How to Use Edge Model

Edge Model is a free Bayesian football prediction tool. It tracks the probability that each team wins their league or tournament, updates those probabilities after every result, and shows you where the model disagrees with bookmaker odds. You do not need to sign up to use any of it — just open the app and start exploring.

This guide covers everything from loading your first league to reading the market edge pills on fixtures. It takes about five minutes end to end.

Step 1 — Open the app

Go to edgemodel.tech in any browser. No account, no download, no installation. The app loads with the FIFA World Cup 2026 tournament by default. If the league season has already started and you want current standings and fixtures, switch to league mode in the next step.

Step 2 — Load a league

Click 📊 League in the header. The league switcher shows all available competition presets as cards. Competitions with existing saved progress display a green ▶ Resume badge — if you are returning after the season has already started and results have been ingested, this means your standings are already current.

• 1
  Click 📊 League in the header

  This opens the competition switcher. If you are already in league mode, this re-opens the switcher so you can change the competition.
• 2
  Click a preset card to stage it

  Cards without a Resume badge start from calibrated pre-season priors (historical goals data). Cards with a Resume badge restore previously recorded results.
• 3
  Click Load League

  The Gamma-Poisson model initialises. Championship probabilities, standings, and the match predictor all populate with data from that competition.

Built-in league presets include the Premier League, La Liga, Bundesliga, Serie A, and Ligue 1, with results ingested automatically every day. The NBA, NFL, SA Betway Premiership, and Botswana Premier League are also available.

Step 3 — Read the championship probability chart

The horizontal bar chart below the stat cards shows P(Champion) — the model's current probability that each team wins the league. Teams are sorted in descending order. The gold bar is the outright leader; blue bars are the next tier.

Each bar includes 95% credible interval error bars. Early in the season, with few results recorded, the intervals are wide — the model is uncertain because it has not seen much evidence yet. As results accumulate and the Gamma posteriors tighten around each team's true attack and defense strength, the intervals narrow. A team with a narrow interval is one the model is confident about; a wide interval signals uncertainty that could swing either way.

The key insight from the chart is that league position and championship probability are not the same thing. A team sitting third with several games in hand against weaker opposition can have a higher P(Champion) than the current table leader. The model projects probability over all remaining fixtures, not just the current standings snapshot.

Step 4 — Read the standings table

The full standings table lives below the chart. It shows the standard columns — played, won, drawn, lost, goals for, goals against, goal difference, and points — plus a P(Champion) column. Teams are sorted by points; ties are broken by championship probability, not goal difference.

For American sports presets (NFL), the draw column relabels as T (ties). For sports with no draws (NBA), the draw column is hidden entirely.

Step 5 — Use the match predictor

The Match Predictor panel lets you analyse any head-to-head matchup in the loaded league. Select a home team and an away team from the dropdowns — both update automatically after each new result, so the probabilities always reflect the current model state.

The predictor outputs three things:

• W
  Win / Draw / Loss probabilities

  Derived from the Skellam distribution over the current Gamma-Poisson posteriors. The expected goal rates for each side are shown alongside the probabilities.
• Z
  Decisive vs. too close to call

  Rather than a raw probability threshold, the predictor uses a Skellam Z-score (expected goal margin ÷ natural variance) to determine whether a result is decisive. A Z-score ≥ 0.5 shows a clear winner; below 0.5 shows "Too close to call" with the slight favourite noted.
• E
  Market edge (if odds are available)

  When bookmaker odds are loaded, the panel shows the difference between the model's win probability and the market's implied probability for each outcome. A positive edge means the model thinks the outcome is more likely than the market does.

Step 6 — Check upcoming fixtures for edge

Switch to the Fixtures tab. This panel lists all scheduled matches for the loaded competition with the model's predicted outcome for each one.

When the model's win probability for an outcome diverges from the bookmaker's implied probability by five percentage points or more, a coloured edge pill appears on the fixture card. A pill labelled +8% Home means the model thinks the home team is 8pp more likely to win than the odds imply. Negative edge (odds are more generous than the model) is shown in a different colour.

The edge pills load in the background after the fixture list renders — you will see a brief shimmer animation while they fetch. Fixtures without market coverage show no pill; the model prediction is still there.

What automatic ingestion means for you

For the five major European leagues, you do not need to record results manually. A background scheduler fetches finished match scores from football-data.org every morning (and multiple times across the day on weekends to capture Saturday and Sunday results). The Gamma posteriors update automatically and all open tabs receive the new state via Server-Sent Events — no page refresh required.

A Last updated timestamp in the standings header and the fixtures panel tells you exactly when the model last incorporated new results. If you open the app an hour before the weekend ingestion run, you will see that timestamp and know the current state reflects Friday's results.

Creating a free account

You do not need an account to use any of the prediction features. An account adds two things: your preferences (last loaded league, favourite teams, active tab) sync to the server so they restore on any device, and you can participate in Group Picks challenges against friends.

Register by clicking the account icon in the header. Registration takes under a minute — email, password, and consent to the privacy policy. Once signed in, every preference change is synced automatically in the background.

Tournament mode

The app works equally well for knockout tournaments. Switch to tournament mode using the 🏆 Tournament button in the header. Preset tournaments include FIFA World Cup 2026 and WAFCON 2026. The underlying model is a Dirichlet distribution — each result transfers Bayesian weight from the eliminated team to the winner, so championship probability concentrates on fewer and fewer remaining teams as the bracket progresses.

Tournament mode includes a Bracket Simulator where you can click through any hypothetical path and see how each choice reshapes championship probabilities, without touching the live state.