Random teams

Teams are drawn in your browser, by the scripts listed below. The list never leaves this page.

Enforced by your browser

This page's Content Security Policy allows a single destination, the site's audience measurement (connect-src https://stats.r-no.fr, see below), and forbids any form submission (form-action 'none'), so even a bug could not send what you type or draw anywhere. You can read the policy at the top of the page source.

Audience measurement

Like every page of the site, this one counts its visit with Umami, installed on R-no’s own server (stats.r-no.fr). It receives the address of the page, without parameters or anchor, the site you came from (its domain only), your browser, system, language and screen size, and the country, region and city deduced from your IP address, which is not kept. It never receives what you type, nor the result of the tool. No cookie, nothing stored in your browser; if your browser asks not to be tracked (Do Not Track), nothing is measured. The data is erased after 25 months.

• /commun/mesure.js: the filter applied before each sending, written for R-no Toolbox. 1.4 kB
  SHA-256 225eec74fd12623d82fc05ee7d222be629ce454d081c129e8d1e249ed37ee716
• https://stats.r-no.fr/m.js: Umami’s tracker, open source (MIT), served by the measurement server. Its fingerprint is not shown here: it changes with each version of Umami, and you cannot check from your browser that it is the published code. For that part, you have to trust R-no.

What is stored

No cookie, and nothing in the address. The list and the teams are never stored: they disappear when you reload or close the page. Your language is given by the address itself (/en/ or /fr/): nothing is stored to remember it. Like any website, the server records the download of the page and its files in its access logs (IP address, date and time, address requested, browser name). It never receives what you type or draw.

Code that runs here

• app.js: the team split, written for R-no Toolbox. Not minified, no third-party library. 10.1 kB
  SHA-256 db010b46e3a83224b756adc2364b0c4ce4d7b88a12540a506839b4afc4552217
• textes.js: the French messages of the tool, shown by its script (the rest of the French page is written into its HTML when the site is built). It only defines texts, with no logic. 5.9 kB
  SHA-256 91ed8abd0196a2ebe0b164b1259e9439877484802eda32aa7a7c326bcfb95936
• /commun/langue.js: the language helper, the same on every page of the site. It reads the language of the page and gives the tool its messages in that language. It stores nothing. 1.6 kB
  SHA-256 80f76f69773ef628d36d09bf8387cfedebda4529210d7120649b6f5603310631

Where the randomness comes from

Your browser's cryptographic random generator, crypto.getRandomValues (Web Crypto API). Draws that would favour some values are rejected rather than folded with a modulo, so every outcome is equally likely. The list is shuffled (Fisher-Yates), then cut into teams of the planned sizes. With constraints, the whole shuffle is redone until they are all respected: every valid split keeps exactly the same chance, which patching the teams by hand would not guarantee.

Check it yourself

Download a script and compare its fingerprint with the one above:
curl -s https://r-no.fr/random/teams/app.js | shasum -a 256

What a result proves

Nothing, to anyone else. A screenshot or a copied result does not show how many times the draw was repeated before this one, nor whether the page was modified. It is fine for deciding among yourselves; for a draw that others must be able to check (a contest, for instance), a verifiable draw tool is planned.

Limits

At most 1,000 people, and each name must be unique. Teams are balanced by headcount only: the tool knows nothing about levels or ages. With many constraints, it gives up after 10,000 tries and says so, even if a valid split might exist.