Top-five recipient concentration

What it measures

The share of clean Base micro-band transactions received by the top five to_wallet values in the current last full week. The source validation defines the interval as 2026-06-01T00:00:00 <= timestamp < 2026-06-08T00:00:00, the micro-band as chain='base' AND amount_usdc BETWEEN 0.0005 AND 5, and clean as wash_flag IS NULL.

How it is produced

The canonical query filters the full-week population once, ranks recipients by transaction count, sums the first five rows, and divides by the same filtered denominator. The decomposition query exposes each top recipient's individual share.

Canonical query

WITH clean AS (
  SELECT to_wallet
  FROM payments
  WHERE chain='base'
    AND amount_usdc BETWEEN 0.0005 AND 5
    AND wash_flag IS NULL
    AND timestamp >= '2026-06-01T00:00:00'
    AND timestamp < '2026-06-08T00:00:00'
),
top5 AS (
  SELECT to_wallet, COUNT(*) AS n
  FROM clean
  GROUP BY to_wallet
  ORDER BY n DESC
  LIMIT 5
),
totals AS (SELECT COUNT(*) AS clean_tx FROM clean)
SELECT
  (SELECT clean_tx FROM totals) AS clean_tx,
  SUM(n) AS top5_tx,
  ROUND(100.0 * SUM(n) / (SELECT clean_tx FROM totals), 6) AS top5_share_pct
FROM top5;

Wallet decomposition

WITH clean AS (
  SELECT to_wallet
  FROM payments
  WHERE chain='base'
    AND amount_usdc BETWEEN 0.0005 AND 5
    AND wash_flag IS NULL
    AND timestamp >= '2026-06-01T00:00:00'
    AND timestamp < '2026-06-08T00:00:00'
),
totals AS (SELECT COUNT(*) AS clean_tx FROM clean)
SELECT
  to_wallet,
  COUNT(*) AS n,
  ROUND(100.0 * COUNT(*) / (SELECT clean_tx FROM totals), 6) AS pct_of_clean
FROM clean
GROUP BY to_wallet
ORDER BY n DESC
LIMIT 10;

Limits and assumptions

• The metric inherits the static-rule risk in wash_flag. If the classifier misses a pattern, both the clean denominator and the top-five share can be wrong.
• The week boundary is UTC Monday through Sunday, based on the source note's current date of 2026-06-08.
• The deterministic replacement recomputes clean eligibility from a versioned classifier at query time or materializes a wash_flag_version, and always publishes the week interval with the value.

How to refute this

1. Re-run the canonical SQL and get a different clean count or top-5 count.
2. Recompute wash flags and show some top-5 rows should not be clean.
3. Use a different full-week boundary and show the concentration materially changes.

Last validated

2026-06-08