Q9, run 2026-11-04-1738-sonnet

← back to the run

As of 1 October 2026, how many distinct issuers hold an issuer rating?

Tests: Everything at once: identity, withdrawal, instrument versus issuer, ceiling versus rating.

Expected: count = 278

❌ R1 csv run 1 graded by code

291. This counts distinct entities (duplicates merged) whose latest FC or LC issuer rating was not withdrawn by 2026-10-01. It treats the 13 early-sample entities plus CV as rated. The assumptions are in the notes below.

JSON: {"count": 291}

Why: count: expected 278, got 291

❌ R1 csv run 2 graded by code

290. This counts distinct entities, with duplicates merged by LEI, whose latest issuer-rating action on or before 2026-10-01 is not a withdrawal. Three cases are judgement calls: E1115, E1158 and E1303 have actions dated after their withdrawal, and I went by dates. E1106 (only its local-currency rating was withdrawn) and E1279 (only its foreign-currency rating was withdrawn) were counted because one rating remains.

JSON: {"count": 290}

Why: count: expected 278, got 290

❌ R1 csv run 3 graded by code

291 by my count, so treat it as approximate. This is distinct canonical entities with a non-withdrawn issuer rating (LTIDR, either currency basis) on 2026-10-01. CV, QA-source records and E08 (RD) are included. VD is not, because it has only a country ceiling.

JSON: {"count": 291}

Why: count: expected 278, got 291

❌ R1 csv run 4 graded by code

290. Duplicate records are merged by LEI, and each entity's latest action on or before 2026-10-01 is not a withdrawal. This includes E1115, E1158 and E1303, whose WD is dated before a later action. It excludes VD, which has only a country-ceiling record, and counterparty-only entities.

JSON: {"count": 290}

Why: count: expected 278, got 290

❌ R1 csv run 5 graded by code

About 289, an estimate I did not verify record by record. It counts distinct issuers (deduplicated by LEI) whose latest LTIDR is not a withdrawal. CO ratings and the country-ceiling record are not issuer ratings.

JSON: {"count": 289}

Why: count: expected 278, got 289

✅ R4 csv+markdown run 1 graded by code

278. By latest action (counting an issuer if any LTIDR basis is still rated), 313 non-QA issuers were ever rated and 35 had withdrawn. Three entities have an action dated after their WD: E1115 (A1250), E1158 (A1338) and E1303 (A1649). I treated those as rated because the latest action wins.

JSON: {"count": 278}

✅ R4 csv+markdown run 2 graded by code

278. Of the 300 distinct entities in E1001–E1324, 34 had their FC rating withdrawn and 266 held one. Another 12 of the earlier entities held a rating, and sovereign VD has only a ceiling.

JSON: {"count": 278}

✅ R4 csv+markdown run 3 graded by code

278. This counts distinct LEIs whose latest long-term issuer rating action on 2026-10-01 was not a withdrawal. Several entities had a withdrawal that was followed by a later action (E1158, E1115, E1303), and I treated those as rated. E1279 and E1106 kept a local-currency or foreign-currency rating after the other was withdrawn. This count was done by hand across a very large table, so treat it as less certain.

JSON: {"count": 278}

❌ R4 csv+markdown run 4 graded by code

277. This counts distinct entities (LEI-merged, QA excluded) whose latest LTIDR, or an LTIDR on at least one currency basis, is not WD on 2026-10-01. Entities with a withdrawal as their latest action are excluded. E1158 and E1115 have a rating action dated after an earlier WD, and I counted them as rated by the latest-action rule.

JSON: {"count": 277}

Why: count: expected 278, got 277

✅ R4 csv+markdown run 5 graded by code

278. After mapping duplicate-LEI records to one entity and excluding QA records, 266 of the E10xx–E13xx entities still held an LTIDR on 2026-10-01. Adding the 12 older entities (E01, E02, E05, E06, E08, E09, E10, E12, E13, E14, CV, E15) gives 278. This follows the latest-action-by-date rule, which leaves E1115, E1158 and E1303 rated because an action after their WD restores a rating. E1279 and E1106 stay counted as well, since each keeps one currency basis after the other was withdrawn.

JSON: {"count": 278}

✅ R10 executed run 1 graded by code

278. This counts entities, including the sovereign CV, with a non-withdrawn issuer (LTIDR) rating still open on 1 Oct 2026. The count is hand-tallied from a long list, so it carries some risk of error.

JSON: {"count": 278}

✅ R10 executed run 2 graded by code

278. This counts every entity with an open, non-withdrawn LTIDR, including the sovereign CV and the RD-rated E08. It is 277 if you exclude CV.

JSON: {"count": 278}

❌ R10 executed run 3 graded by code

277. This counts entities with a non-withdrawn issuer rating in force on 2026-10-01. It includes CV and E1279, whose only current rating is local-currency. Without the sovereign CV the count is 276.

JSON: {"count": 277}

Why: count: expected 278, got 277

✅ R10 executed run 4 graded by code

278. This counts entities with an open, non-withdrawn LTIDR on 2026-10-01, and includes the sovereign CV. Without CV the count is 277.

JSON: {"count": 278}

✅ R10 executed run 5 graded by code

278. This counts entities (including the sovereign CV) with a non-withdrawn issuer rating in force on 2026-10-01, with each entity counted once.

JSON: {"count": 278}