2026-W28 — frozen weekly edition
Merit-only venue scoring within (chain × trade-size) cohorts, rolled into one weekly artifact. Frozen at publish: this page's numbers and hashes never change after release.
This edition was computed from an offline fixture (no live observations this week). It validates the pipeline end-to-end but does not measure live market execution. Treat every figure below as pipeline output, not a market claim; the observation source is part of the frozen artifact.
Headline RQI — withheld
No cohort reached the per-cohort sample floor of 30 observations this week, so the headline number is withheld rather than published on thin evidence. The coverage that was observed is reported below — counted, not hidden.
Where the observations came from
source_manifest null (computed offline — no active source manifest)
generated_at 2026-07-06T03:16:17.647669Z
Per-cohort states — ranked and withheld alike
| Chain | Size bucket | Obs | Claim state | Cohort RQI |
|---|---|---|---|---|
| arbitrum | small | 2 | insufficient_sample | withheld |
| ethereum | large | 3 | insufficient_sample | withheld |
| ethereum | small | 5 | insufficient_sample | withheld |
What was excluded, counted
insufficient_sample_observations 10
degraded_source_periods 2026-W28: 2 stale
Version + hashes
methodology_hash sha256:7524b117d588fadd01a0054ff49d7f9b411b92cb2134dbf8e4341738fa4e2aaf
sample_floor 30
Changing any weight, ceiling, or floor flips the methodology hash — and with it the artifact hash — so a cited edition can never be silently re-scored. Full approach on /methodology.
Standing citation block
Read before quoting
- The Route Quality Index is merit-only: it is a pure function of observed execution evidence with no sponsorship, partner-tier, or boost input. Concentration is a PENALTY, so a venue cannot buy rank by routing more volume through itself.
- Cohorts with fewer than the sample floor of observations are flagged 'insufficient_sample' and are EXCLUDED from the headline index — they are counted, not hidden, and never ranked into a public number.
- Realized-outcome coverage is sparse during launch; most cohorts will read 'insufficient_sample' until they pass the per-cohort sample floor, so the headline RQI may be null in early weeks.
- Scores are always computed WITHIN a cohort = (chain, route_size_bucket). A venue that is strong for small stable swaps on one chain is a different claim from the same venue on large trades on another chain; we never blend them.
- Stale observations (older than the freshness ceiling) are reported as degraded source periods and dampen the freshness component; they do not silently inflate a quality score.