2018-19 NBA Hoops Premium Box Set

• G-001 EXTRACTION_SPEC 5.2 BASE-CARD IDENTITY COLLISION, naming both releases: hoopspremium-bkb-2018-19 (this release) and paninihoops-bkb-2018-19 describe the same 300 physical card faces under two release ids. BOTH OVERLAP FIGURES, measured with csv.DictReader on card_number, player_name and team_or_school: POSITIONAL (row index against row index) 299 of 300; BY-CARD-NUMBER 299 of 300. Zero card numbers appear on one side only. Both are reported because they can diverge badly - the 2020-21 pair of this same line reads 175 of 270 positionally against 269 of 270 by card number, because a single row numbered "3b" shifts every row after it. Here the two agree because both files are a clean contiguous 1-300 in the same order. The single non-matching row is card 92 - see G-006. This release is the #/199 Premium Box Set printing of that checklist; the parent is the pack-issued unnumbered printing. Built per 5.2 rather than deduplicated or double-counted. The parent's own gap G-014 anticipated this release by name and says its "Premium" base-set parallel row and its "Premium Boxed Set Autographs" set row belong here; that release was NOT edited by this build (one agent, one release) and a resolver pass should decide which side keeps them. Sources: https://www.cardboardconnection.com/2018-19-panini-nba-hoops-premium-box-set ; https://www.tcdb.com/ViewSet.cfm/sid/188426/2018-19-Hoops-Premium ; https://www.tcdb.com/Inserts.cfm/sid/182809/2018-19-Hoops
• G-002 parallels.csv is intentionally EMPTY (header only), and the evidence for the absence is stronger this season than in 2019-20 without being a statement of absence. (1) tcdb's parent inventory, read twice under two distinct URLs, renders a "Parallel Sets (17)" heading whose 17 members were listed in full, and exactly ONE of them is this product ("Premium", sid 188426); no sub-parallel of it is catalogued, and the whole page contains only two rows whose name includes "Premium" (https://www.tcdb.com/Inserts.cfm/sid/182809/2018-19-Hoops). (2) cardboardconnection's page for this product lists the base cards as a single line, "Serial numbered #/199. Full set per box.", with no second tier (https://www.cardboardconnection.com/2018-19-panini-nba-hoops-premium-box-set). (3) Panini states "Every card in this special version of 2018-19 NBA Hoops Basketball is numbered to just 199" and "Just 199 Hoops Premium Factory Sets have been produced." (https://blog.paniniamerica.net/panini-america-unveils-online-only-2018-19-nba-hoops-basketball-premium-box-set/) - a first-party description of one uniform tier and one production quantity, which is positive evidence that there is no second tier, though it is still not a source saying in words that the box set had no parallels. All six later seasons of this line likewise carry zero base-run parallel rows; that was treated as precedent to VERIFY, not as a result to assume, and the verification above was run independently of it.
• G-003 THE UNIT TRAP OF EXTRACTION_SPEC 5.4 FIRES A SEVENTH TIME ON THIS LINE AND THE GENERAL FORM base_set_size + 1 HOLDS AGAIN; 301 IS WRITTEN IN NO FIELD. The words around every number, with their units named: cardboardconnection prints the LABEL AND VALUE "Product Configuration: 301-card factory set" - the unit is cards per factory set - and immediately decomposes it in a box-contents list reading "Full base set (300 cards)" and "1 Hoops Tribute Autograph". Panini names the unit inside the value in one sentence: "a 301-card offering that includes 300 2018-19 NBA Hoops Basketball cards printed on premium holographic board with gold foil-highlights and one autograph from a current superstar or retired legend numbered to just 10". base_set_size is 300, from tcdb's "Total Cards: 300", from "Inside each 2018-19 Panini NBA Hoops Premium Box Set is a 300-card checklist", from "of the 300 cards in each Premium Factory Set", and from 300 rows and tiles counted one by one. Earlier seasons of this line recorded the same trap with different unit wordings: 2019-20 "305"/"303" cards per sealed box, 2020-21 and 2021-22 "271-card factory set" beside "Full base set (270 cards)", 2022-23 and 2023-24 "301" beside "300", 2024-25 "Box Sets - 301 cards per box set". Sources: https://www.cardboardconnection.com/2018-19-panini-nba-hoops-premium-box-set ; https://blog.paniniamerica.net/panini-america-unveils-online-only-2018-19-nba-hoops-basketball-premium-box-set/ ; https://www.tcdb.com/ViewSet.cfm/sid/188426/2018-19-Hoops-Premium
• G-004 NOT "NBA Hoops Premium Stock", which is a different and ongoing retail line, and the negative rests on positive enumerations rather than on silence. https://www.tcdb.com/ViewAll.cfm/sp/Basketball/year/2018 lists exactly three 2018-19 products whose title contains "Hoops": "2018-19 Hoops" (sid 182809), "2018-19 Hoops Winter" (sid 188435) and "2018-19 Hoops Dallas Mavericks SGA" (sid 663408). No Premium Stock of any kind appears. The same index for the following season, https://www.tcdb.com/ViewAll.cfm/sp/Basketball/year/2019, DOES carry "2019-20 Hoops Premium Stock" (sid 245171), so the term is present in tcdb's vocabulary one season later and genuinely absent here - a positive control on the index rather than an untested absence. The parent's own inventory contains exactly two rows whose name includes "Premium" and none containing "Stock" (https://www.tcdb.com/Inserts.cfm/sid/182809/2018-19-Hoops). A targeted web search for 2018-19 "Premium Stock" returns only 2019-20 and later pages. Nothing in this release derives from a Premium Stock source.
• G-005 CHANNEL IS SOURCED TWICE AND CANNOT BE WRITTEN, for the seventh consecutive season of this line. Panini: "And the only place to get it is at the official online store of Panini America." and the post is headed "Panini America Unveils Online-Only 2018-19 NBA Hoops Basketball Premium Box Set" (https://blog.paniniamerica.net/panini-america-unveils-online-only-2018-19-nba-hoops-basketball-premium-box-set/). cardboardconnection: "The sets are sold online by Panini for $500 a box." (https://www.cardboardconnection.com/2018-19-panini-nba-hoops-premium-box-set). In the EXTRACTION_SPEC 2.3 taxonomy that is channel=panini-direct with channel_basis=sourced-exclusive, but release.csv has no channel column and this release has no parallel row to carry one, so the value lives in prose only. This is a schema expressiveness gap, not a research gap: unlike 2019-20's G-005, which recorded a channel nobody stated, here the channel is stated by the manufacturer and there is still nowhere structured to put it. All seven seasons of this line now in data/ hit it.
• G-006 ONE-ROW DIVERGENCE AGAINST paninihoops-bkb-2018-19, AND IT IS A REAL RENDERING DIFFERENCE INSIDE ONE SOURCE, NOT A TRANSCRIPTION ERROR BY EITHER BUILD. Card 92: this release writes "J.R. Smith"; paninihoops-bkb-2018-19 writes "JR Smith". Both are correct about the page they read. tcdb renders "J.R. Smith" on THIS product's set, confirmed twice under two distinct URLs (https://www.tcdb.com/Gallery.cfm/sid/188426/2018-19-Hoops-Premium?PageIndex=10 and the same page with &sort=0, asked as a narrow single-tile question), and renders "JR Smith" on the PARENT set, likewise confirmed twice (https://www.tcdb.com/Gallery.cfm/sid/182809/2018-19-Hoops?PageIndex=10 and the same with &sort=0). The same source spells one player two ways on two sids in one season. This single row is why both overlap figures in G-001 read 299 of 300 rather than 300 of 300; the card numbers and teams match exactly. The line is not joinable on player_name for this card without an alias. Neither release was edited.
• G-007 NO NON-INTEGER CARD NUMBER AND NO UNISSUED NUMBER THIS SEASON, CHECKED RATHER THAN ASSUMED, because this line has sprung that trap twice (2020-21 rendered "3b" with a "UER" flag and no card 98; 2023-24 rendered two rows numbered 209 and no 208). All 300 numbers 1-300 were observed exactly once and contiguous, with no suffixed, lettered, duplicated or unissued number, across nine Checklist.cfm readings under varied URLs and eight Gallery.cfm pages. https://www.tcdb.com/Errors.cfm/sid/188426/2018-19-Hoops-Premium states "No errors or variations have been added for this set yet." and https://www.tcdb.com/Gallery.cfm/sid/188426/2018-19-Hoops-Premium?PageIndex=26 renders no "UER" token on card 260. No dated correction marker is needed on any base row. NOTE THE CAVEAT IN G-014: Errors.cfm is a FALSE NEGATIVE on the parent of this product, so its silence here was NOT treated as the evidence - the contiguous enumeration was. [CORRECTED 2026-09-20: THE Errors.cfm SENTENCE QUOTED ABOVE IS NOT EVIDENCE AND THE CONCLUSION DRAWN FROM IT DOES NOT HOLD. Measured 2026-09-20 on tcdb sid 182809: Errors.cfm returns that identical sentence for a set whose own renderings carry UER on at least two cards (card 260 Mo Bamba, "UER: Statline says Undrafted, was actually drafted 6th", confirmed under &sort=0; card 67 Justin Jackson, "Career stat line is that of Kosta Koufos"). The endpoint returns the same string on a set that provably HAS errata and on one that appears not to, so it discriminates nothing and its silence is uninformative. WHAT SURVIVES: any claim here about base_set_size, about the base run being contiguous, or about no numbered block sitting past the base run, because those rest on the independent contiguous enumeration also cited in this note - the last rendered Gallery page, the empty page after it, and the per-row flag column. WHAT DOES NOT SURVIVE: the narrower claim that NO VARIATION SET EXISTS. A variation set renders as b-suffixed rows (101a/101b) that a base-run enumeration is not designed to surface, and Errors.cfm was the documented route to it, so that claim is now unsupported rather than disproved - it may still be true. Not re-fetched; recorded so it is not read as settled. See reference/caveats.csv CAV-020.]
• G-008 SOURCES DISAGREE ON THE ROOKIE COUNT BY EXACTLY ONE CARD, AND THIS BUILD WROTE THE PER-CARD SOURCE'S ANSWER. tcdb renders the literal token "RC" on 41 cards: 233 and 241-280. Observed twice and independently - https://www.tcdb.com/Checklist.cfm/sid/182809/2018-19-Hoops?PageIndex=3 answered 233 and 241-273 when asked only which numbers carry "RC", and https://www.tcdb.com/Gallery.cfm/sid/182809/2018-19-Hoops?PageIndex=24 through 28 rendered "RC" on 233 and on all of 241-280 tile by tile. Panini's own launch post instead says "of the 300 cards in each Premium Factory Set, 20 are from the Tribute subset, 240 are from the base set and 40 are Rookie Cards featuring the likes of DeAndre Ayton, Marvin Bagley III, Luka Doncic, Trae Young, Jaren Jackson Jr. and more" (https://blog.paniniamerica.net/panini-america-unveils-online-only-2018-19-nba-hoops-basketball-premium-box-set/) - 40, not 41, and its block arithmetic 240 + 40 + 20 = 300 puts card 233 (Svi Mykhailiuk, a 2018 draftee sitting inside the 1-240 block) on the base side. rookie_flag=yes is written for 233 because tcdb's is a per-card designation from the only source that publishes per-card designations, and because paninihoops-bkb-2018-19 reads the same way, while Panini's sentence is a block-composition summary of where cards come from rather than a statement about which cards carry an RC designation. The disagreement is recorded rather than smoothed over; a reader who prefers the first-party count should read 40 rookies at 241-280.
• G-009 ROOKIE FLAG RESTS ON THE PARENT'S RENDERINGS, NOT ON THIS PRODUCT'S OWN, AND A POSITIVE CONTROL WAS RUN BEFORE TREATING THE SILENCE AS INFORMATIVE. tcdb renders no rookie designation anywhere on this product's set: https://www.tcdb.com/Gallery.cfm/sid/188426/2018-19-Hoops-Premium?PageIndex=25 renders "SN199" on tiles 241-250 and no "RC" on any of them, while https://www.tcdb.com/Gallery.cfm/sid/182809/2018-19-Hoops?PageIndex=25 renders "RC" on all ten of the same card numbers. So the renderer suppresses RC on a parallel gallery while still rendering the serial flag there - the same behaviour hoopspremium-bkb-2019-20 recorded in its G-008 - and this product's silence on RC carries no information. rookie_flag was therefore read off the PARENT for the identical card numbers, with full coverage of 1-300 and no range assumed: Checklist.cfm PageIndex 1 (rendered 1-78, no RC), PageIndex 2 (rendered 101-173, no RC) and PageIndex 3 (rendered 201-273, RC on 233 and 241-273), topped up tile by tile from Gallery.cfm pages 8 (71-80), 9 (81-90), 10 (91-100), 18 (171-180), 19 (181-190), 20 (191-200), 24 (231-240), 25, 26, 27, 28 (241-280), 29 (281-290) and 30 (291-300). No second source publishes RC designations for this checklist; the only independent corroboration is a retailer title, "Luka Doncic 2018-19 Panini Hoops Premium Box Set Rookie Card RC 022/199" against card 268 (https://www.steelcitycollectibles.com/i/luka-doncic-2018-19-panini-hoops-premium-box-set-rookie-card-rc-022-199), which confirms one card inside the block and nothing else.
• G-010 beckett.com is still unreachable, as it has been for every pass this project has recorded, and its failure mode on this URL is the 302. https://www.beckett.com/news/2018-19-nba-hoops/ returns HTTP 302 to https://maintenance.beckett.com/ and the body was never read. The page title indexed by search - "2018-19 Panini NBA Hoops Basketball Checklist, Team Sets, Release Date" - shows Beckett catalogues the PARENT, but nothing about this product was obtained from it and nothing from it was written. The usual third structural source is therefore absent, which matters for EXTRACTION_SPEC 5.4: see G-016 on whether this release's two sources are really two.
• G-011 A SET PRINT RUN THAT sets.csv HAS NO COLUMN FOR, recorded here in full. Premium Boxed Set Autographs is uniformly serial numbered to 10: tcdb's set-level note reads "Serial numbered to 10" and every one of its 20 rows renders the literal tokens "AU, SN10" (https://www.tcdb.com/ViewSet.cfm/sid/188437/2018-19-Hoops-Premium-Boxed-Set-Autographs ; https://www.tcdb.com/Inserts.cfm/sid/182809/2018-19-Hoops); cardboardconnection states "one Tribute autograph numbered to 10 copies" and Panini "one autograph from a current superstar or retired legend numbered to just 10". EXTRACTION_SPEC 2.2 forbids putting that figure in card_count, and it is not a parallel tier of a parent set name, so it has nowhere structured to live. card_count 20 is distinct cards.
• G-012 SET-NAME DIVERGENCE ACROSS SOURCES AND ACROSS THIS LINE'S OWN SEASONS, recorded because a resolver has to match on what sources print. For the autograph set this release writes tcdb's "Premium Boxed Set Autographs"; cardboardconnection heads the same checklist "Tribute Premium Box Set Autographs Checklist"; and the six later seasons of this line in data/ carry four further spellings - hoopspremium-bkb-2019-20 and -2020-21 "Premium Box Set Autographs", -2021-22 and -2022-23 "Hoops Tribute Premium Box Set Autographs", -2023-24 "Base Hoops Tribute Premium Box Set Autographs", -2024-25 "Base Hoops Tribute Premium Box Set Autograph". Five names for one recurring object across seven seasons. paninihoops-bkb-2018-19 already records the same divergence in its G-015. For the product itself tcdb shortens the name to "Premium" - see G-021. Nothing was renamed away from a source.
• G-013 MARKETPLACE CONTAMINATION INVESTIGATED AND NOTHING WRITTEN, AND THIS SEASON CARRIES A CONFUSION THAT THE LATER SEASONS DO NOT. (1) The known fabricated parallel names for this line - "Premium Box Set Silver Foil" and "PREMIUM STOCK FOIL /199" - were searched for against 2018-19 and returned no 2018-19 listing at all; every "Premium Stock" hit belongs to 2019-20 or later, i.e. to the different retail line. (2) A bare "FOIL" token in a listing title would be merely descriptive here, because the product's own copy calls the stock foil: Panini says the cards are "printed on premium holographic board with gold foil-highlights" and cardboardconnection says "the set is printed on holographic board with gold foil accents and is serial numbered out of 199". No parallel was written off any listing title. (3) NEW FOR THIS SEASON: the parent's own "Silver" base parallel is ALSO serial numbered to 199 - tcdb's Notes cell for sid 188432 reads "Serial numbered to 199" and checklistcenter's parallel list prints "Silver #/199" - so a marketplace title carrying only a card number and "/199" does NOT identify a Premium card. A real example of the ambiguity: an eBay title "2018-19 Panini NBA Hoops - Luka Doncic #268 Silver /199 (RC)" sits alongside "2018-19 Panini NBA Hoops Premium Box Set /199 Jrue Holiday #181". Only titles naming "Premium Box Set" explicitly were used as corroboration. Sources: https://www.tcdb.com/Inserts.cfm/sid/182809/2018-19-Hoops ; https://www.checklistcenter.com/2018-19-panini-nba-hoops-basketball-card-checklist/ ; https://blog.paniniamerica.net/panini-america-unveils-online-only-2018-19-nba-hoops-basketball-premium-box-set/ ; https://www.cardboardconnection.com/2018-19-panini-nba-hoops-premium-box-set
• G-014 SPEC DEFECT, MEASURED: tcdb's Errors.cfm IS A FALSE NEGATIVE ON THIS PARENT, AND EXTRACTION_SPEC 3.2 CALLS IT "the route to any variation set" WITHOUT QUALIFICATION. https://www.tcdb.com/Errors.cfm/sid/182809/2018-19-Hoops states "No errors or variations have been added for this set yet." - yet the same set's own renderings carry uncorrected-error tokens on at least two cards: https://www.tcdb.com/Gallery.cfm/sid/182809/2018-19-Hoops?PageIndex=26 renders card 260 Mo Bamba with the literal tokens "RC" and "UER" and the text "UER: Statline says Undrafted, was actually drafted 6th" (confirmed under a second URL with &sort=0), and https://www.tcdb.com/Checklist.cfm/sid/182809/2018-19-Hoops?PageIndex=1 reports card 67 Justin Jackson carrying "UER" with the note "Career stat line is that of Kosta Koufos". https://www.tcdb.com/Errors.cfm/sid/188426/2018-19-Hoops-Premium returns the identical "no errors" sentence for this product's own sid, so the intended positive control FAILED - the endpoint gave the same answer on a set that provably has UER rows and on one that appears not to, which means its silence is uninformative on both. THE CONSEQUENCE IS THE 2020-21 TRAP: that season of this line rendered "3b" flagged "UER" with no card 98, exactly the kind of thing Errors.cfm is supposed to surface. An agent relying on Errors.cfm alone to detect a misprinted base number would have missed it. The clean base run recorded in G-007 rests on the contiguous enumeration of all 300 numbers, not on Errors.cfm. [CORRECTED 2026-09-20: THE Errors.cfm SENTENCE QUOTED ABOVE IS NOT EVIDENCE AND THE CONCLUSION DRAWN FROM IT DOES NOT HOLD. Measured 2026-09-20 on tcdb sid 182809: Errors.cfm returns that identical sentence for a set whose own renderings carry UER on at least two cards (card 260 Mo Bamba, "UER: Statline says Undrafted, was actually drafted 6th", confirmed under &sort=0; card 67 Justin Jackson, "Career stat line is that of Kosta Koufos"). The endpoint returns the same string on a set that provably HAS errata and on one that appears not to, so it discriminates nothing and its silence is uninformative. WHAT SURVIVES: any claim here about base_set_size, about the base run being contiguous, or about no numbered block sitting past the base run, because those rest on the independent contiguous enumeration also cited in this note - the last rendered Gallery page, the empty page after it, and the per-row flag column. WHAT DOES NOT SURVIVE: the narrower claim that NO VARIATION SET EXISTS. A variation set renders as b-suffixed rows (101a/101b) that a base-run enumeration is not designed to surface, and Errors.cfm was the documented route to it, so that claim is now unsupported rather than disproved - it may still be true. Not re-fetched; recorded so it is not read as settled. See reference/caveats.csv CAV-020.]
• G-015 THE FETCH LAYER INVENTED TWO DIFFERENT GLOSSES FOR ONE FLAG TOKEN IN ONE PASS, confirming EXTRACTION_SPEC 3.4's warning with a fresh instance. Asked about the literal token "UER" on https://www.tcdb.com/Gallery.cfm/sid/182809/2018-19-Hoops?PageIndex=26 the layer volunteered "(Unofficial Error Report)"; asked about the same token on https://www.tcdb.com/Checklist.cfm/sid/182809/2018-19-Hoops?PageIndex=3 it volunteered "(Undesired Error Report)". Neither expansion appears on either page and they contradict each other. Only the literal string "UER" was recorded anywhere in this release, and no meaning was assigned to it. A third narrow re-ask that explicitly forbade explaining what tokens mean returned the token alone.
• G-016 MY TWO STRUCTURAL SOURCES MAY NOT BE INDEPENDENT, BUT THE LOAD-BEARING ONE IS FIRST-PARTY. tcdb and cardboardconnection agree on every value here (300 cards, #/199, a 20-card Hoops Tribute autograph checklist #/10 at 281-300), and cardboardconnection's parallel and autograph sections for the Hoops line track tcdb's set inventory closely enough that independence cannot be assumed - the same caution hoopspremium-bkb-2019-20 raised in its G-010. What is different this season is that blog.paniniamerica.net is the MANUFACTURER and is not derived from either: it independently states the 301/300 decomposition, the #/199, the #/10 autograph, the online-exclusive channel and the launch date. Marketplace corroboration is also genuinely independent: Steel City Collectibles and eBay seller titles match card numbers 17, 181 and 268 against this checklist. With Beckett unreachable (G-010) the usual third structural source is still gone. The single-sourced values in this release are: the $500 price (cardboardconnection only, and it is written in no field), and the ISO release date, for which cardboardconnection's "10/31/2018" and Panini's dateline are two sources but of the same day.
• G-017 EXTRACTION_SPEC 4.2 COPY CHECK FIRED TWICE, NEAR MAXIMUM BOTH TIMES, AND WAS INVESTIGATED RATHER THAN DELETED. (1) Base Set against data/paninihoops-bkb-2018-19/base_cards.csv: 299 of 300 positionally and 299 of 300 by card number. This is expected and stated by a source in words - "Inside each 2018-19 Panini NBA Hoops Premium Box Set is a 300-card checklist straight from 2018-19 NBA Hoops." - and is exactly what EXTRACTION_SPEC 5.2 describes. A complete overlap alone does not settle product versus parallel, because it is equally what a full base parallel looks like; what settles it is in the release notes and G-021. (2) Premium Boxed Set Autographs against the base Hoops Tribute block 281-300: 20 of 20, same players on the same numbers (281 Stephen Curry, 296 Kobe Bryant, 300 Allen Iverson). That is what the set is by construction - the signed printing of the Hoops Tribute subjects. Neither set is a renamed copy and neither was deleted. EXTRACTION_SPEC 4.5 SUSPICIOUS UNIFORMITY was also weighed: this line's seven seasons do share one formula (a 300 or 270 card base #/199 plus a 20-card Tribute autograph #/10), which is a template-fill smell, but here the formula is confirmed by a first-party launch post written in 2018 and the season's own naming differs from every later season's (tcdb "Premium" and "Premium Boxed Set Autographs" against later "Premium Box Set"), which is evidence against a template having been filled.
• G-018 RELEASE DATE IS SOURCED AND EXPLICITLY NOT AN ESTIMATE, WHICH BREAKS THIS LINE'S PATTERN AND IS WORTH STATING. cardboardconnection prints the label and value "Release Date: 10/31/2018" with no qualifier word, asked for twice under two distinct URLs with the qualifier specifically requested if present. Panini's launch post is datelined October 31, 2018 in both the byline and the body and reads "Company officials today launched the online-exclusive 2018-19 NBA Hoops Basketball Premium Box Set", and a narrow re-ask confirmed the word "Estimated" appears nowhere on that page. 2018-10-31 is therefore written. Three other seasons of this line (2019-20, 2023-24, 2024-25) wrote UNKNOWN, two of them because the only date any source stated was labelled "Estimated Release Date"; this season is not in that position. Note the date sits one week after the parent's 2018-10-24, which is itself part of the product-versus-parallel evidence.
• G-019 PRICE IS SINGLE-SOURCED AND IS WRITTEN IN NO FIELD. cardboardconnection: "The sets are sold online by Panini for $500 a box." (https://www.cardboardconnection.com/2018-19-panini-nba-hoops-premium-box-set). Panini's own post states no price, and its store page https://store.paniniamerica.net/2018-19-panini-hoops-premium-box-set-nba-trading-cards.html 302s to https://www.paniniamerica.net/2018-19-panini-hoops-premium-box-set-nba-trading-cards.html, which renders only "You need to enable Javascript to run this app." and no product content. release.csv has no price column in any case; the figure is recorded here so a later pass does not have to re-find it.
• G-020 NO PACK ODDS AND NO SCARCITY TIER ARE RECORDED FOR EITHER SET, and the four odds columns plus scarcity_tier are empty on both rows. This is a sealed complete factory box set with fixed contents rather than a pack product - cardboardconnection's box-contents list reads "Full base set (300 cards)" and "1 Hoops Tribute Autograph", and "Full set per box." - so per-pack odds do not exist in the ordinary sense, and no source states any odds figure. The nearest thing to an odds statement is the guarantee of one autograph per box, which is contents, not odds.
• G-021 STRUCTURAL DIVERGENCE BETWEEN THIS CATALOGUE AND tcdb, STATED RATHER THAN RESOLVED SILENTLY. tcdb does NOT give this product a parent sid of its own: https://www.tcdb.com/ViewAll.cfm/sp/Basketball/year/2018 lists only three 2018-19 Hoops-named products and none of them is the Premium Box Set, and the object is filed instead as one row among the parent's "Parallel Sets (17)" (sid 188426, name shortened to "Premium") with its autograph filed among the parent's "Insert Sets (47)" (sid 188437). So the single most structural source in this pass files this as a PARALLEL. It is catalogued here as a RELEASE because reference/brands.csv already carries the brand hoopspremium, because six later seasons of the line are built as releases in data/, and above all because the manufacturer describes it as a separately launched, separately sold, separately configured product with its own exclusive autograph set - the evidence set out in the release notes. tcdb files the 2019-20, 2020-21, 2021-22, 2022-23 and 2024-25 entries the same way, so this is the line's normal shape and not a 2018-19 anomaly. Recorded so that nobody reading tcdb alone concludes this release was invented.
• G-022 ENDPOINT BEHAVIOUR THAT CONTRADICTS EXTRACTION_SPEC 3.2, MEASURED ON THIS RELEASE. The spec says the two endpoints fail differently and that the difference is diagnostic: "Checklist.cfm?PageIndex=N past the last page renders no rows rather than erroring." On https://www.tcdb.com/Checklist.cfm/sid/188426/2018-19-Hoops-Premium?PageIndex=4 the page rendered no checklist rows but ALSO rendered tcdb's home-page boilerplate ("The archive of trading cards and all things trading card related. Checklists, galleries, forums, track your collection, trade, and more"), which is the documented signature of a NON-EXISTENT resource, not of an empty page. The two failure signatures the spec relies on to be distinguishable were not distinguishable here. The conclusion that nothing is numbered past 300 therefore does NOT rest on that fetch; it rests on Gallery.cfm reporting "Page 30 of 30", on tcdb's header "Total Cards: 300", and on the contiguous enumeration of 1-300.
• G-023 CONTRADICTION AGAINST AN EXISTING RELEASE'S CITED EVIDENCE, FOUND AND NOT REPAIRED (one agent, one release). data/paninihoops-bkb-2018-19/parallels.csv cites checklistcenter as the "second source" in the visual_notes of its "Premium" base-set parallel row (https://www.checklistcenter.com/2018-19-panini-nba-hoops-basketball-card-checklist/). That page does not support the row. Asked twice under two distinct URLs, and the second time as a narrow single-string question, it answers that the string "Premium" does not appear anywhere on the page, nor does "Box Set" or "boxed"; its parallel list prints in full as "Blue; Purple; Purple Winter; Red Back; Teal Explosion; Winter; Yello; Silver #/199; Green #/99; Blue Checkerboard #/75; Red #/49; Teal #/49; Artist Proof #/25; Orange #/25; Orange Explosion #/25; Red Checkerboard #/15; Gold #/10; Artist Proof Black 1/1" - eighteen names, none of them Premium. Its base-set line does corroborate the parent, "300 Cards - Including 20 Hoops Tribute Cards". EXTRACTION_SPEC 3.3 source calibration was run on the same page and it FAILED on questions that can be checked: it renders "Green #/99" where tcdb renders "Green Checkerboard" with "Serial numbered to 99", "Yello" for "Yellow", "Red Back" for "Red Backs", and it lists "Winter" and "Purple Winter" which tcdb files under a separate sid. checklistcenter was therefore used for NO value in this release. Either the page changed since 2026-09-17 or the citation was wrong when written; the parent was not edited either way.
• G-024 THIS BUILD CONTRADICTS specs/2018-19_SCOPE_MEMO.md AND THE TASK BRIEFING'S OPEN QUESTION, WITH EVIDENCE. The memo's section 1 table classifies "Panini NBA Hoops Premium Box Set", released 2018-10-31, as a "variant" that "reads as" an "identical 300-card checklist to Hoops, premium stock, #/199", and its following note says the two Premium Box Sets "share a base checklist with their parent product" and that "channel_taxonomy.csv may be the better home". The checklist claim is confirmed - 299 of 300 both positionally and by card number, see G-001 - but the classification is not: this is a separately launched product with its own release date, its own box configuration, its own price, its own exclusive sales channel and its own exclusive 20-card autograph set that appears in no pack of the parent, all stated by the manufacturer in a dated launch post. A shared checklist is not a sufficient condition for a parallel; it is equally what a box-set repackaging looks like, and channel_taxonomy cannot express an exclusive chase set. The memo's phrase "premium stock" is also unfortunate and was checked rather than inherited: the stock is described as "premium holographic board with gold foil-highlights", and the separate NBA Hoops Premium Stock line does not exist in 2018-19 (G-004). Recorded as a correction to the memo, which this build did not edit.

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