EuroLeague Analytics

Add EuroLeague to the AI you already use.

Connect once, then ask in your own words — every lineup, every shot, every second on court. No app, no spreadsheet.

Connect See how it works
Claude EuroLeague Analytics connected

EuroLeague Analyticsreading the play-by-play

Their most efficient five this season:

  • 13Biberovic
  • 92Birch
  • 20Hall
  • 8Horton-Tucker
  • 4Melli

+24.5 net rating over 214 possessions together

EuroLeague Analyticscounting possessions

Paris Basketball, by some distance:

78.0 possessions per game counted from the event stream, not estimated from a box score

EuroLeague Analyticsfiltering by margin and clock

Last two minutes, within three points:

646 possessions and you pick the thresholds — clutch is a filter here, not a fixed definition

A recorded answer from the warehouse. Works the same in any assistant that speaks MCP.

Ask where the shots came from.

Any game, any player, any stretch of the floor — the coordinate the league recorded, not a zone and not a guess from a box score.

Olympiacos78 Real Madrid79 Final Four championship game, 21 May 2023

0 shots placed

Every field goal attempt, placed at the coordinate the league recorded. Free throws have none, so they are absent.

Placing every attempt in the game

Ask which five actually work together.

Four Fenerbahce players stay. Swap Birch for Baldwin, and the team turns over.

Every answer shows the possessions behind it, so a small sample cannot hide.

The club supplies each position; the spots on this floor are illustrative.

— net rating
— possessions together

Now ask it yourself.

The same window, and this time you drive it. Each answer was produced by the same query your own assistant would run — no number here was typed by hand.

You Your assistant EuroLeague Analytics connected

Ask me something. Pick one of the questions below and I will go and read it.

Connect it once.

There is one address, and it is the whole of the setup. Any assistant that speaks MCP takes it — there is nothing to install, and nothing to configure afterwards.

https://euroleague-analytics-mcp.fly.dev/mcp

Verified — connected and answering

  1. Open Settings, then Connectors.
  2. Choose Add custom connector and paste the address above.
  3. Sign in with Google when Claude opens the sign-in window.
  4. Ask it something. Eleven tools appear under the connector's name.

In Claude Code instead:

claude mcp add --transport http euroleague https://euroleague-analytics-mcp.fly.dev/mcp

Free to use. Read-only: eleven tools that answer questions and none that change anything. The full compatibility matrix

Ask it something hard.

The hard part is the reading: asking the next question, rejecting a weak answer, and naming what was left out.

Three real sessions against the running server. Every figure and caveat below came back from it.

It asks the next question for you.

One threshold is an opinion. A useful answer moves the threshold and shows you what happens, then tells you where the number stops being worth quoting.

C Claude EuroLeague Analytics connected

Fenerbahce look good late in games. Check it — how do they score in the last two minutes with the margin inside three? Does it hold if I loosen the definition? And tell me whether the number is trustworthy.

EuroLeague Analytics comparing three clutch thresholds

Over the whole season Fenerbahce scored 112.8 per 100 possessions. Narrow to the last five minutes within five points and it rises to 116.4, over 159 possessions. Narrow again — last two minutes, within three — and it jumps to 144.2.

Be careful with that last one. It rests on 43 possessions, about half a game's worth, and five of them straddled a substitution. The direction is real; the size of it is not established. These are counted possessions, not a box-score estimate, so do not compare the figure with a pace number from anywhere else.

144.2 per 100 in the clutch on 43 possessions — which is why it says so

It takes its own answer back.

Every lineup response carries the possessions behind it and a warning that small samples are noise. An assistant that reads its own tool output will act on that — and the second answer is the real one.

C Claude EuroLeague Analytics connected

Which is Panathinaikos's best five this season?

EuroLeague Analytics checking the sample behind each lineup

My first answer was not an answer. At a 25-possession threshold the best Panathinaikos unit is Hernangomez–Holmes–Nunn–Osman–Shorts at +76.6 net, over 54 possessions. That is about a quarter of a game. A number like that is noise wearing a suit.

Raise the bar to 100 possessions and exactly one unit survives: Grant–Hayes-Davis–Lessort–Nunn–Osman, +7.55 over 133 possessions. That is the only five-man number here I would repeat out loud.

+7.6 net rating over 133 possessions — the one unit that qualifies

It tells you what it left out.

Games that fail a validation invariant are held back from every answer, and the response says how many and why. Nothing is deleted and you can ask for them — but you are told before you quote a number, not after.

C Claude EuroLeague Analytics connected

How many games is all of this built on — and is there anything you are not telling me?

EuroLeague Analytics checking coverage and exclusions

732 games are loaded across two seasons and 53 of them are held back: 22 in 2024-25 and 31 in 2025-26. Take this season — 371 of 402 are used. Seventeen of the 31 failed the possession gate, twelve had a statistical event attributed to a player believed to be off court, three disagreed with the official box score on minutes, and one had a substitution state I could not resolve.

I did not have to go looking for that. Every answer carries it, so I cannot quote you a number without being told what is missing from it. The games are not deleted and you can ask me to include them — but then I would say so.

679 of 732 games used and the 53 left out are named, not hidden

What is behind it.

732
games loaded, across 2024-25 and 2025-26, with 2026-27 filling nightly
53
games held back because they failed a validation check, and left out of every answer unless you ask for them
11
tools, all read-only, all documented in the repository

Counting statistics are the official box score. Possessions, lineups, on/off and every per-100 rate are this project's own reconstruction from the play-by-play, and every answer says which it gave you. Read the source