2016-17 Select

• G-001 ENUMERATION COMPLETENESS AND ITS LIMIT. The set inventory rests on tcdb's parent Inserts.cfm, fetched SECTION BY SECTION per caveats.csv CAV-047 rather than as a whole page: https://www.tcdb.com/Inserts.cfm/sid/147246/2016-17-Panini-Select returned the heading 'Parallel Sets (15)' and 15 numbered rows, and https://www.tcdb.com/Inserts.cfm/sid/147246/2016-17-Panini-Select?x=1 returned 'Insert Sets (55)' and 55 numbered rows. Both counts match their own section headings and both were counted by me, not taken from the fetch's tally. No refusal and no truncation occurred on either ask. 15 + 55 = 70 rows, and the 55 insert rows decompose exactly into 10 named sets plus 45 colour tiers of those sets, which is an independent arithmetic check that nothing was dropped. WHAT IS NOT ESTABLISHED: Inserts.cfm is documented to omit real sets entirely, and only ONE source enumerates this release's set inventory - cardboardconnection names the same ten insert/auto sets and no eleventh (https://www.cardboardconnection.com/2016-17-panini-select-basketball-cards), but the two may be one lineage (caveats.csv CAV-037). A set tcdb does not file and cardboardconnection does not print would be invisible to this build.
• G-002 base_print_run IS 'unnumbered' WITH basis 'assumed' AND NO SOURCE STATES IT IN WORDS. The evidence is indirect and strong: tcdb's parent renderers print no serial token on any of rows 1-300 while the SAME flag column on those same rows positively renders RC on 60 of them, and the same renderers print SN299, SN199, SN175, SN149, SN99, SN75, SN60, SN49, SN25, SN10, SN5 and SN1 on this release's own parallel sids. EXTRACTION_SPEC section 2.1 classifies exactly that shape as 'assumed' and forbids promoting it to 'sourced'. Panini's own first-look post describes the tiers as 'increasing in scarcity with each tier' (https://blog.paniniamerica.net/panini-america-provides-detailed-first-look-at-2016-17-select-basketball-gallery/), which is an insertion-rate statement and not a print run, so base_print_run is 'unnumbered' and NOT 'varies-by-tier'. select-bkb-2017-18, select-bkb-2018-19 and select-bkb-2019-20 all write the same pair for the same reason.
• G-003 TIER PRINT RUN AND TIER CHANNEL ARE UNRESOLVED ON ALL THREE TIERS. base_card_tiers.csv carries print_run=UNKNOWN, serial_numbered=UNKNOWN and channel=UNKNOWN for Concourse, Premier Level and Courtside. Two independent sources state the tiers' NAMES, RANGES and COUNTS - Panini America's own post (https://blog.paniniamerica.net/panini-america-provides-detailed-first-look-at-2016-17-select-basketball-gallery/) and cardboardconnection (https://www.cardboardconnection.com/2016-17-panini-select-basketball-cards) - and eight tier-exclusive tcdb parallel sids prove both boundaries of each range on rendered pages. Neither source states a print run or a trade channel for any tier. caveats.csv CAV-024 is explicit that UNKNOWN here means unresolved, NOT unnumbered.
• G-004 COVERAGE OF THE SERIAL-TOKEN ENUMERATION ON THE FIFTEEN BASE PARALLELS, STATED EXACTLY. Every one of the fourteen reachable base-parallel sids was probed on Checklist.cfm with one fetch per 100-card block it covers, so a per-block change of value would have been caught; the five 300-card tiers (Silver, Tie-Dye, Gold, Green, Black) were each probed on all three of their pages and none changes value. The eight block tiers additionally had their last Gallery page opened, which both proves the upper boundary and reads the tokens on the final ten rows. WHAT IS NOT COVERED: tcdb's fetch layer truncates, so between 41 and 86 of each 100-row page actually rendered. Unseen rows by tier: Blue 79-90 (12 of 100); White 77-90 (14); Maroon 83-90 (8); Orange 76-90 (15); Light Blue and Purple and Neon Yellow the interior rows their pages dropped plus 181-190; Copper 279-290; Silver, Tie-Dye, Gold, Green, Black roughly a quarter of each of their three pages. The residual risk is a single-card exception inside a block that no fetch rendered. TRUNCATION CENSUS MEASURED ON THIS RELEASE, for caveats.csv CAV-046: my own counts of rendered rows on 100-row Checklist pages were 41, 50, 50, 50, 50, 51, 56, 67, 70, 72, 75, 76, 76, 78, 82 and 86 of 100. Nothing fell below 41. TWO INTERIOR-DROP SHAPES were seen, where the row count disagrees with the number span: Light Blue rendered 50 rows over the span 101-175 and Purple 50 rows over 101-184. Varying the sort did NOT reliably help here - on the parent's page 1 the default sort reached card 81 while ?sort=0 reached only 76, and on page 3 the default sort reached 274 while ?sort=0 reached 277.
• G-005 CHANNEL IS UNKNOWN ON ALL 60 PARALLEL ROWS AND channel_basis IS EMPTY ON ALL 60, DELIBERATELY. The only channel statement found anywhere is a PRODUCT-level one on Panini America's own first-look post: 'Each hobby box of 2016-17 Select Basketball (five cards per pack, 12 packs per box, 12 boxes per case) will deliver three autograph or memorabilia cards and 12 Prizm parallels.' (https://blog.paniniamerica.net/panini-america-provides-detailed-first-look-at-2016-17-select-basketball-gallery/). That places the product in hobby boxes; it does not claim any tier is exclusive to hobby, and it says nothing about whether a retail configuration existed. CLAUDE.md and EXTRACTION_SPEC section 2.3 both say that when you decline to assign a channel the channel is UNKNOWN and the basis is EMPTY - 'unsupported' describes a KNOWN channel you could not source and would be rejected by the database here. cardboardconnection states no channel for any parallel on its product page (https://www.cardboardconnection.com/2016-17-panini-select-basketball-cards).
• G-006 A PARALLEL tcdb LISTS BUT WHOSE PAGE CANNOT BE OPENED. 'Pink Prizm', sid 614736, is the tenth row of tcdb's 'Parallel Sets (15)' section on the parent (https://www.tcdb.com/Inserts.cfm/sid/147246/2016-17-Panini-Select), rendered on two differently-parameterised fetches of that page with an EMPTY Notes cell, the second of which returned its href as https://www.tcdb.com/Checklist.cfm/sid/614736/2016-17-Panini-Select-Pink-Prizm . That URL returns '404 - Page Not Found'; so does https://www.tcdb.com/ViewSet.cfm/sid/614736/2016-17-Panini-Select-Pink-Prizm and a ViewSet.cfm call under an arbitrary slug. This is NOT the documented non-existent-sid signature, which is the tcdb HOME PAGE rendering rather than a 404, so it is a new failure shape worth recording. Consequence: the row is written with print_run, serial_numbered and scope_type all UNKNOWN - its roster, its tier scope and its print run are all unread. No other source corroborates it: cardboardconnection's base-parallel sentence names ten tiers and no Pink (https://www.cardboardconnection.com/2016-17-panini-select-basketball-cards), Panini's own first-look post does not mention Pink (https://blog.paniniamerica.net/panini-america-provides-detailed-first-look-at-2016-17-select-basketball-gallery/), and a web search for a 2016-17 Select Pink Prizm returned nothing. Its sid sits in a 614xxx range while the other fourteen sit in 1486xx.
• G-007 FOUR SET COUNTS DISAGREE BETWEEN tcdb AND cardboardconnection AND ARE RECORDED RATHER THAN MERGED. (1) DUETS: cardboardconnection heads 'Duets Relics Set Checklist' at '24 cards' (https://www.cardboardconnection.com/2016-17-panini-select-basketball-cards); the union of three fully enumerated tcdb pages is the complete run 1-25, twenty-five distinct numbers each rendered with a player pair - the base page carries #17, the Black page carries #8, #12 and #22, the Purple page carries #19 and #25. sets.csv records 25. (2) IN FLIGHT SIGNATURES: cardboardconnection says '14 cards'; tcdb's Orange parallel renders 'Total Cards: 18' and enumerates 18 rows, and the set's number span reaches 20. sets.csv records 18. (3) ROOKIE SIGNATURES: THREE figures - cardboardconnection '34 cards', tcdb's four colour parallels 'Total Cards: 34', tcdb's own base page header 'Total Cards: 35' against 36 distinct numbers that base page actually renders, the identical 36-number sequence returned by two differently-varied fetches. sets.csv records 36. (4) SIGNATURES: cardboardconnection '35 cards' against 'Total Cards: 36' on tcdb's base page and on all four of its colour parallels. sets.csv records 36. In every case the figure written is the one carried by a page whose header and its own enumerated rows agree, and the other figure is quoted here beside it.
• G-008 THREE INSERT/AUTOGRAPH SETS HAVE A PER-CARD VARIABLE PRINT RUN OF THEIR OWN THAT sets.csv CANNOT CARRY. sets.csv has no print_run column (EXTRACTION_SPEC section 2.2) and these are not parallel tiers, so the figures survive only here and in each set's notes. DIE CUT AUTOGRAPHS (sid 148633): its 58 enumerated rows carry AU plus four distinct serial tokens - SN49, SN60, SN75 and SN99. SIGNATURES (sid 148671): AU plus SN99 and SN149. THROWBACK MEMORABILIA (sid 148682): its 36 enumerated rows carry MEM plus SN199 on most rows, SN122 on #23 Eric Gordon, SN85 on #19 LeBron James and SN50 on #22 Michael Beasley - and tcdb's own Inserts.cfm Notes cell for that set reads 'MEM, SN199 or less' (https://www.tcdb.com/Inserts.cfm/sid/147246/2016-17-Panini-Select), a CEILING, which EXTRACTION_SPEC section 2.3 says is not a print run. Two further sets are uniform and their figures are likewise unrecordable in the schema: Die Cut Rookie Autographs is AU+SN199 on all 38 rendered rows, Rookie Signatures is AU+SN299 on all 36, Sparks is MEM+SN199 on all 17, In Flight Signatures is AU+SN99 on all 17, Duets is MEM+SN149 on 19 of 20. Rookie Swatches and Select Swatches are UNNUMBERED on a within-product positive discriminative test. This is the fifth-consecutive-season expressiveness gap recorded in caveats.csv CAV-041, in its sets.csv form.
• G-009 EIGHT SET-SCOPE PARALLEL TIERS HAVE A PER-CARD VARIABLE PRINT RUN AND CARRY UNKNOWN/UNKNOWN. All 45 set-scope sids were probed and each was enumerated in full or to a row count equal to its own Total Cards, so these are enumerations and not samples. The eight that disagree with themselves: Die Cut Autographs Pulsar (sid 148636) AU plus {SN35, SN49, SN60}; Die Cut Autographs Scope (148637) AU plus {SN25, SN49}; Duets Gold Prizms (148646) MEM plus {SN10, SN5}, the SN5 on card #4 Blake Griffin / Chris Paul; Duets Purple Prizms (148647) MEM plus {SN99, SN78}, the SN78 on card #25 James Harden / Patrick Beverley; Duets Tie-Dye Prizms (148648) MEM plus {SN25, SN10}, again on card #4; Sparks Gold Prizms (148679) MEM plus {SN10, SN2}, the SN2 on card #24 Brandon Knight; Sparks Tie-Dye Prizms (148681) MEM plus {SN25, SN5}, again on card #24; Throwback Memorabilia Tie-Dye Prizms (148687) MEM plus {SN25, SN21}, the SN21 on #2 Luol Deng and #24 Isaiah Canaan. Each single-row exception was confirmed on a second differently-varied fetch, or corroborated by the SAME card being the exception in a sibling tier of the same set, which is why none is treated as a dropped token. The other 37 set-scope tiers are uniform. EXTRACTION_SPEC section 2.3 forbids writing the modal token, so print_run and serial_numbered are UNKNOWN on all eight and the real figures survive only in visual_notes and here. Note the structural result: 8 of 45 set-scope tiers (18 percent) are variable, and 0 of the 14 READABLE base parallels are - which reproduces the standing generalisation that variability concentrates in the chase sets and not in the base ladder, and that nothing at the 1/1 end varies (every Black and every Power tier in this release is uniform SN1).
• G-010 tcdb'S OWN BASE PAGES ARE SHORT OF THEIR OWN PARALLEL PAGES ON SIX SETS, AND FOUR OF THE MISSING CARDS ARE NAMED HERE BY DIFFING. DIE CUT AUTOGRAPHS: base page 58 rows against the Scope parallel's 60; the two cards the base page lacks are #20 Carmelo Anthony (read off https://www.tcdb.com/Checklist.cfm/sid/148637/2016-17-Panini-Select-Die-Cut-Autographs-Scope?PageIndex=1&x=1 ) and #60 Andrew Wiggins (read off https://www.tcdb.com/Checklist.cfm/sid/148634/2016-17-Panini-Select-Die-Cut-Autographs-Mosaic?PageIndex=1 ). DIE CUT ROOKIE AUTOGRAPHS: base page 38 rows against the Mosaic parallel's 39; the missing card is #18 Patrick McCaw (read off https://www.tcdb.com/Checklist.cfm/sid/148639/2016-17-Panini-Select-Die-Cut-Rookie-Autographs-Mosaic?PageIndex=1&x=1 ). IN FLIGHT SIGNATURES: base page 17 rows against Orange's 18; the missing card is #11 Chris Paul. DUETS: base page 20 rows, and the Black and Purple pages together supply #8, #12, #19, #22 and #25. SELECT SWATCHES: base page 60 rows against Black's 64. SPARKS: base page 17 rows against Black's 25. CONSEQUENCE FOR A LATER CHASE-CHECKLIST PASS: the parent-side page is not the right enumeration for these sets; the largest colour tier is. TWO GENUINELY UNISSUED NUMBERS were established under caveats.csv CAV-044 by opening parallel pages rather than trusting the base page: Die Cut Rookie Autographs #24 is absent from the base, Mosaic AND Scope pages, and Duets has no number missing once the three pages are unioned.
• G-011 FOUR SET NAMES DIFFER BETWEEN tcdb AND cardboardconnection AND OUR ROWS USE tcdb'S. tcdb heads them 'Duets', 'Sparks', 'Die Cut Autographs' and 'In Flight Signatures'; cardboardconnection heads the same four 'Duets Relics Set Checklist', 'Sparks Relics Set Checklist', 'Die-Cut Autographs Set Checklist' and 'In-Flight Signatures Set Checklist' (https://www.cardboardconnection.com/2016-17-panini-select-basketball-cards). The differences are an added 'Relics' and a hyphen. tcdb's strings are used because tcdb is the structural source this build enumerates from and CHASE_CHECKLIST_SPEC requires insert_cards.set_name to match sets.csv byte for byte; a later pass working from cardboardconnection will need this mapping. Note also that cardboardconnection heads the three base tiers 'Concourse Level Base Set Checklist', 'Premier Level Base Set Checklist' and 'Courtside Level Base Set Checklist' while its own prose - and Panini's own post - call them 'Concourse', 'Premier Level' and 'Courtside'; base_card_tiers.csv uses the prose forms, which also match select-bkb-2017-18, select-bkb-2018-19 and select-bkb-2019-20.
• G-012 ROOKIE FLAGS ARE TAKEN FROM tcdb'S RC TOKEN AND NOTHING ELSE. Sixty of the 300 base rows render 'RC' - 19 in Concourse, 21 in Premier Level and 20 in Courtside. rookie_flag='no' rather than UNKNOWN is licensed because the flag column is live in BOTH directions on every page used: each rendered page carries RC on some rows and not on others. caveats.csv CAV-003 records that most Panini checklist sources do not print RC designations at all, so this is a single-source claim; cardboardconnection was not read for rookie designations. A player can hold several base cards here - Buddy Hield is #1, #122 and #270 - so a flag difference between tiers for one player would not be visible to this build.
• G-013 NO VARIATION LAYER WAS LOOKED FOR AND ITS ABSENCE IS NOT ESTABLISHED. Errors.cfm was not fetched for this release. caveats.csv CAV-020 (the Errors.cfm caveat, per CAV-048's disambiguation) records that its silence is worthless anyway - it returns the same 'no errors or variations' sentence on a set that provably has UER rows - and the 2016-17 scope memo section 7.3 found it returning that sentence on both 2016-17 parents it was tried on. What DOES survive: base_set_size=300 and base-run contiguity rest on an independent contiguous enumeration, and every base card_number is a plain integer with no b-suffix rendered on any page opened. What does NOT survive: the claim that no variation set exists. Gallery.cfm is documented to strip a 'b' suffix, and seven of this release's Courtside rows are known only from Gallery tiles.
• G-014 SOURCE INDEPENDENCE, STATED PLAINLY. This build rests on THREE lineages, which is one more than the 2016-17 scope memo's Phase 0 had: tcdb (structural), cardboardconnection (editorial) and blog.paniniamerica.net (first-party). The first-party post is what makes the tier structure genuinely multi-sourced - it states the tier names AND their ranges in one sentence and independently states the tier-exclusivity of the Prizm ladder. checklistcenter, checklistinsider, sportscardchecklist and waxstat were NOT swept; per caveats.csv CAV-037 a constructed checklistcenter URL that 404s is evidence of nothing, so none was constructed. Beckett is confirmed unreachable (CAV-033) and was not budgeted for. Marketplace corroboration was not attempted. Where a print run rests on tcdb alone - Light Blue Prizms, Purple Prizms, Neon Yellow Prizms, Copper Prizms and all 45 set-scope tiers - source_confidence is 'single'. Where cardboardconnection merely OMITS a figure, as for Silver Prizms and Tri-Color Prizms, that is an omission and not a statement (EXTRACTION_SPEC section 2.1), so those rows are 'single' too and not 'multi'.
• G-015 THE STRUCTURAL FINDING OF THIS BUILD, RECORDED BECAUSE parallels.csv CANNOT SHOW IT AT A GLANCE: A PARALLEL'S ROSTER IS NOT THIS RELEASE'S BASE RUN, AND THE FIFTEEN BASE PARALLELS FALL INTO FOUR DIFFERENT SHAPES. Five span the whole 300 (Silver, Tie-Dye, Gold, Green, Black - each 'Total Cards: 300', pagination to page 3). Eight cover exactly one 100-card tier and are filed scope_type=block: Blue, White, Maroon and Orange over Concourse 1-100; Light Blue, Purple and Neon Yellow over Premier Level 101-200; Copper over Courtside 201-300. ONE covers TWO tiers and not the third - Tri-Color Prizms renders 'Total Cards: 200' with page 1 starting at card 1 and page 2 starting at card 101 and no page 3, so it spans 1-200 and excludes Courtside. One is unreadable (Pink Prizm, see G-006). Panini's own first-look post states the mechanism in words - 'each base tier features unique Prizm offerings' (https://blog.paniniamerica.net/panini-america-provides-detailed-first-look-at-2016-17-select-basketball-gallery/) - which is why the eight tier-exclusive tiers are a finding about the product and not about tcdb's cataloguing. CONSEQUENCE FOR THE SCHEMA: block rows carry scope_set_id NULL unconditionally, so all ten block-and-UNKNOWN rows in this release had to be globally unique by NAME; they are, with no name composed by me - every one is the string tcdb prints. CONSEQUENCE FOR THE CATALOGUE: select-bkb-2017-18 files all seven of its whole-run base parallels as scope_type='base-set' and twelve others as block. A consumer comparing the two seasons will find 2016-17 with only 5 base-set-scope base parallels of 15. That is a real difference in the products, not a convention divergence, but it is worth stating because the two seasons look alike.
• G-016 A parallel_family PATTERN THE DERIVATION LIST DOES NOT HAVE. parallel_family was derived by AST-lifting FAMILIES and family_of() out of pipeline/fix_schema.py, per caveats.csv CAV-039, and NOTHING was added to that file. The derivation returns 'other' for the two rows named 'Power' - the Power tier of Die Cut Autographs (sid 148635) and of Die Cut Rookie Autographs (sid 148640). 'Power' is a real Select die-cut finish name that sits beside Mosaic, Pulsar and Scope in the same ladder, and those three DO resolve to families of their own ('Mosaic', 'Pulsar', 'Scope'), so one member of a four-member visual system lands in 'other' while its three siblings do not. The derived value is written as returned. Per CAV-019 (the FAMILIES substring caveat, per CAV-048's disambiguation) any new short pattern needs a corpus-wide dry run before it is added, and adding one is not this build's call.
• G-017 ONE ROW'S FLAG COLUMN CAME BACK EMPTY ON TWO FETCHES AND IS STILL NOT EVIDENCE OF ABSENCE. Duets (sid 148643) row #13, Dirk Nowitzki / Deron Williams, renders an EMPTY flag column on https://www.tcdb.com/Checklist.cfm/sid/148643/2016-17-Panini-Select-Duets?PageIndex=1&sort=0 and again on the same endpoint under the default sort, while all 19 other rows of that page render 'MEM, SN149'. The same card renders 'MEM, SN99' on the Purple page, 'MEM, SN49' on the Copper page and 'MEM, SN1' on the Black page, so the card exists and is serial numbered in every tier that carries it. print_run=149 is written for the Duets base tier on the strength of the 19 rows that did render, and the blank is treated as the dropped token caveats.csv CAV-042 describes. What is NOT established is that card #13's own Duets-tier run is 149 rather than something lower - three of this release's other tiers have exactly one card numbered lower than the rest, so a single low card is a live possibility here. Closing this needs a differently-varied third endpoint on that one row.

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