The cycle
Monthly, anchored to UTC. The nightly capture runs at 23:30 UTC and the archive and integrity jobs follow it, so every date below assumes the prior day's snapshot is settled and verified.
| Day (UTC) | Event | Published? |
|---|---|---|
| T−7 — 25th | Review cut. Eligibility and weights computed from the trailing 30 days ending on the 24th. | no |
| T−5 — 27th | Provisional constituent list published, with per-row evidence and the concentration block. Appeals window opens. | yes |
| T−1 — 31st (or month end) | Appeals window closes, 17:00 UTC. | — |
| T — 1st | Rebalance effective, 00:00 UTC. Final list and change log published. | yes |
The provisional list is published five days before the rebalance takes effect. That gap exists so an issuer can see and contest its own row before it becomes the basis of anything settled. An index that publishes constituents and effects them simultaneously has no appeals process, whatever its documentation says.
The window is the invitation to contest. It is not a promise of a decision before T. An appeal filed in it is handled under the published service levels in §03.3 — acknowledged within 5 business days, resolved within 30 calendar days of that acknowledgement — and those are the only timing commitments Paddock makes.
Where an appeal filed against a provisional list is upheld after the rebalance has taken effect, the change is given effect off-cycle under §02.3 rather than waiting a month. Filing early still helps; it just is not a guarantee, and this page will not print one it cannot keep.
What is published at each step
Provisional (T−5) and final (T) both carry:
- every constituent with its matched domain, matched wallet(s), 30-day settled volume and transaction count,
probe_status, and evidence URL; - the full concentration block (§01.4);
- the change log against the previous period: entries, exits, and weight moves beyond a stated threshold, each with a reason;
- the methodology version in force.
Exits name their reason. “Fell below the settled-volume floor”, “wallet added to the circular denylist”, “attribution withdrawn on appeal” are different events and are not collapsed into “removed”.
Off-cycle changes
Three, and only three, justify changing the index between rebalances.
- A constituent's wallet is added to the circular/flagged or facilitator denylist. Its settlement is no longer attributable demand.
- An upheld appeal (see §03) that materially changes a constituent's eligibility or weight — including one filed against the provisional list and decided after T.
- A discovered error in the published basis — a figure computed on the wrong denominator, a row whose evidence does not resolve.
Anything else waits for the cycle. Price moves, volume spikes, news, and requests from issuers are not off-cycle events.
An off-cycle change is published with the same disclosure as a rebalance and is recorded in the corrections log when it moves a published level.
The discontinuity rule — the one that matters most
A methodology change is not a rebalance. This is the rule most likely to be violated under time pressure, and it is the one that would do the most damage, because the damage is invisible in the number itself.
When eligibility, weighting, exclusions, or any input basis changes:
- Record it as a dated discontinuity before the first level is published on the new basis, naming the date, the change, and the cause.
- Bump the methodology version (§01.6).
- Annotate the series at that date. Any chart crossing the boundary shows the annotation. Never smooth across it, never splice, never restate prior levels on the new basis.
- Never present a cross-boundary comparison as a market movement. A change in what is measured is not a change in what is being measured.
The standing example is the 2026-06-10/06-12 event, where the basis moved by roughly 10× and any trend line crossing it shows a measurement change as market growth.
A second is live: an unexplained ~7× step in daily settled transactions on 2026-08-17 that cannot currently be attributed to market movement or to an upstream re-basing, because the only parity check available compares our series against the same upstream. Where cause is undetermined, the annotation says undetermined — it does not guess.
Data-quality holds
The rebalance does not proceed on a basis known to be incomplete.
Hold and publish the hold, rather than rebalancing quietly, when any of:
- the snapshot series has a gap in the trailing window;
- the recipient capture was truncated in a way that moves eligibility (an incomplete recipient list is the steady state today — the hold is for a change in truncation that would move constituents, not for its steady state);
- the circular classification is stale beyond its weekly cycle, so exclusions rest on an old detection;
- a source's disclosure marker shows a failed or carried fetch in the window.
A held rebalance keeps the prior period's constituents in force and says so. Carrying the previous list with a stated reason is honest; rebalancing on a basis you know is broken is not.
A hold is the only permitted reason to move off the dates in §02.1. A licensee's exposure is never one — see §04.3.