Esports Analytics
Live CS2 and Dota 2 win probabilities, round by round, checked against the market before anything is traded.
About
Overview
Esports Analytics follows live CS2 and Dota 2 matches and works out, round by round, how likely each team is to win the map and the series. It sets that probability against the live market price, records every signal it finds in a journal, and acts only when every risk guard agrees.
Live match state arrives two ways: from a round-by-round match data feed, and from OCR that reads the score off the broadcast itself, so a slow feed does not leave the engine blind. Each game has its own model, trained on historical matches and calibrated against past results.
Every signal passes confidence, price and edge checks before an order is placed. A rolling kill switch pauses trading after a losing run and sends an alert, a review gate stops each strategy at a set number of fills until an operator signs off, and a new game runs in shadow mode, recording what it would have done, before it trades.
Built as Python services under systemd on AWS, with XGBoost models calibrated in scikit-learn, SQLite for state, Discord for alerts and a Next.js operations dashboard that works on desktop and phone.
Media
Screens & stills
Highlights
What makes it work
Win probability, round by round
Separate CS2 and Dota 2 models price each map and series while it is played, trained on historical matches and calibrated against past results.
Two live sources
A match data feed and broadcast OCR both report the score. The dashboard shows how far behind each one is, so stale data is visible before it matters.
Guards before every order
Confidence, price and edge checks on every signal. A rolling kill switch pauses after a losing run, and a review gate stops at a set number of fills.
One screen to run it
Balance, signal journal, live matches, gates and feed health on one dashboard that refreshes every 30 seconds, on desktop or phone.
Have a project in mind?
Tell us what you're planning: a website, a game or a video. We'll reply with how we'd approach it, a rough timeline and a cost estimate.