Output B — Hub Identification

Daily ranking of which coins lead vs. lag within each cluster, with the Granger p-value for each lead-lag relationship.

Cadence

Daily, at the configured snapshot time (default 00:00 UTC).

Schema

The pydantic model is memenet.schemas.hubs.HubSnapshot.

Example payload:

{
  "timestamp": "2026-05-02T00:00:00Z",
  "hubs": [
    {
      "cluster_id": "cluster_0",
      "hub_coin": "BONK",
      "hub_score": 0.87,
      "leads": [
        {"coin": "WIF", "lag_minutes": 30, "granger_pval": 0.003}
      ]
    }
  ]
}

How to consume

Snapshots are written to partitioned Parquet under outputs/hubs/ and queryable via DuckDB. P-values are Benjamini-Hochberg corrected over the full pair set evaluated on the snapshot day.