2016-17 Optic

• G-001 RELEASE DATE CONTESTED AND NOT MERGED. tcdb's parent page renders "Release Date: April 14, 2017", unqualified (https://www.tcdb.com/ViewSet.cfm/sid/146985/zzz-optic-probe-slug). THREE other pages print April 5: Panini's own blog, published 2017-02-20, "Set to arrive in hobby shops across North America on April 5" (https://blog.paniniamerica.net/2017/02/20/panini-america-provides-detailed-first-look-at-2016-17-donruss-optic-basketball/); blowoutbuzz, published 2017-01-31, "Arrives: April 5" (https://blowoutbuzz.wordpress.com/2017/01/31/first-buzz-2016-17-donruss-optic-nba-retail-only-details-for-now/); and cardboardconnection's product page, "Release Date: 4/5/2017" (https://www.cardboardconnection.com/2016-17-donruss-optic-basketball-cards). 2017-04-14 is written and April 5 is preserved here. TWO CANDIDATE EXPLANATIONS WERE TESTED AND BOTH FAILED. (1) That cardboardconnection's figure was an artefact of its season index, which 2016-17 Phase 0 proved is not a release-date field: REFUTED - the index row also reads "Apr 05, 2017" (https://www.cardboardconnection.com/sports-cards-sets/nba-basketball-cards/2016-2017-basketball-cards), so index and product page agree and the figure is cardboardconnection's own. (2) That April 5 and April 14 are separate retail and hobby dates: REFUTED - blowoutbuzz gives retail and hobby details on one page under one arrival date and states no second date. WHAT DECIDED IT: publication distance. All three April 5 pages are PRE-RELEASE FORECASTS, two of them measurably 6.5 and 9 weeks ahead of the event, and they plausibly descend from one Panini announcement rather than being three lineages. tcdb's is the only retrospective record. This is the same shape the 2016-17 scope memo already adjudicated for Studio (Panini blog forecast "May 31" against tcdb "June 9"; tcdb preferred). UNRESOLVED: whether April 14 is a slipped street date or tcdb's own later entry date. A retailer invoice or a dated first-break post would settle it.
• G-002 NO base_card_tiers.csv WAS WRITTEN, AND THIS DIVERGES FROM BOTH SEASONS ABOVE THIS ONE IN THE CATALOGUE. The 150/50 structure is real and measured: tcdb renders the flag token "RC" on every parent row 151-200 and on no row 1-150. What is NOT sourceable in 2016-17 is a NAME for the 151-200 block. optic-bkb-2017-18 and optic-bkb-2018-19 each carry a base_card_tiers.csv row named "Rated Rookies" for 151-200, and in those seasons the name is supported by tcdb rendering the tokens "RR, RC" on that block. In 2016-17 tcdb renders ONLY "RC": I asked Gallery.cfm?PageIndex=18 directly whether "RR" appears anywhere on the page and the answer was no (https://www.tcdb.com/Gallery.cfm/sid/146985/2016-17-Donruss-Optic?PageIndex=18&sort=0). cardboardconnection was asked whether it names a subset within the base set and a fetch returned the word "Rookies" with a 151-200 range; that was NOT written, because this release also contains an insert set literally called "The Rookies" (tcdb sid 150406) and the answer was not reproduced as a verbatim quoted heading, so it fails the cross-check-set-NAMES test. Writing "Rated Rookies" would have been inherited from an adjacent season, which the 2016-17 brief forbids. CONSEQUENCE: card_tier is empty on all 200 base rows and a consumer comparing 2016-17 against 2017-18 sees a tier structure appear from nowhere one season up. TO RESOLVE: a 2016-17 source naming the block in words, e.g. a Panini sell sheet or the wrapper.
• G-003 77 OF THE 88 INSERT-SCOPE PARALLEL ROWS CARRY print_run=UNKNOWN / serial_numbered=UNKNOWN, AND UNKNOWN HERE MEANS UNRESOLVED, NOT UNNUMBERED. Their EXISTENCE is sourced - every one has its own tcdb sid taken from the "Insert Sets (102)" section of https://www.tcdb.com/Inserts.cfm/sid/146985/2016-17-Donruss-Optic, which I counted myself at exactly 102 rows. Their print runs are not, for a specific measured reason: this page is the CAV-036 DONRUSS SHAPE, where the two sections of one page behave differently. Its "Parallel Sets (13)" section splits cleanly - 9 annotated, 4 blank - and a row-level probe of all 13 confirmed the split in BOTH directions, so blank there means unnumbered. Its "Insert Sets" section does not: 94 of its 102 cells are blank, and of the 11 annotated insert cells I probed at row level, 11 of 11 are serial numbered. A blank cell in that section therefore carries zero information and the both-directions test cannot be run there. Writing "unnumbered" off those blanks would have been the exact false-value failure CAV-036 exists to prevent. A per-sid row-level token sweep of the remaining 77 sids was NOT run in this build - it is roughly 77 further fetches - and that is the whole reason the value is missing. TO RESOLVE: sweep each sid on Checklist.cfm for its distinct token set.
• G-004 THREE AUTOGRAPH SETS HAVE A PER-CARD VARIABLE PRINT RUN THAT THIS SCHEMA CANNOT EXPRESS, AND NO FIGURE WAS WRITTEN FOR ANY OF THEM. Dominator Signatures (tcdb sid 150343), Hall Dominator Signatues (150353) and Rookie Dominator Signatures (150367) each render the distinct token set {AU, SN25, SN50, SN99} - three different serial figures inside one set - enumerated rather than sampled. Per EXTRACTION_SPEC 2.3 a tier whose own rows disagree gets no modal token, and sets.csv has no print_run column at all, so the finding survives only in those rows' notes and here. This is the fifth-consecutive-season expressiveness gap recorded in CAV-041; these three are SETS rather than base-scope parallels, so nothing had to be deleted, but nothing could be recorded as a value either. NOTE ALSO A SUSPICIOUS-UNIFORMITY SIGNAL, INVESTIGATED AND NOT ACTED ON: all three render the IDENTICAL token set, which is the template-fill smell of fabrication check 5. Their rosters differ substantively (current stars, Hall of Famers, rookies) and they are three sibling sets of one family in one product sharing a serial ladder, which is structurally expected, so the signal was investigated rather than treated as proof. It is recorded so a later pass can re-test it.
• G-005 FOUR tcdb "Total Cards" HEADERS DISAGREE WITH MY OWN COUNT OF THE RENDERED ROWS, AND EACH IS RECORDED AS A QUESTION ABOUT THE UNIT RATHER THAN AS PROOF THE HEADER IS STALE. Hall Dominator Signatues (sid 150353): header 29, 30 rows rendered over 1-30. Rookie Dominator Signatures (150367): header 29, 30 rows over 1-30. Rookie Signatures (150381): header 45, 44 rows over a 1-50 span. Dominator Signatures (150343): header 35, 35 rows over a 1-40 span, so five numbers in the span are unissued and the exact set of them is not resolved. The header figure was written to card_count in every case, because on 2017-18 National Treasures a header disagreeing with a row count turned out to be RIGHT (it counted distinct issued numbers). Two of the four are off by exactly one in the same direction, which is consistent with one convention rather than four separate errors. NOT resolved. TO RESOLVE: an independent roster for each set.
• G-006 SIGNATURE SERIES CARD COUNT IS CONTESTED BETWEEN ITS OWN BASE PAGE AND ITS OWN PARALLELS, AND THE TIERS ARE NOT NESTED. The base sid 150390 states "Total Cards: 46" and renders 46 rows over a 1-50 span - self-consistent. Its six probed parallels state SIX DIFFERENT totals over the same span: Black (150391) 49, Blue (150392) 47, Gold (150393) 46, Gold Vinyl (150394) 47, Green (150395) 49, Pink (150397) 47. No page is a superset of any other, which is the non-nested-tier shape. 46 was written because it is the base page's own header agreeing with the base page's own rendered rows, but CAV-044 shows a parent can be rows SHORT of its own parallel and that the evidence shape is identical to the case where the header is right - so the larger figures may be the true roster. The sets.csv tiebreaker that normally settles this is unavailable: this release's sets.csv was built FROM tcdb, so it is not an independent third party. NOT resolved. TO RESOLVE: an independent 2016-17 roster for Signature Series.
• G-007 AN EVIDENCED HOLE AND AN EVIDENCED DISTRIBUTION ROUTE IN ROOKIE SIGNATURES, RECORDED RATHER THAN DROPPED. tcdb's Inserts.cfm Notes cell on Rookie Signatures Holo (sid 150387) reads verbatim "#38 does not exist. Some cards were distributed via Panini Rewards." (https://www.tcdb.com/Inserts.cfm/sid/146985/2016-17-Donruss-Optic), and the Rookie Signatures base sid 150381 independently renders 38 as absent from its 1-50 span - two renderings of the product agreeing. Signature Series Purple (sid 150398) carries a related cell, "Some cards were distributed via Panini Rewards". These are the EXCH shape: a statement that converts an unresolved placeholder into evidenced absence, and the spec says to record them and never drop them. They are carried in those rows' notes. NOT acted on further: per CAV-044 a hole is not writeable as unissued until a parallel page of the same product has been opened for that number specifically, and I did not open one for #38. Card-level checklists are a separate pass.
• G-008 EVERY SOURCED CHANNEL ON THIS RELEASE'S BASE PARALLEL LADDER IS A RETAIL FORMAT, WHICH REVERSES AGAINST THE SEASON DIRECTLY ABOVE IT. cardboardconnection states verbatim: "Among the many Opti-Chrome Prizm parallels are Holo, Purple, Orange (#/199 - Fat Pack), Red (#/99 - Blaster), Blue (#/49 - Retail Box), Aqua (#/25 - Gravity Feed), Pink (#/25 - Fat Pack), Gold (#/10 - Blaster), Green (#/5 - Gravity Feed), and Black (1/1 - Retail Box) editions." (https://www.cardboardconnection.com/2016-17-donruss-optic-basketball-cards). Reproduced character for character on a second, cache-busted fetch of the same page before any value was written. optic-bkb-2017-18 records Blue, Gold, Green, Holo and Gold Vinyl as channel=hobby; here the same colour names are Fat Pack, Blaster, Retail Box and Gravity Feed. This is the "nineteen lines reversed against their adjacent season" shape and it is recorded, not smoothed. TWO CAVEATS ON WHAT WAS WRITTEN. (1) Fat Pack, Retail Box and Gravity Feed have NO enum value; those rows carry channel=retail with the SKU in channel_sku, which is the forcing that reference/channel_taxonomy.csv already records as a known defect for fat-pack, hanger, cello and jumbo. Blaster does have an enum value and is written as blaster. (2) The quote states WHERE each parallel is found and never uses the word "exclusive", so channel_basis is sourced-presence and not sourced-exclusive on all eight. Single source (cardboardconnection alone), so channel_confidence=single.
• G-009 THREE OF THE THIRTEEN BASE PARALLELS HAVE NO CHANNEL AT ALL AND ARE NOT ABSENT FROM THE PRODUCT. Checkerboard, White Sparkle and Gold Vinyl do not appear in cardboardconnection's channel sentence quoted in G-008, and Holo and Purple appear in it with no SKU annotation. All five therefore carry channel=UNKNOWN with an EMPTY channel_basis, per EXTRACTION_SPEC 2.3: when you decline to assign a channel the basis is empty, not "unsupported". The product-level hobby statement in Panini's blog ("arrive in hobby shops") was deliberately NOT propagated onto any tier it does not cover. Note Gold Vinyl is a 1/1 and Black is a 1/1 sourced to "Retail Box", which sit oddly together and may mean the sentence is a retail-ladder sentence rather than the whole ladder. NOT resolved.
• G-010 ZERO OF THE THIRTEEN BASE PARALLELS AND ZERO OF THE ELEVEN PROBED INSERT PARALLELS ARE PER-CARD VARIABLE, AND THE DENOMINATORS BEHIND THAT ARE PARTIAL. Each of the thirteen base tiers states "Total Cards: 200" and was swept on Checklist.cfm pages 1 and 2 under ?sort=0. Rows actually rendered per tier ranged from 75 to 159 of 200, so between 41 and 125 rows per tier were NEVER RENDERED to me, on every tier. Every rendered row of every tier carried exactly one token (or none, on the four unnumbered tiers) and no tier showed a second serial token anywhere. The reliable generalisations held: nothing at the 1/1 end varies (Black and Gold Vinyl are uniform SN1 across 158 and 71-plus rendered rows) and variability concentrated entirely in the base tiers of autograph SETS (G-004), not in colour tiers. But "no second token was rendered" is weaker than "no second token exists", and a tier could carry a different figure on a row no fetch showed me. TO RESOLVE: Gallery.cfm top-ups across the unrendered spans of each tier.
• G-011 ONE FETCH-LAYER TALLY WAS SELF-CONTRADICTORY AND ITS COUNT IS NOT RELIED ON ANYWHERE. The Gold Vinyl base parallel probe (sid 150310, Checklist.cfm?PageIndex=2&sort=0) reported "100 rows" over a card-number span of 101-181, which cannot both be true of 81 distinct numbers. Per EXTRACTION_SPEC the substantive clause was taken and the tally discarded: what that row rests on is the DISTINCT TOKEN SET (SN1 only, zero tokenless rows), which is unaffected by the miscount, plus page 1's 71 rows over 1-72 and tcdb's own Inserts.cfm statement "Serial numbered to 1" and cardboardconnection's "Gold Vinyl 1/1". Two further fetches this build produced arithmetic disagreeing with their own lists (the Dominator Signatures token counts summed to 41 over a stated 40 rows; the Rookie Signatures absent-number list contradicted itself in the same sentence) and in both cases only the distinct token set and the verbatim strings were used.
• G-012 THE fetch-layer COPYRIGHT REFUSAL DOCUMENTED ON THIS EXACT PARENT DID NOT RECUR, AND THE REMEDY IS RECORDED FOR THE NEXT AGENT. 2016-17 Phase 0 measured Inserts.cfm/sid/146985 refusing outright ("I need to decline this task as written...") and then volunteering a confident hedged summary anyway. In this build the refusal did NOT fire, because no whole-page ask was ever made: the page was asked for ONE NAMED SECTION per fetch - "Parallel Sets" alone (13 rows, plus its heading text), then "Insert Sets" alone under ?sort=0 (102 rows, numbered so the count was explicit). Both returned complete sections with their Notes cells, no refusal and no truncation, and the 13 + 102 = 115 rows reconcile exactly with the contiguous sid block 150298-150412. NO VALUE IN THIS RELEASE COMES FROM A RESPONSE THAT REFUSED THE UNDERLYING REQUEST. Recorded because the shape of the ask is the finding: per-section, named, with the row count demanded explicitly.
• G-013 TRUNCATION ON Checklist.cfm WENT LOWER THIS BUILD THAN ANY FIGURE IN THE 2016-17 SCOPE MEMO. Phase 0's census of rendered rows on 100-row pages was 50, 50, 50, 75, 79, 86, with the memo saying "nothing below 50". My own census across 25 page fetches includes 25 (Checkerboard sid 150298 page 1) and 40 twice (Holo 150299 page 2, Purple 150300 page 2), so the floor is now 25 of 100. The row-count-versus-number-span tell fired on essentially every page-1 fetch - Checkerboard rendered 25 rows spanning card numbers 1-73, so interior rows were silently dropped rather than a clean tail cut. ?sort=0 beat truncation again on the parent (71 rows to 82). CONSEQUENCE FOR PLANNING: the 60%-of-a-page top-up budget was NOT enough here; on the worst page it would have needed 75%. This does not affect the base run, which was completed to 200 with Gallery.cfm top-ups and both boundaries proved, but it is why G-010's denominators are partial.
• G-014 Errors.cfm WAS NOT CONSULTED AND NO CLAIM HERE RESTS ON ITS SILENCE. Per CAV-020 the endpoint returns the identical "no errors or variations" sentence on sets that provably have errata and on sets that appear not to, so it discriminates nothing and 43 rows in this corpus were marked for concluding from it. WHAT SURVIVES WITHOUT IT: base_set_size, the contiguity of 1-200, and the absence of any numbered block past the base run, all of which rest on an independent contiguous enumeration with both boundaries proved. WHAT IS THEREFORE NOT ESTABLISHED: whether a VARIATION set exists. A variation renders as b-suffixed rows that a base-run enumeration is not designed to surface, and Gallery.cfm strips the b suffix, so this is unsupported rather than disproved. It may still be true.
• G-015 NO BOX-SET OR PREMIUM PRODUCT EXISTS UNDER THIS PARENT, ON POSITIVE EVIDENCE. All 115 child sids of parent 146985 were enumerated from Inserts.cfm and not one row name contains "Premium" or "Box Set" (https://www.tcdb.com/Inserts.cfm/sid/146985/2016-17-Donruss-Optic). This reproduces the direct yes/no asks recorded in the 2016-17 scope memo section 5a and uses the endpoint that memo names as the right one, rather than absence from a season index that is known to under-count exactly this class of product. The brands.csv line opticpremium therefore has no 2016-17 release, and that is measured rather than assumed. The related identity question in EXTRACTION_SPEC 5.2 does not arise here.
• G-016 SEVENTY-FIVE OF 101 PARALLEL ROWS LAND ON parallel_family="other", AND NO FAMILY WAS MINTED OR ADDED. parallel_family and channel_class were derived by AST-lifting FAMILIES, family_of() and CHANNEL_CLASS out of pipeline/fix_schema.py so the repo keeps exactly one definition (CAV-034); fix_schema.py itself was not run and was not edited. Where the names land: Holo -> "Holo" (12 rows), Gold Vinyl -> "Vinyl" (12), Checkerboard -> "Checkerboard" (1), White Sparkle -> "Sparkle" (1), and everything else -> "other" (75). The 75 are the plain colour names - Red, Blue, Green, Gold, Orange, Pink, Purple, Aqua, Black - for which FAMILIES has no pattern, plus Pink, which matches the explicit ("Pink","other") entry that exists precisely to stop "Ink" matching inside it. That is the derivation working as designed, not a miss. NO NEW FAMILY WAS ADDED to fix_schema.py, per CAV-039 and the brief. Flagged so that if a later pass decides plain colours deserve families, this release moves with the rest of the corpus rather than alone.
• G-017 INSERT AND AUTOGRAPH CARD-LEVEL CHECKLISTS WERE DELIBERATELY NOT WRITTEN. insert_cards.csv does not exist for this release; that pass is governed by CHASE_CHECKLIST_SPEC and is out of scope here. The 14 insert and autograph sets are in sets.csv with sourced counts and tcdb sids, and every one renders PLAIN INTEGER card numbers rather than letter codes - which is unusual for autograph sets on this manufacturer and is recorded here because it changes the calibration strategy that pass will need: there are no coded sets in this release to calibrate an aggregator against, so the per-set-type split CHASE_CHECKLIST_SPEC relies on has no purchase and marketplace titles will be the only tiebreaker.

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