What CompactStats is

CompactStats is a football research tool for the goals markets — over 2.5, over 3.5, both teams to score — and match results. It covers 150+ leagues and cups across five seasons, refreshes several times a day, and shows a calibrated probability for every fixture rather than a tip.

Calibrated means what it sounds like: when a rating says 60%, that outcome should happen about 60% of the time. We test that claim by replaying tens of thousands of past matches, making each prediction using only information that existed before kick-off, and comparing what we said with what happened. Every weighting in the model earned its place that way, and the studies are published in full — including the ideas that failed.

How to read a fixture

Each row shows our pick, the probability behind it, and the edge — our probability minus the bookmaker's, once their margin is stripped out. The edge column is the one that matters for value: a 72% call priced at odds implying 75% is a confident opinion and a poor bet, while a 55% call priced at 48% is the reverse. Confidence and value are different things, and the site keeps them in separate columns on purpose.

Open any fixture for the full picture: each side's season form at home, away and overall; how their goals split between halves; head-to-head history; league table position; a score projection; and the callouts that flag things the raw numbers hide — a side returning from a long break, a 2,000km midweek trip, or a contradiction with our own outlier boards.

Where we say nothing

Roughly a third of the fixtures we cover carry no pick at all, and that is deliberate. European qualifier rounds graded a coin flip across 1,150 historical games, so we stopped calling them. Cup ties where a side has played fewer than three league matches give the model nothing team-specific to work with. And in August 2026 we deleted our Over 3.5 under-calls despite a 75% strike rate, because always saying "under" scores about 70% anyway — the full reasoning. Empty cells are not gaps in the product; they are the places a number would have been decoration.

The record

Finished games are graded against official results, and since 30 July 2026 every rating is frozen before kick-off into a version-controlled ledger. Improvements to the model apply from that moment forward and cannot reach backwards, so what you saw before a match is what stays on the page afterwards — wins and misses alike. The daily strongest-combination selections lock the same way each morning, before play.

The research behind it

Every model choice on this site came from a study we ran on our own data, and each one is published in full — including the tests that failed and the ideas we abandoned.

• The call we deleted for winning Our Over 3.5 under-calls were landing at 75%. We removed them, because always saying "under" scores 70% — and five points of that is not skill.
• Early-season league averages lie Six matches into a season a league can look like it has changed character. We measured how much of that is real: almost none of it.
• What we learned auditing 104 leagues We tested the model league by league to find where it should stay quiet. The honest answer surprised us: almost nowhere — yet.
• Why four strong picks make a weak bet Four 70% calls together are not a 70% bet. They are a 24% bet — and that gap is where accumulator marketing lives.
• Grading your own homework If a prediction site can rewrite yesterday's ratings, its record means nothing. Here is the mechanism that stops us doing it.
• When a model should say nothing Roughly a third of the fixtures we cover get no pick at all. The rules for staying quiet, and what each one cost us to learn.
• Home/away form splits are ~94% noise A 30,582-match study: most home/away over-rate gaps are indistinguishable from coin flips — and models that trust them get worse.
• European qualifiers grade like coin flips 1,150 UEFA qualifier games: our goals calls landed 51.4% — so we stopped making them.
• Over 3.5 needs three answers, not two A ~30% event breaks binary calls: the honest model said Under on 99% of rows. Calibration on 30,582 matches fixed the display, not the maths.
• What a 70% call really means 44,728 walk-forward matches: calls rated 67%+ landed 68.5% — calibrated, streaky, and not automatically valuable.
• Why an honest model almost never picks a draw 45,556 matches: the highest draw probability our model ever produced was 46.8% — draws are real, calibrated, and almost never the most likely outcome.
• Travel, breaks, and the European hangover Long trips cost win probability but not goals; form streaks die over long breaks; and the fabled post-Europe slump barely exists.

Start with today's league and team stats, read the methodology, or browse all research. CompactStats is an independent site operated from Northern Ireland. It is a research aid, not betting advice, and it is for over-18s.