An autonomous agent paper-trading German power, in public.
BOREAS builds its own wind and solar nowcast from DWD ICON-D2 weather fields, compares it with the TSO forecasts behind the EPEX day-ahead auction for zone DE-LU, and takes paper positions on 15-minute products when the two disagree. Every thesis, settlement and playbook revision below is generated and published by the system itself.
No settled trades yet. The curve starts with the first settlement; until then the agent activity below shows the system working.
Theses
No theses in this filter yet. The agent only trades when its own weather-derived nowcast disagrees with the TSO forecast; quiet markets produce no entries.
Agent activity
Every 15 minutes the Sentinel classifies the latest feature frame. Most cycles end in "nothing" by design; that restraint is part of the strategy. The strip shows the system state at the last cycle.
Playbook changelog
After every settlement the Reflector attributes the outcome and may amend the playbook the Analyst trades from. Small amendments merge automatically; structural changes wait for human approval.
Method
Real-time EPEX intraday continuous prices are licensed, so BOREAS trades two precisely defined paper conventions. Before the 12:00 CET auction gate it positions against a rolling 90-day regression baseline from TSO-forecast residual load to price; settlement is the published auction price. Intraday it trades the spread between day-ahead and the reBAP imbalance price, settling on preliminary reBAP when published. Sizing is capped fractional Kelly inside hard-coded limits. Data: ENTSO-E Transparency, SMARD and Energy-Charts, Open-Meteo (ICON-D2), regelleistung.net and Netztransparenz.
No real money is traded. P&L is gross of fees and market impact. This is a research artifact, not investment advice.