2018-19 Select

• G-001 Release date 2019-03-06 rests on two sources, cardboardconnection ('Release Date: March 6, 2019', https://www.cardboardconnection.com/2018-19-panini-select-basketball-nba-cards) and tcdb's parent set page ('Release Date: March 6, 2019', https://www.tcdb.com/ViewSet.cfm/sid/194154/2018-19-Panini-Select). Neither labels itself an estimate, so section 3.2's estimated-date rule is not engaged. Beckett, the usual third structural source, was not reachable in this pass, and blowoutcards' First Buzz page returned an Incapsula challenge resource instead of page content, so no editorial third source was read.
• G-002 56 of the 74 parallel rows -- every insert- and autograph-scope row -- carry print_run=UNKNOWN and serial_numbered=UNKNOWN because no source read in this pass states a run for any of them. This is a genuine coverage hole, not an evidenced absence. The reason the usual cheap test is unavailable is specific and worth recording: tcdb's parent inserts page https://www.tcdb.com/Inserts.cfm/sid/194154/2018-19-Panini-Select does render a Notes column, but it carries an image-completion percentage on every one of its 88 rows across BOTH sections ('0.0% of images added', '19.2% of images added', '74.8% of images added'). There are no blank cells and a non-blank cell is not a serial statement, so the split that licenses reading blank as unnumbered does not exist on this parent in either section. This is the Contenders/Hoops-Winter shape of the six measured in section 3.2, and it is the shape on the Parallel Sets section as well as the Insert Sets section. Resolving these 56 rows needs one within-sid serial probe per sid; the sids are recorded in each row's visual_notes.
• G-003 PRINT RUN CONTRADICTION on the two Premier Level die-cut tiers, resolved in favour of the majority and recorded here. cardboardconnection (https://www.cardboardconnection.com/2018-19-panini-select-basketball-nba-cards) states 'Blue Die-Cut #/185' and 'Maroon Die-Cut #/135'. Written instead: Blue 249 and Maroon 175. Blue 249 is carried by tcdb's row-level token SN249 under two different renderers (https://www.tcdb.com/Checklist.cfm/sid/199443/2018-19-Panini-Select-Blue-Prizms-Die-Cut?PageIndex=1&sort=0 and https://www.tcdb.com/Gallery.cfm/sid/199443/2018-19-Panini-Select-Blue-Prizms-Die-Cut?PageIndex=10&sort=0), by PSA CardFacts' 'Blue (#/249)' (https://www.psacard.com/cardfacts/basketball-cards/2018-panini-select/33453), and by a physical serial in a marketplace title, '2018-19 Panini Select - Premier Level Patrick Mills #177 Blue Prizm /249' (https://www.ebay.de/itm/155284629482). Maroon 175 is carried by tcdb's SN175 and PSA's 'Maroon (#/175)'. SOURCE CALIBRATION, recorded because it is the reason for the call: cardboardconnection was tested on eight figures this build could check independently and passed seven (Light Blue 299, Red 199, White 149, Neon Green 75, Purple 99, Orange 65, Copper 60, Tie-Dye 25) and failed on Blue and Maroon, plus a coverage error on Green (G-004). PSA passed on every figure checkable against a tcdb row-level token. A reviewer wanting to reverse this should note that the eBay title is rendered in eBay-catalogue form rather than as free seller text, so it is weaker corroboration than a raw seller title would be.
• G-004 COVERAGE CONTRADICTION on Green Prizms. cardboardconnection lists Green under Premier Level and Courtside only and omits it from Concourse (https://www.cardboardconnection.com/2018-19-panini-select-basketball-nba-cards). Written instead: base-set scope, all three tiers. tcdb renders 'Total Cards: 300' with SN5 on every row (https://www.tcdb.com/Checklist.cfm/sid/199430/2018-19-Panini-Select-Green-Prizms?PageIndex=1&sort=0) and its gallery renders the Concourse card 1 tile itself, '1 - Stephen Curry SN5 - Golden State Warriors' (https://www.tcdb.com/Gallery.cfm/sid/199430/2018-19-Panini-Select-Green-Prizms?PageIndex=1&sort=0); PSA lists 'Green (#/5)' under all three tiers. Recorded rather than smoothed over.
• G-005 SCOPE UNRESOLVED on Gold Prizms, written as block scope 'Concourse; Premier Level' with value_contested=yes. tcdb positively enumerates exactly 200 cards, contiguous 1-200, SN10 throughout, and its last gallery page ends at '200 - Kyle Lowry SN10' with nothing above 200 (https://www.tcdb.com/Gallery.cfm/sid/199428/2018-19-Panini-Select-Gold-Prizms?PageIndex=20&sort=0). But PSA CardFacts (https://www.psacard.com/cardfacts/basketball-cards/2018-panini-select/33453) and cardboardconnection (https://www.cardboardconnection.com/2018-19-panini-select-basketball-nba-cards) BOTH list a Gold #/10 under Courtside as well. Those two sources are independent of each other -- they disagree on Blue, Maroon and Green -- so this is two positive assertions against one catalogue's non-holding, and absence from tcdb is not evidence of absence. The narrower, positively enumerated scope was written because it is the one backed card by card; if a Courtside Gold is ever rendered or listed, this row should become base-set scope. A targeted search for a marketplace listing of a Courtside Gold (card 201-300, /10) returned nothing in this pass. Note the consequence: under the written scope, Courtside has no Gold tier while Concourse and Premier Level do.
• G-006 CARD COUNT CONTESTED on Throwback Memorabilia: tcdb renders 'Total Cards: 38' (https://www.tcdb.com/ViewSet.cfm/sid/199496/2018-19-Panini-Select-Throwback-Memorabilia) while checklistcenter states '2018-19 Panini Select - Throwback Memorabilia Set - 40 Cards' (https://www.checklistcenter.com/2018-19-panini-select-basketball/). 38 was written because it accompanies an enumerated letter-coded checklist (TM-ACR ... TM-WMT), but section 3.2 records that tcdb's 'Total Cards' header has been seen counting four different things, so this is not settled. Twelve of the fourteen chase sets in this release agree exactly across these same two sources; only this one and Autographed Memorabilia (G-007) do not. Resolving it needs a row count off the rendered checklist rather than either header.
• G-007 CARD COUNT CONTESTED on Autographed Memorabilia: tcdb renders 'Total Cards: 26' (https://www.tcdb.com/ViewSet.cfm/sid/199422/2018-19-Panini-Select-Autographed-Memorabilia) while checklistcenter states '2018-19 Panini Select - Auto Memorabilia Set - 30 Cards' (https://www.checklistcenter.com/2018-19-panini-select-basketball/). 26 was written. Note the two sources also differ on the set's NAME ('Autographed Memorabilia' against 'Auto Memorabilia'); tcdb's form is used, and it is the form the parent inserts page and all four of the set's parallels use.
• G-008 SET-LEVEL PRINT RUNS THAT THE SCHEMA CANNOT HOLD. sets.csv has no print_run column (section 2.2) and must not carry a run in card_count, so these figures are recorded only in prose and here. Rookie Signatures is uniformly serial numbered: tcdb renders the literal token 'AU, SN199' on EVERY row (https://www.tcdb.com/ViewSet.cfm/sid/199470/2018-19-Panini-Select-Rookie-Signatures), a property of the whole 40-card set. Four further chase sets are PER-CARD VARIABLE and so have no single run at all: Autographed Memorabilia renders SN199, SN167, SN160, SN149, SN99 and SN69 within one set; In Flight Signatures renders SN199, SN99 and SN49; Signatures renders SN199, SN99 and SN49; Rookie Jersey Autographs renders SN199 and SN99. Section 2.3 records that the schema cannot express 'serial numbered, run varies', which is why these are held here rather than in a column.
• G-009 Zebra Prizms is called 'Zebra SP' by cardboardconnection (https://www.cardboardconnection.com/2018-19-panini-select-basketball-nba-cards) in all three tiers -- a short-print claim. No source read in this pass quantifies it and tcdb renders no serial token on any Zebra row, so scarcity_tier was left EMPTY rather than set to 'short-print' off a two-letter abbreviation. PSA lists the tier as plain 'Zebra' with no SP and no figure (https://www.psacard.com/cardfacts/basketball-cards/2018-panini-select/33453). The same caution applies to every other row in this release: no stated pack odds were found for any set or parallel, so stated_odds, odds_evidence_url and odds_evidence_quote are empty throughout both parallels.csv and sets.csv.
• G-010 SPEC DEFECT MEASURED THIS PASS, opposite in direction to the one section 3.2 records. The spec says ViewSet.cfm and Checklist.cfm do not render the flag column the same way, and instructs that every row-level serial probe be run on Checklist.cfm because ViewSet was seen dropping a flag. On this release Checklist.cfm is the endpoint that dropped it: https://www.tcdb.com/Checklist.cfm/sid/199440/2018-19-Panini-Select-White-Prizms?PageIndex=1&sort=0 returned an EMPTY flag column on every row of White Prizms, while Gallery.cfm on the same sid renders 'SN149' in every tile title (https://www.tcdb.com/Gallery.cfm/sid/199440/2018-19-Panini-Select-White-Prizms?PageIndex=1&sort=0) and both PSA and cardboardconnection state #/149. Had the Checklist rendering been trusted as the spec directs, a /149 tier would have been recorded as unnumbered -- and it would have passed the discriminative test, because sibling sets in the same renderer DID show tokens. Recommendation: the rule should be that a NEGATIVE flag result on any single endpoint must be confirmed on a second endpoint before it is written, in both directions. Three endpoints disagreeing pairwise is now measured on this project.
• G-011 SUSPICIOUS-UNIFORMITY CHECK RUN AND CLEARED, recorded because the signal was strong. This release's parallel ladder is IDENTICAL to the one already in data/select-bkb-2019-20: Light Blue 299, Red 199, White 149, Neon Green 75, Tie-Dye 25, Neon Orange Pulsar 13, Gold 10, Green 5, Black 1, and on the die-cuts Blue 249, Maroon 175, Purple 99, Orange 65. Section 4 check 5 treats identical structure across releases as a template-fill smell, and a plausible failure mode here is a catalogue back-filling 2018-19 from 2019-20. It was investigated rather than deleted and it is genuine: PSA CardFacts states the figures for 2018-19 specifically and independently of tcdb, and two marketplace listings carry physical serials on real 2018-19 cards -- '#177 ... Blue Prizm /249' and '#172 Maroon Prizm /175'. The one source that does NOT match 2019-20 is cardboardconnection, and it is the source that is wrong (see G-003). Panini reused the ladder between the two seasons.
• G-012 TRANSCRIPTION CONTROL AND ITS ONE DISAGREEMENT. All 300 base rows were read from the parent sid, and the three Checklist.cfm pages were each re-anchored against a named player to catch the fetch layer's documented constant-shift renumbering: page 1 returned Deandre Ayton at 2 and Luka Doncic at 25, page 2 returned Aaron Gordon at 101 and Doncic at 122, page 3 returned Shai Gilgeous-Alexander at 201 and Doncic at 229. Each anchor matches the same card number on an independently fetched parallel sid (Light Blue card 1, Blue Die Cut card 101 'Aaron Gordon', Copper card 201 'Shai Gilgeous-Alexander'), and the three Checklist/Gallery overlaps (71-79, 171-172, 271-276) agree row for row, so no shift occurred. ONE DISAGREEMENT, resolved: a summarising fetch of https://www.tcdb.com/ViewSet.cfm/sid/199443/2018-19-Panini-Select-Blue-Prizms-Die-Cut reported card 179 as 'Tyreke Evans (Sacramento Kings)', while the parent gallery renders '179 - Tyreke Evans - Indiana Pacers' (https://www.tcdb.com/Gallery.cfm/sid/194154/2018-19-Panini-Select?PageIndex=18&sort=0). Indiana Pacers is written; Evans played for Indiana in 2018-19 and for Sacramento the prior season, so the summary appears to have supplied a team from outside the page. A further instance of section 3.4's fabricated-detail hazard.
• G-013 CONVENTION CHECK against the adjacent season of this same line, per the closing instruction of section 6. data/select-bkb-2019-20 sets base_set_size=300 and files Concourse/Premier Level/Courtside as three base_card_tiers rows with base_print_run=unnumbered/assumed, which is exactly the shape written here, so the two seasons of Select agree and no divergence needs repair. Section 5.3 had nothing to exclude in either season. ONE DELIBERATE DIFFERENCE: the 2019-20 build files its per-tier colour parallels as scope_type=base-set with the tier name folded into parallel_name ('Concourse Silver Prizm', 'Premier Level Blue Die-Cut Prizm'), whereas this build files a parallel that covers only part of the base run as scope_type=block with the tier named in `scope`, which is what section 2.3 defines block for. The effect is that 2019-20 reports 33 base-set-scope base parallels and 2018-19 reports 6 base-set-scope plus 12 block-scope, for products of the same shape. A consumer comparing the two seasons will see different scope_types for the same kind of object. This build follows the current spec; the 2019-20 rows are the ones that would need a repair pass, and nothing was touched there.
• G-014 checklistcenter lists five further 'sets' that are NOT sets and were filed as parallels or not at all: 'Select Swatches Tie-Dye Set - 8 Cards', 'Select Swatches Gold Set - 9 Cards', 'X-Factor Memorabilia Tie-Dye Set - 12 Cards', 'X-Factor Memorabilia Gold Set - 9 Cards' and 'X-Factor Memorabilia Black Set - 23 Cards' (https://www.checklistcenter.com/2018-19-panini-select-basketball/). Each is a colour parallel of a set already in sets.csv, and its stated count is that parallel's own roster rather than a distinct-card count -- section 5.4 unit 4 exactly. They are carried as parallels.csv rows scoped to Swatches and X-Factor Memorabilia. The counts are NOT written anywhere, because parallels.csv has no roster-size column; they are recorded here so a later pass can see that a partial roster is known for four of the 56 unresolved insert parallels.
• G-015 SINGLE-SOURCE INVENTORY. The set inventory itself -- which sets exist and which parallels belong to which set -- rests on ONE source, tcdb's parent inserts page https://www.tcdb.com/Inserts.cfm/sid/194154/2018-19-Panini-Select. checklistcenter corroborates the fourteen chase set NAMES and twelve of their counts, but it does not list the 56 insert-scope parallels at all, and PSA covers only base tiers. Section 3.2 records that Inserts.cfm can omit real sets entirely, so a set missing from that page would be invisible to this build. Errors.cfm was NOT used as a completeness check in either direction, because it is measured to return its no-errors sentence on sets that provably carry UER rows and therefore discriminates nothing; the base run's cleanliness here rests on the contiguous 1-300 enumeration instead.

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