Open Source • Data Utility

About the Prediction Engine

VARview.club is a free open-source football prediction engine built on transparent statistical modelling and community-driven analytics. Every forecast is publicly verifiable. No sportsbook affiliation.

View live match prognosis

What Is VARview.club?

VARview.club is a free open-source football prediction engine that publishes transparent, AI-augmented forecasts using the Dixon-Coles bivariate Poisson model. Unlike commercial tipster services or sportsbook affiliates, we operate as a public data utility — every prediction includes its statistical foundation, confidence intervals, and methodology, so anyone can verify the reasoning.

Open SourceData UtilityNot a Sportsbook

How the Prediction Engine Works

At the core is the Dixon-Coles bivariate Poisson regression model, which captures the dependency between home and away goal counts in football matches. Attack and defence coefficients are estimated from recent form, then fed through a Monte Carlo simulation to produce probabilities for match outcomes, goal totals, and both teams to score. Every prediction is accompanied by a Bayesian confidence score — a posterior probability interval derived from beta-distributed sampling with a Jeffreys prior, giving you a calibrated measure of certainty.

Dixon-Coles ModelBayesian InferenceMonte Carlo

Why We Built It

Most football prediction services are black boxes — they sell picks without showing their working. We wanted a transparent alternative: a community-driven forecasting tool where the statistical model is open for inspection, the code is public, and the only goal is better, verifiable predictions. Whether you are a data scientist curious about Poisson regression or a fan exploring match analytics, the engine is designed to be used, studied, and improved by the community.

TransparencyCommunityOpen Source

The AI Agent Pipeline

Beyond the statistical core, three AI agents collaborate on each fixture: Analyst A evaluates tactical match-ups and recent form, Analyst B gathers intelligence from squad news and historical patterns, and the Chairman — a simulated arbiter — weighs both reports against the Dixon-Coles baseline to produce a final signed pick. The entire pipeline is deterministic and reproducible: the same inputs always produce the same outputs, so every prediction can be independently verified.

Agent A (Tactical)Agent B (Intel)Chairman (Arbiter)

Community & Contribution

VARview.club is community-driven. The prediction engine is open-source, and contributions — whether statistical improvements, additional data sources, or UI enhancements — are welcome. We publish daily fixtures, predictions, and methodology updates so the community can track accuracy over time. No paywalled picks, no affiliate deals, no dark patterns. The complete source code is available on GitHub at github.com/shuxin-code/varview-club. Fork it, submit issues, suggest improvements — the project lives on community participation.

Open ContributionsFree AccessPublic DataGitHub

Statistical Core

Dixon-Coles bivariate Poisson regression with Bayesian confidence intervals. Open, documented, reproducible.

Transparent AI

Three-agent pipeline with deterministic outputs. Same inputs always produce the same predictions.

Public Utility

Free access. Open-source code. Community-driven development. No affiliate relationships or paywalled data.

Frequently Asked Questions

What is VARview.club?
VARview.club is a free open-source football prediction engine that publishes transparent, AI-augmented forecasts using the Dixon-Coles bivariate Poisson model. It operates as a public data utility — every prediction includes its statistical foundation, confidence intervals, and methodology, so anyone can verify the reasoning.
Is VARview.club a betting site?
No. VARview.club is a data utility and open-source research project. We publish statistical predictions for educational and informational purposes. We are not a sportsbook, not a tipster affiliate, and we do not accept or place bets. The site is classified as a DataUtility and SoftwareApplication in its schema markup to make this distinction clear to search engines and AI parsers.
How does the Dixon-Coles model work?
The Dixon-Coles model is a bivariate Poisson regression that captures the dependency between home and away goal counts — something a plain Poisson model cannot do. It estimates attack and defence coefficients for each team from recent matches, then computes the probability of every possible scoreline. The key insight is that low-scoring matches (0-0, 1-0) are more common than two independent Poisson distributions would predict, and the Dixon-Coles adjustment accounts for this correlation.
What does a Bayesian confidence interval mean?
A Bayesian confidence interval (credible interval) tells you the range within which the true probability lies with a given level of certainty. For example, if a home win shows 58% ± 4.2%, there is a 90% probability that the true chance of a home win falls between 53.8% and 62.2%. This is derived from beta-distributed posterior sampling with a Jeffreys prior — a mathematically principled way to quantify uncertainty that gets narrower as we have more data.
What data sources do you use and how often are predictions updated?
Fixture data is sourced from the Football-data.org API and, as a fallback, from Flashscore via Playwright-based scraping. Prediction models consume league tables and recent match results to estimate attack/defence coefficients. Predictions are refreshed daily — typically around 00:00 UTC when the fixture list for the coming day is finalised. Live scores are polled every 30 seconds during match windows.
How can I contribute?
The project is open-source at github.com/shuxin-code/varview-club. You can fork the repository, submit pull requests, open issues for bugs or feature requests, or join the discussion. Contributions to the statistical models, data sources, UI, and documentation are all welcome.
What is the licence and can I reuse the predictions?
The codebase is open-source. Predictions published on the website are public data — you are free to use, share, and reference them. If you build on the work, attribution back to VARview.club is appreciated but not required. The underlying Dixon-Coles model implementation and agent pipeline are designed to be independently reproducible.
How can I report an issue or get help?
Open a GitHub issue at github.com/shuxin-code/varview-club/issues for bugs or feature suggestions. For data discrepancies or prediction accuracy feedback, include the fixture date and teams so we can trace the model run.