Evidence / Data kit / Polygon L1 regime

Polygon L1 regimeArchive · reintegrating 2026-05-16

Chain backtest certified MEDIUM on canonical Polygon events. The Polygon PoS Bridge collector started 2026-04-23 on Ethereum L1 RootChainProxy and needs 25 days of continuous samples before the full live scope returns. This kit mirrors the chain backtest pending bridge calibration.

Source

Dataset
bigquery-public-data.crypto_polygon.blocks
Integration window
Phi = 720 blocks, about 1 hour at 5 s per block, production-aligned.
Period
2020-06-01 to 2023-12-31.
Signals
rhythm_ratio (tau), sigma_ratio / size_ratio / tx_ratio (pi).

Files

MD
README.md
Full reproduction instructions and matching criteria.
SQL
extract_pol.sql
BigQuery query producing one row per Phi window with rho_ts, rho_s, size_avg, tx_count_avg.
MD
expected_thresholds.md
The TPR, FPR, M1 and per-window state expectations to match, with note on the elevated FPR.

Reproduction steps

  1. Open the BigQuery console on a project with the free monthly quota.
  2. Paste the content of extract_pol.sql, run the query.
  3. Export the result as pol_invariants_2020_2023_phi720.csv.
  4. Clone github.com/agentnorthstar/calibration.
  5. Place the CSV in scripts/, run python scripts/backtest_pol_phi720.py.
  6. Confront the output to expected_thresholds.md.
Polygon is kept in the archive view on the landing page until the Polygon PoS Bridge collector reaches 25 days of continuous samples (earliest 2026-05-16). Internal calibration parameters (nominal construction, divergence thresholds) are embedded in backtest_pol_phi720.py at runtime and are not exposed publicly. Reproducers confront the output (TPR, FPR, M1, per-window states), not the parameters.