A personal Counter-Strike 2 telemetry project
01Individual match history · 65 tournaments
Track map-to-map variance of HLTV Rating 3.0 across tier-1 tournaments, comparing 5, 10, 15, and 20-map moving averages against career baselines.
02Rolling momentum & streaks · Tier-1 CS2
Evaluate current player trajectory with rolling 10-map ratings, plain-language form verdicts, and recent match outcome breakdowns.
03Spatial combat telemetry · 598,000+ events
Interactive top-down radar spatial projections rendering player duel locations, directional kill vectors, and trade attribution on official competitive maps.
How it works
An automated Python pipeline downloads and verifies tournament demos across 65 tier-1 events. Uses browser TLS impersonation, request jitter, circuit-breaker error handling, and SQLite WAL mode to track match processing state without data loss.
Processes >380 GB of raw match replays into a compact 472 MB partitioned Parquet datastore (>99.8% storage reduction). Extracts and normalizes 598,000+ combat events with sub-tick precision.
Transforms in-game world coordinates into 2D radar map space. Computes directional kill vectors and executes 3-second sliding-window algorithms to identify and credit trade kills.
Next.js App Router renders tournament analytics, Canvas-based form curves, and interactive radar duel heatmaps. Pre-computed manifest indexes enable instant player lookups and shareable state.
Next.js and TypeScript on the site, Python and Polars for the data pipelines, and Snappy Parquet / Supabase for the cleaned telemetry.
Next.js · TypeScript · Python · Polars · Snappy Parquet · SQLite · Supabase · Canvas · SVG