How to use the site

The home page is a match list. Four tabs sit at the top:

• Live — Matches being played right now, with the score and the minute. It refreshes itself once a minute. When a match ends it leaves this tab and moves to Results.
• Upcoming — The default view. Only matches not yet played, nearest first. Matches in progress and finished matches do not appear here — they have their own tabs. One exception: a match whose kick-off has passed but whose result the provider hasn't sent yet waits here, because it would otherwise fall into no tab at all.
• Results — Matches already played, newest first. This is where last night's score is. Past analyses are never deleted; they simply don't appear in the upcoming list.
• Pick of the day — The at most three matches the model is most confident about that day. The rule is fixed and published: at least 60% on an outright result, or, if nothing clears that, at least 80% on a double chance. If no match clears the bar the tab is not left empty — the day's most confident call is shown and labelled as below the threshold. This tab is open to Pro members; the past record of these picks sits underneath it.
• League selector — The round buttons under the tabs. The number beside each one is how many matches of data we hold for that league. Picking a league keeps whichever tab you were on.

Opening a match shows the probabilities the model computed for it, and what those probabilities rest on.

• Probability bars — Home win / draw / away win. The three always add up to 100%. The bar length visualises the number; the measure is the number itself, not the bar.
• How many matches it was computed from — The bottom of the page always states it: how many matches played BEFORE kick-off in that league the ratings came from, and how many matches the less-played of the two teams has. No data from after the match is ever used — that rule is what stops the numbers from looking falsely good.
• Some matches get no probability at all — If either team has played fewer than the threshold in that league, no percentage is shown and the reason is written out. For newly promoted teams this is normal for the first few weeks: a percentage built on thin data is a guess dressed up as certainty.
• The referee section — If a referee has been appointed, their past card averages are shown. For MOST unplayed matches no referee is known yet and the page says so plainly — nothing is broken; the provider simply publishes appointments close to kick-off.

The threshold: a team needs at least 10 matches in that league. Below it, no probability is published.

The model claims no certainty; it states a probability. If an outcome was given 70%, then it NOT happening in three matches out of ten is the expected behaviour. A single match neither confirms nor refutes the model.

The right question is: across the matches where it said 70%, is it right about 70% of the time? That is answered over thousands of matches, not one — and the track record page is exactly that answer.

This is why no number here is advice. A high probability does not mean “certain”, only “more likely”.

The page that measures how accurate the model has been. Three things are worth knowing before you read it:

• Brier score — lower is better — The squared gap between the probability given and what actually happened. A confident, correct forecast approaches zero; a confident, wrong one grows. Saying 50% when unsure is penalised less than being sure and wrong — which is exactly what the measure wants.
• Why a baseline always sits next to it — A Brier score alone is meaningless: is 0.61 good or bad, compared to what? So the score of a simple method (the historical base rate) on the same matches is shown beside it. The model adds something only if it scores lower than that.
• Why incorrect predictions aren't hidden — The page is unfiltered and comes straight from the database. Filtering out the incorrect ones would flatter the number and destroy the measurement — a platform that tells its success story with examples it picked itself isn't measuring, it's marketing.

Matris is not in an app store, but it can be added to your phone's home screen as a shortcut. Once added it opens full screen, without the browser bar — it behaves like an installed app. Nothing to download, no storage used, no account needed.

• iPhone / iPad — Open the site in Safari or Chrome, tap the share button (the arrow pointing up) in the address bar, then choose “Add to Home Screen”.
• Android — In Chrome, tap the three dots at the top right and choose “Add to Home screen” or “Install app”. On some phones the prompt appears on its own.
• Desktop — In Chrome and Edge, click the install icon at the right of the address bar. The site opens in its own window, without tabs.