Zcash methodology test on real targets, 2026-09-17 (second run)

Source: chains/zcash/data/methodology_test_2026-09-17.md in the repository, rendered here without changes. Source file

Full tables are in METHODOLOGY.md section 7.3. This file is the short record of what was tested, what held and what changed.

What was re-derived, and how#

Target	Primary source read	Independent of the published run?
ZIP 271 fund `t3ev37Q2uL1sfTsiJQJiWJoFzQpDhmnUwYo`	`GetTaddressTxids` 3,146,400 to 3,486,770 and `GetTaddressBalance` on `zec.rocks`, `zcash.mysideoftheweb.com`, `na.zec.rocks`; ZIP 271 text	Yes: a from-scratch gRPC, transaction and script parser (not this folder's `scripts/`), scores computed by hand from METHODOLOGY.md 4.6
ZCG fund `t3cFfPt1Bcvgez9ZbMBFWeZsskxTkPzGCow` (control)	Same calls, window 3,476,771 to 3,486,770 (not the scorer's window)	Yes
`L1` concentration input	`GetBlockRange` coinbase outputs, published window and a fresh window 3,484,771 to 3,486,770, two operators; coinbase input text of 200 sampled blocks via `GetTransaction`	Yes
ZIP 0 citations	`zips/zip-0000.rst` on `main` and the rendered `zips.z.cash/zip-0000`	Yes

Result#

Check	Outcome
Fund thresholds, pubkeys, balances	Match: both 2-of-3, keys disjoint, `t3ev37...` balance 78,183.4093 ZEC on three operators
Fund scores by hand	50 / 39 / 0 / 100 / 100, composite 32, capped 32: match
`L1` `k50` / `k25`	3 / 1 on both windows, both operators, all blocks and attributed-only blocks: match, `multisigScore` 50
`L1` lineage and synthetic key	1 independent lineage (Zakura still a declared fork), key `0xB8e62E6AeB4C400b6F3e66cAb69ACe55439035c3`: match

Rules changed by the test#

Open point	Resolution
Unspent P2SH rule	"score 0" replaced by an unresolved-P2SH floor (adminKey 5, multisig 16, composite 7), also applied to unrecognised script shapes and operator disagreement. The old code placeholder (20/20) ranked an unknown script above a known 1-of-3
Double counting of miner concentration	Scored once, in `L1` `multisigScore`; `oracleAuthorityScore` discloses it without scoring; the lightwalletd lineage input stays in `adminKeyScore` only
Exact ZIP 0 citation	Exact sentences now quoted; "must always hold a ZIP Editor seat" corrected to ZIP 0's "The current design of the ZIP Process dictates ..."
Found during the test	Script suffix validation added; concentration window made tip-relative and two-operator

None of these changes moves a published score today. Sensitivity (not scored): if the two untagged payout addresses ranked 1st and 5th belonged to one entity, L1 multisigScore would be 35 and the L1 composite 29.