About KillFeed News

What we do

Every day we pull the day's professional matches, rank the tournaments by prestige so the page shows the fixtures that actually matter, and attach context to each one: recent form, head-to-head history, an ELO-based win probability, and a short AI-written preview. News briefs are generated and fact-anchored against the same match data.

How the ELO system works

ELO is a rating system originally built for chess. Every team starts on a baseline rating of 1000. When two teams meet, their ratings produce an expected result:

expected(A) = 1 / (1 + 10 ^ ((ratingB - ratingA) / 400))

After the match, each team's rating moves by K × (actual − expected), where K is 32. Beating a stronger side moves your rating a long way; beating a much weaker side barely moves it at all. Ratings are rebuilt chronologically from stored professional results, and only completed matches count.

The win probability shown on each card is simply the two teams' expected scores normalised to 100%. There is no draw option — esports matches resolve to a winner. Form, ELO and head-to-head shown for a given day reflect the state going into that day, so a result that lands this afternoon changes the ratings tomorrow, not retroactively.

We publish our own hit rate openly — the accuracy banner on the homepage shows how often the higher-rated side actually won over the last 1, 7 and 30 days. Predictions are for entertainment and analysis only.

Where the data comes from

Match fixtures, scores and tournament metadata come from PandaScore. Story imagery is royalty-free material from Wikipedia and Wikimedia Commons. Written previews and briefs are AI-generated from that underlying data and reviewed for tone, not rewritten by hand.

Who built it

KillFeed News is built and run by Karl Tomkins through Rubber Duck Labs Ltd, an independent UK software studio. Feedback, corrections and partnership enquiries are welcome at hello@rubberducklabs.uk.

Connect on LinkedIn, or support the running costs on Ko-fi.