2016-17 Panini Hoops

• G-001 These three base parallels are written print_run=unnumbered / serial_numbered=no on a POSITIVE DISCRIMINATIVE TEST and not on an absence: no serial token renders on any row of sids 137572, 137573 or 137585 on either Checklist.cfm (pages 1-3) or Gallery.cfm (page 30), while those SAME two endpoints render SN149 on sibling sid 137574 and SN10 on sibling sid 137578 of the same parent. That is the strongest form CAV-042 allows. It is still ONE source: two tcdb endpoints are two endpoints of one source, and cardboardconnection printing no figure for these three is an omission, not a statement of fact (EXTRACTION_SPEC 2.1). source_confidence=single on all three. A second independent lineage would settle them.
• G-003 base_print_run=unnumbered with base_print_run_basis=assumed. NO source states in words that the 2016-17 Hoops base cards carry no serial numbering. The evidence is strong but indirect: tcdb renders a serial token on every row of eleven of the fourteen base parallel sids and NO token on any of the 300 base rows, and cardboardconnection prints "300 cards." with no figure. EXTRACTION_SPEC 2.1 is explicit that this shape is assumed, not sourced, and it has not been talked up.
• G-004 CARD COUNT CONTESTED, both readings sourced, 20 written. tcdb sid 137617 renders the header "Total Cards: 10" and renders card numbers 11-20 only, while the SAME tcdb page carries the Notes line "#1-10 issued in NBA2K17 video game" and cardboardconnection states "20 cards. First ten cards are exclusive to 2K17 Legend Edition video games." Both statements are true of different things: the set has 20 numbered cards and tcdb catalogues only the ten that reached packs. card_count=20 is written because two independent lineages support it and because tcdb's own Notes line names #1-10 as issued. NOTE FOR THE Total Cards LEDGER: this is a SIXTH thing tcdb's Total Cards has been measured counting - rows the site has catalogued, excluding issued numbers it has not. The ten video-game cards have no card-level source and are not written anywhere.
• G-005 SET NAME CARRIES A SOURCE TYPO AND WAS NOT SILENTLY CORRECTED. tcdb's own heading for sid 137629 reads "2016-17 Hoops - NBA FInals MVP" with a capital I in FInals; cardboardconnection prints "Finals MVP"; paninihoops-bkb-2017-18 carries a set named "Finals MVP". The tcdb string is written verbatim per CLAUDE.md (names are what sources print, and a resolver has to match them) rather than repaired to a third string that no source prints. A naming decision is needed across the line.
• G-006 THE SET IS SERIAL NUMBERED AT FOUR DIFFERENT RUNS AND sets.csv CANNOT HOLD ANY OF THEM. The distinct token set over every rendered row of sid 137622 was enumerated, not sampled: SN2016 on cards 1-44, SN999 on 45-66, SN499 on 67-79, SN199 on 80-86 (Checklist.cfm page 1 gave 69 rows, ?sort=0 reached 75, Gallery.cfm pages 8 and 9 closed 71-86). These are per-playoff-round runs on different cards, not parallel tiers, so they are not parallels.csv rows either; EXTRACTION_SPEC 2.2 gives sets.csv no print_run column, so the figures survive only in that row's notes and here. This is the same expressiveness gap CAV-041 records, in a new location.
• G-007 RED HOT SIGNATURES IS FILED AS A PARALLEL HERE AND AS A SET ONE SEASON UP. This build files "Red Hot Signatures" (sid 137587) and "Rookie Red Hot Signatures" (sid 137589) as parallels.csv rows scoped to "Hot Signatures" and "Rookie Hot Signatures", because each renders a roster IDENTICAL to its parent set - 1-60 with #53 absent, first row Cody Zeller, last row Andrew Wiggins; and 1-40 with 37 rendered, first row Brandon Ingram, last row Anthony Barber - and adds only SN25 to the parent's AU, which is the definition of a parallel; cardboardconnection agrees, listing "Red Hot #/25" as a PARALLEL CARDS line under each set. data/paninihoops-bkb-2017-18/sets.csv instead carries "Red Hot Signatures" (50) and "Red Hot Signatures Rookies" (40) as set_type=auto rows, with no corresponding parallels.csv rows. Same line, adjacent seasons, opposite convention, and nothing in either release tells a consumer which it is under. Recorded rather than smoothed, per EXTRACTION_SPEC 6.
• G-008 FAMILIES HAS NO PATTERN FOR tcdb'S APOSTROPHE FORM. tcdb names sid 137592 "Bird's Eye View Artist's Proof", so this release carries a parallel named "Artist's Proof" alongside four named "Artist Proof". family_of() returns "Artist Proof" for the latter and "other" for the former, because FAMILIES carries the pattern ("Artist Proof","Artist Proof") and nothing matching "Artist's Proof". The derived value was written as family_of() returns it and NO entry was added to pipeline/fix_schema.py (CAV-039: 268 rows already carry hand-written families and a corpus re-derivation would destroy them, so the list is not edited from inside a slice). An ordered-list fix would have to be checked corpus-wide before being called safe, since a pattern for the apostrophe form must precede "Artist Proof" or the shorter one swallows it - the CAV-019 Cosmo/Cosmos shape.
• G-010 CHANNEL ON THE INSERT-SCOPE TIERS IS from-note, NOT sourced. cardboardconnection writes, per insert set, a block of the shape "20 cards. Hobby Only." followed by a separate "PARALLEL CARDS: Artist Proof #/25, Holo Gold #/10." line. The Hobby Only / Retail Only sentence is a statement about the SET; applying it to the set's tiers is an inference, so those rows carry channel_basis=from-note and channel_confidence=single rather than sourced-exclusive. The five RETAIL-ONLY base parallels are different: cardboardconnection names those tiers individually under the heading "RETAIL-ONLY PARALLEL CARDS", so they carry sourced-exclusive. Rookie Remembrance has no Hobby/Retail sentence at all, so its Prime tier is channel=UNKNOWN with an EMPTY basis, not unsupported (migrations/003 rejects that pairing).
• G-011 https://www.tcdb.com/Errors.cfm/sid/134315/2016-17-Hoops returns "No errors or variations have been added for this set yet." NOTHING IS CONCLUDED FROM THAT (CAV-020: the endpoint returns the identical sentence on sets that provably carry UER rows, so its silence discriminates nothing). What does survive is independent: the base run is contiguous 1-300 with no b-suffixed row rendered on any of six Gallery pages or three Checklist pages. Whether a photo-variation layer exists for 2016-17 Hoops is UNSUPPORTED, not disproved.
• G-012 FOUR SETS WHOSE HEADER AND NUMBER SPAN DISAGREE, EACH RECONCILED ON THE DISTINCT-NUMBER UNIT, WITH THE INDIVIDUAL HOLES UNIDENTIFIED. Great Significance: header 96, Gallery page 10 renders 95-100, so four numbers in 1-100 are unissued and I did not identify which four (its Gold parallel renders 82 rows spanning 1-85 with 56 and 61-63 absent, which is consistent but is a truncated render, not a complete one). Hot Signatures: header 59, rows 1-60 with #53 absent - CAV-044 satisfied, because its PARALLEL page (Red Hot Signatures, sid 137587) independently renders the same 1-60 span with the same hole. Rookie Hot Signatures: header 37, span 1-40, three numbers unissued and not individually identified; its parallel sid 137589 renders the same span. Rise and Shine Memorabilia: header 39, rows 2-40, no card 1 - its Prime parallel sid 139507 renders the same 2-40 span, so CAV-044 is satisfied. No unissued number is recorded anywhere as a fact; card-level work belongs to the CHASE_CHECKLIST_SPEC pass.
• G-013 FIVE SET/TIER NAMES DIFFER BETWEEN THE TWO SOURCES AND tcdb'S FORM WAS TAKEN EACH TIME, BECAUSE tcdb IS THE SOURCE THAT ALSO SUPPLIES THE SID. tcdb "Great Significance" vs cc "Great SIGnificance Autographs" (and paninihoops-bkb-2017-18 carries "Great SIGnificance"); tcdb "Rise and Shine Memorabilia" vs cc "Rise N Shine Memorabilia" (2017-18 carries "Rise N Shine Memorabilia"); tcdb "Rookie Hot Signatures" vs cc "Hot Signatures Rookies"; tcdb "Spark Plugs" vs cc "Sparkplugs"; and the High Flyers gold tier, which tcdb names "High Flyers Gold" while cc prints "Holo Gold #/10" exactly as it does for the four sets whose tcdb name IS Holo Gold. The last one matters most: this release therefore carries a set-scope parallel named "Gold" under High Flyers and four named "Holo Gold" elsewhere, which a resolver keyed on the cc form will not match.
• G-014 NOT ATTEMPTED IN THIS PASS, BY DESIGN. insert_cards.csv is governed by CHASE_CHECKLIST_SPEC and is a separate pass; EXTRACTION_SPEC 2 forbids writing it here. All 42 child sids are recorded in sets.csv and parallels.csv notes, so that pass does not need to re-enumerate them. One item is worth carrying forward: sid 137586 Hot Signatures renders EXCH on one of its rows alongside AU on 59 - an exchange/redemption card, which converts an unresolved placeholder into evidenced absence and must be recorded rather than dropped.
• G-015 TWO COVERAGE CHECKS THE SPECS REQUIRE, BOTH RUN, BOTH NEGATIVE, AND THE SECOND IS THE WEAKER OF THE TWO. (a) CHASE_CHECKLIST_SPEC records tcdb filing some of a Hoops release's sets under a SIBLING parent named "Hoops Winter" in three consecutive seasons. tcdb's own 2016 season index (https://www.tcdb.com/ViewAll.cfm/sp/Basketball/year/2016), read across every section heading, contains exactly ONE entry whose name contains "Hoops": "2016-17 Hoops", sid 134315. There is no sibling parent for 2016-17, so that hazard does not apply here. (b) EXTRACTION_SPEC 5.5's mirror says this season's Update blocks can be filed under the NEXT season's parent. https://www.tcdb.com/Inserts.cfm/sid/153685/2017-18-Hoops was asked directly whether any row's name contains "2016-17" or "Update" and answered NO. THAT NEGATIVE IS BOUNDED: it rests on a single fetch of a page type that is documented to truncate on large parents, and no second differently-varied fetch was run, so it is weaker evidence than (a). No prior-season subset from another product was found under THIS parent either - all 56 rows of its Inserts.cfm carry 2016-17 Hoops names - so EXTRACTION_SPEC 5.5's file-them-in-sets.csv rule had nothing to act on.

If you know a source that answers one of these, tell us — name the question's label and where to look.