How BetGlitch Works: From Provider Data to a Verified Record

Introduction: Smarter Predictions Through Data

The sports prediction landscape is crowded with tipsters who rely on gut feelings, outdated heuristics, and selective memory. BetGlitch takes a different approach. We take probability data from a specialist football data provider, apply our own ranking filters, and freeze selected picks before kickoff — wins and losses alike — across European competitions. We do not claim a predictive edge of our own; whether our filtering adds value is something we are still measuring in public.

One distinction runs through everything below, so it is worth stating first. A live signal is our current ranked output for a fixture and it can change at any time before kickoff. A published claim is a signal we froze, and it can never change. Most live signals are never published. Only settled published claims count towards the verified record — pending claims count towards nothing — and every published claim stays visible afterwards, including the losing ones.

Our mission is simple: replace guesswork with an auditable record. BetGlitch takes probability data from a specialist football data provider, applies its own filtering to surface a small number of candidates, and freezes each published pick before kickoff so the result can be checked afterwards. This article explains exactly how that works.

Where the probabilities come from

At the core of BetGlitch is a specialist football data provider's probability model. We do not build, train or own a predictive model, and we do not claim to. There is no second model and no averaging of opinions: one provider supplies a probability vector per market — match result, over/under 2.5, both teams to score, double chance — and we work from those numbers.

We rank those probabilities against recorded market prices. For each fixture and market we record:

Signal score: A relative ranking score, not a calibrated probability. It orders candidates; it does not state a chance of winning.
Probability gap: How far the leading outcome sits above the next one. A narrow gap means the market is close to a coin flip.
Recorded price: The exact odds, bookmaker, market and capture time behind every published pick.

We are deliberately careful about what this does and does not prove. A ranking score is not a calibrated probability, and we have not yet demonstrated that our filtering beats the market price. That is precisely why the public record exists: to test it in the open rather than assert it.

Filtering: Being Selective

Raw provider probabilities alone are not enough, so BetGlitch narrows them before anything is surfaced. These filters are selection rules, not validated predictors — they decide what we look at, and we have not shown that they improve results:

Signal-score floor: Candidates below a minimum score are not surfaced. The threshold is a selection rule we tune, not a validated accuracy boundary.
Price gating: A candidate is only considered when we hold a verified market price for the exact market being ranked, with complete provenance. Without one, value cannot be assessed at all.
Odds range: We concentrate on moderate odds. This bounds variance; it is not evidence that we price long shots correctly.
Competition-specific handling: Some competitions are excluded or handled more conservatively. These lists are maintained by hand and revised as the sample grows.
Probability gap: How far the leading outcome sits above the next one. A narrow gap means the market is close to a coin flip.

We used to publish an accuracy figure against the probability-gap filter. We have withdrawn it. It was measured on a sample that included predictions priced from the wrong market, so it was not a valid measurement of anything, and we would rather show no number than a number we cannot stand behind.

The signal score

Every candidate that passes the filters carries a signal score from 0 to 100. It ranks our relative preference among the available outcomes in one market. It is not a calibrated probability: a score of 66 does not mean the outcome happens 66% of the time, and a score of 70 is not "safer" than one of 60 in any way we have validated. We publish no tiers, no bet/no-bet decision, no fair odds and no numeric expected value, because each of those would assert a calibration standard we do not have.

Data Pipeline: From Source to Prediction

BetGlitch sources its probability and fixture data from a single specialist provider, SportMonks. Our pipeline operates in several stages:

Data ingestion: Fixture data, team information and recorded odds are pulled continuously for every covered competition.
Probability ingestion: Provider probabilities are collected for each fixture and market type, with every outcome retained rather than only the favoured side.
Filtering and ranking: The filters above narrow the candidates, and the survivors are ranked to produce a signal score.
Publication: Some selections — not all — are frozen before kickoff with their signal score, recorded price, bookmaker, provenance and immutable timestamps. A published claim is never edited or withdrawn.
Settlement: After full-time a published claim is graded against the fields frozen at publication. Results are inserted, never overwritten.

Coverage

BetGlitch indexes European competitions spanning the top divisions of major footballing nations, selected second tiers, domestic cups and the UEFA club tournaments — the Premier League, La Liga, Bundesliga, Serie A, Ligue 1, Eredivisie, Liga Portugal, Süper Lig and more. Not every indexed competition produces signals: the UEFA tournaments are searchable but are not currently swept, so a fixture can be visible without ever being ranked. The current figures are shown on the about page, which reads them from the same source the product does.

What Makes BetGlitch Different

Several qualities set BetGlitch apart from traditional tipsters and other prediction platforms:

Ranking, not tipping: We rank provider probabilities against recorded prices and only surface a candidate when it clears our filters.
Immutability: A published claim freezes its selection, signal score, price, bookmaker and provenance. It cannot be edited, re-priced, re-graded or withdrawn.
Consistency: The same rules run on every fixture. There are no favourite teams and no retrospective reinterpretation of a result.
Breadth: Sweeping every covered competition across four markets at once is beyond the capacity of an individual tipster.
Published losses: Losing claims stay up at the same prominence as winning ones. Odds are never rewritten.

What Actually Counts

It would be easy to claim that we publish everything we produce. It would also be false, and the difference matters more than the slogan. BetGlitch generates far more live signals than it publishes; only the ones we freeze become claims. So the honest version is narrower:

Live signals can change until kickoff, and count towards nothing.
Every published claim is immutable and stays permanently visible, win or lose.
Only settled published claims count in the verified record.
Pending claims count towards nothing until the match finishes.
Void and cancelled claims stay visible but are excluded from the relevant totals.

This makes our record smaller than the number of signals we produce, and that is deliberate. Settlement is graded against the fields frozen at publication, results are inserted rather than overwritten, and nothing is deleted. Visit the track record page to check any of it yourself.

BetGlitch is a free public beta, is not a bookmaker, and does not accept bets. Nothing here is advice, no output guarantees profit, and you can lose all the money you wager.