Documentation
To run the project, start with the team quickstart. The research guides explain existing results and the evidence still missing.
Project guide and offline demos
Offline demonstrations, data evidence, research candidates and execution boundaries.
Architecture
One repository, workflow modules and a separate reviewed execution handoff.
Team quickstart
Install, configure and run the independent offline packages.
Data and snapshots
Daily acquisition, immutable receipts and input validation.
Study methods and results
Fixed assumptions, accounting, actual results and limitations.
Strategy guide
Signals, fixed pools, failure cases, costs and reproduction.
Strategy controls
Fixed comparisons, verification and publication status.
Robustness checks
US exclusions, recomputed crypto fees and ex-post volume checks.
Fresh data and observations
Bounded request plans, fresh warmup and per-market protocol handoff.
Synthetic process timing
Separate synthetic startup, exit and timeout evidence.
US execution evidence
What daily caches still need before next-session replay.
Manual observation registry
Freeze candidates and preserve a continuous manual account.
Equity prediction study
Read the source-candidate pool, fixed model protocol and three-period results.
Reversal, models and weekly trading
Check what learned models add and how less frequent crypto trading changes costs.
Trading frequency, prediction and attribution
Bounded holding-horizon models, exact cost tolerance, crypto PnL components and unresolved observation inputs.
How a CNN reads price charts
Recovered chart pictures, archived comparisons and what still needs an unseen-data test.