Self-calibrating weather trading bot
What happened
Movez published a full Polymarket weather-trading bot that uses ECMWF/HRRR forecasts, Kelly sizing, EV calculations and auto-calibration—with code on GitHub and strong engagement. It’s a complete end-to-end example of combining external data, sizing, and live calibration that’s adaptable to other event-driven algorithmic projects. (x.com)
Why it matters
Movez’s X thread includes concrete P&L claims for a “vibe-coded” weather bot—an example entry in the thread cites a $64K total and a London trade that grew $197 → $7,342. (unrollnow.com) The GitHub code linked in the ecosystem implements the same stack Movez describes: repos labeled “weatherbot” fetch ECMWF and HRRR (via Open‑Meteo), ingest METAR observations, compute Expected Value per Polymarket bucket, and size trades with a fractional Kelly parameter. Key implementation artifacts are visible in those repos: a bot_v2.py main loop plus config.json and calibration/storage files, hourly scan cadence, KELLY_FRACTION default ~0.25, and calibration.json used to learn per-city forecast accuracy over time. The publication sparked immediate reuse and forks—at least one community repo explicitly credits Movez as the inspiration in its README, and multiple tutorial/video posts surfaced within days that reproduce the ECMWF/HRRR+Kelly approach. Parallel projects and commercial sites mirror the same architecture (multi‑model ensemble, Bayesian/EV filter, quarter‑Kelly sizing, real‑time CLOB streaming) and list dozens of tracked cities and automation features, confirming the pattern of rapid replication after Movez’s post.
Key numbers
- (x.com) Movez’s X thread includes concrete P&L claims for a “vibe-coded” weather bot—an example entry in the thread cites a $64K total and a London trade that grew $197 → $7,342.
- Key implementation artifacts are visible in those repos: a bot_v2.py main loop plus config.json and calibration/storage files, hourly scan cadence, KELLY_FRACTION default ~0.25, and calibration.json used to learn per-city forecast accuracy over time.
Sources
Quick answers
What happened in Self-calibrating weather trading bot?
Movez published a full Polymarket weather-trading bot that uses ECMWF/HRRR forecasts, Kelly sizing, EV calculations and auto-calibration—with code on GitHub and strong engagement. It’s a complete end-to-end example of combining external data, sizing, and live calibration that’s adaptable to other event-driven algorithmic projects. (x.com)
Why does Self-calibrating weather trading bot matter?
Movez’s X thread includes concrete P&L claims for a “vibe-coded” weather bot—an example entry in the thread cites a $64K total and a London trade that grew $197 → $7,342. (unrollnow.com) The GitHub code linked in the ecosystem implements the same stack Movez describes: repos labeled “weatherbot” fetch ECMWF and HRRR (via Open‑Meteo), ingest METAR observations, compute Expected Value per Polymarket bucket, and size trades with a fractional Kelly parameter. Key implementation artifacts are visible in those repos: a bot_v2.py main loop plus config.json and calibration/storage files, hourly scan cadence, KELLY_FRACTION default ~0.25, and calibration.json used to learn per-city forecast accuracy over time. The publication sparked immediate reuse and forks—at least one community repo explicitly credits Movez as the inspiration in its README, and multiple tutorial/video posts surfaced within days that reproduce the ECMWF/HRRR+Kelly approach. Parallel projects and commercial sites mirror the same architecture (multi‑model ensemble, Bayesian/EV filter, quarter‑Kelly sizing, real‑time CLOB streaming) and list dozens of tracked cities and automation features, confirming the pattern of rapid replication after Movez’s post.
