Make decisions with analysis.

Free quantitative probabilities.
Transparent methodology. Measurable performance.

We don't sell promises. We show probabilities.

WHY THE EDGE HOUSE

Transparent methodology

Clear models. No black boxes.

Historical evidence

Based on real data, not opinions.

Measurable performance

Track results. Evaluate over time.

HOW IT WORKS

From data to evidence.

  1. 01

    Model

    We estimate probabilities with a Poisson model of goals per team, built on historical data.

  2. 02

    Publish

    We publish each probability before the match. It is recorded and cannot be changed.

  3. 03

    Track

    We record every outcome, right or wrong, and measure performance.

  4. 04

    Learn

    We evaluate the record. The model only changes by versions, with human approval.

What we said. What happened.

A transparent record of our published predictions with real outcomes.

Real matches. Real results. No opinions.

285
Published
Across 11 leagues
60
Resolved
So far: La Liga only
53.3%
Hit rate
Favorite was right
0.278
Brier Score
n = 60 · Lower is better
Model0.278
Random0.333

Equivalent on the 0–2 scale: 0.555 (random: 0.667)

Model poisson_v1.0 · Oct 5, 2026

HISTORICAL EVALUATION

Probabilities, evaluated.

We evaluate the model against real historical results in a walk-forward backtest: each prediction only uses data available before that match.

Learn more about our methodology →
Historical backtest · not the live track record

Calibration plot

La Liga · 2020-21 → 2024-25 · 1,900 matches · walk-forward

0.00.00.20.20.40.40.60.60.80.81.01.0Predicted probabilityObserved frequency5–10% · n=120 · 7.6% → 10.0%10–15% · n=273 · 12.9% → 13.6%15–20% · n=443 · 17.6% → 15.3%20–25% · n=957 · 22.9% → 21.6%25–30% · n=1363 · 27.3% → 28.1%30–35% · n=566 · 32.2% → 30.6%35–40% · n=391 · 37.6% → 38.4%40–45% · n=370 · 42.4% → 43.0%45–50% · n=327 · 47.5% → 47.1%50–55% · n=231 · 52.3% → 55.0%55–60% · n=226 · 57.2% → 57.1%60–65% · n=154 · 62.4% → 63.6%65–70% · n=103 · 67.2% → 74.8%70–75% · n=80 · 72.3% → 72.5%75–80% · n=48 · 77.8% → 79.2%80–85% · n=35 · 81.8% → 71.4%ModelPerfect calibration

Each match contributes three points (home, draw, away), grouped in 5% bins. Bars are 95% intervals and dot size reflects the number of points. Bins with fewer than 30 points are omitted.

These are historical results, not published predictions. See the live track record

View data
BinPointsPredictedObserved
5–10%1207.6%10.0%
10–15%27312.9%13.6%
15–20%44317.6%15.3%
20–25%95722.9%21.6%
25–30%1,36327.3%28.1%
30–35%56632.2%30.6%
35–40%39137.6%38.4%
40–45%37042.4%43.0%
45–50%32747.5%47.1%
50–55%23152.3%55.0%
55–60%22657.2%57.1%
60–65%15462.4%63.6%
65–70%10367.2%74.8%
70–75%8072.3%72.5%
75–80%4877.8%79.2%
80–85%3581.8%71.4%

Built on data. Not opinions.

The probabilities published by The Edge House are estimates generated by our quantitative models and historical data. They do not represent certainty and do not guarantee outcomes.

Football is a high-variance sport. Injuries, tactical decisions, lineups and other factors can alter any prediction.

Unlock the full model, for free

Sign up for free now and unlock all quantitative probabilities.

Limited time · No card · No cost