiantes (LP) wins

    Expected goals: Talleres 1.4–1.2 Estudiantes (LP)

  • San Lorenzo vs. PlatenseTBC

    • 39%San Lorenzo wins
    • 28%Draw
    • 33%Platense wins

    Expected goals: San Lorenzo 1.1–1.0 Platense

  • Central vs. Estudiantes (RC)TBC

    • 55%Central wins
    • 24%Draw
    • 21%Estudiantes (RC) wins

    Expected goals: Central 1.6–0.9 Estudiantes (RC)

  • Instituto vs. Newell'sTBC

    • 44%Instituto wins
    • 27%Draw
    • 29%Newell's wins

    Expected goals: Instituto 1.3–1.0 Newell's

  • Independiente vs. RiestraTBC

    • 43%Independiente wins
    • 29%Draw
    • 28%Riestra wins

    Expected goals: Independiente 1.1–0.9 Riestra

  • Gimnasia (M) vs. Ind. RivadaviaTBC

    • 42%Gimnasia (M) wins
    • 26%Draw
    • 32%Ind. Rivadavia wins

    Expected goals: Gimnasia (M) 1.3–1.1 Ind. Rivadavia

  • Gimnasia (LP) vs. BelgranoTBC

    • 43%Gimnasia (LP) wins
    • 27%Draw
    • 30%Belgrano wins

    Expected goals: Gimnasia (LP) 1.3–1.0 Belgrano

  • Central Córdoba vs. LanúsTBC

    • 27%Central Córdoba wins
    • 26%Draw
    • 46%Lanús wins

    Expected goals: Central Córdoba 1.0–1.3 Lanús

  • Boca vs. DefensaTBC

    • 43%Boca wins
    • 27%Draw
    • 31%Defensa wins

    Expected goals: Boca 1.3–1.0 Defensa

  • Barracas vs. RacingTBC

    • 44%Barracas wins
    • 27%Draw
    • 29%Racing wins

    Expected goals: Barracas 1.3–1.0 Racing

  • Banfield vs. Atl. TucumánTBC

    • 39%Banfield wins
    • 27%Draw
    • 34%Atl. Tucumán wins

    Expected goals: Banfield 1.2–1.1 Atl. Tucumán

  • Argentinos vs. HuracánTBC

    • 47%Argentinos wins
    • 28%Draw
    • 26%Huracán wins

    Expected goals: Argentinos 1.3–0.8 Huracán

  • Tigre vs. SarmientoTBC

    • 41%Tigre wins
    • 26%Draw
    • 33%Sarmiento wins

    Expected goals: Tigre 1.3–1.1 Sarmiento

Updated 2026-11-01 17:37 (Argentina time)

How good is the model?

Tested on 480 league games of 2025 it had not seen: each game predicted only with the results before it.

Results called
40%always picking the home side: 41%
Log loss
1.078base rates: 1.084 (lower is better)
Brier score
0.653base rates: 0.657 (lower is better)
Draws in the test
32%model expected 34%

In plain words: the model is only a little better than knowing how often home sides win, draw and lose. Argentine football is very even.

Does it say what happens?

Every outcome the model priced, grouped by its chance. If it is honest, the two bars of each group match.

Said 10% to 20% · 48 outcomes

18%
21%

Said 20% to 30% · 395 outcomes

26%
26%

Said 30% to 40% · 761 outcomes

35%
35%

Said 40% to 50% · 212 outcomes

44%
42%

Said 50% to 60% · 24 outcomes

53%
54%

model saidit happened

Fixture coloured by difficulty →How much is home advantage worth? →Turn odds into a probability →

Illustration: Chances of Every Game

Where each number comes from

  1. Attack and defenceFrom results since 2024, the model estimates how many goals each club scores and concedes.
  2. Recent games weigh moreA game played 45 days ago counts half as much as one played today.
  3. Home advantageThe home side gets 12% more expected goals, from the same fit.
  4. From goals to resultThe expected goals of each side give the chance of every score, and from that win, draw or loss.

The test in numbers

MeasureModelBase rates
Log loss1.0781.084
Brier score0.6530.657
Results called40%41% (always home)

480 games of 2025 the model had not seen. For log loss and Brier, lower is better.

Keep going

FixturesWhat each club has left, coloured by difficulty.Home advantageWhat playing at home is worth, with every game since 2024.Odds to probabilityTurn odds into a probability and see the bookmaker margin.

Questions

Why is it right so rarely?

Because the league is very even: 32% of the test games were draws. With that much balance, no model is right often; what it can do is give chances that come true on average.

What is log loss?

A score that punishes the model more the surer it was about something that did not happen. Lower is better. It lets us compare with the base rates.

Does it update itself?

Yes. We check the results every hour and the model is refitted with every new result.