2017-18 Select

• G-001 ENUMERATION COMPLETENESS AND ITS LIMIT. The set inventory rests on ONE Inserts.cfm fetch of the parent (https://www.tcdb.com/Inserts.cfm/sid/166727/2017-18-Panini-Select), which returned 70 + 18 + 1 = 89 rows in a single response with no truncation. I counted the rows myself rather than trusting the section headers, and both sid blocks are contiguous with no gap to probe: 169082-169151 (70 sids) and 169706-169724 (19 sids, the Pink Prizms Marketing promo sitting inside the parallel run at 169716). All four boundaries were re-proved foreign by me with ViewSet.cfm: 169081 = '2002-04 Radio Shack ZipZaps Micro RC' (Total Cards: 75), 169152 = '2016 Panini Gala - Plates and Patches Printing Plates Black' (200), 169705 = '2017 Panini Contenders Optic - Rookie of the Year Contenders Red' (30), 169725 = '1997 Finish Line Phone Pak II' (100). LIMIT: this bounds what is hiding on tcdb under THIS parent and nothing more. Nobody checked the 2018-19 Select parent for a 2017-18 Update block (EXTRACTION_SPEC section 5.5 mirror), and CHASE_CHECKLIST_SPEC records sets filed under a SIBLING product's parent, which no enumeration of the correct parent can surface. NOTE: the 2017-18 scope memo section 7.1 describes sid 169705 as 'a 2017 Panini football set'; it is a 2017 Panini Contenders Optic set. Foreign to this release either way, but the memo's description of it is wrong.
• G-002 base_print_run IS 'unnumbered'/'assumed', NOT 'varies-by-tier', AND THE TASK PROMPT EXPECTED THE LATTER. The base run IS tiered and base_card_tiers.csv is written, but the three tiers differ in INSERTION RATE, not in serial numbering: cardboardconnection's own words are 'Base Concourse cards are most common, followed by short-printed Premier Level (2 per pack) and Courtside (2 per box).' (https://www.cardboardconnection.com/2017-18-panini-select-basketball-nba-cards). No source read in this build states a print run for any of the three tiers, and no serial token appears on any of the 300 base rows on either tcdb endpoint while the same flag column positively renders RC on 48 of them. Writing 'varies-by-tier' would assert three different print runs that nothing states. EXTRACTION_SPEC section 2.1 permits 'assumed' only alongside 'unnumbered', which is exactly this evidence shape. select-bkb-2018-19 and select-bkb-2019-20 are structurally identical (300 cards, same three tier names, same ranges) and both write base_print_run=unnumbered with base_print_run_basis=assumed and UNKNOWN tier print runs, so this build agrees with both siblings. OPEN: whether a source exists that states a per-tier print run or production quantity.
• G-003 TIER PRINT RUN AND 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. The evidence found supports the tier NAME, the RANGE and the COUNT only. The nearest thing to a per-tier distribution fact is cardboardconnection's '(2 per pack)' for Premier Level and '(2 per box)' for Courtside, which are insertion rates inside the same product, not trade channels and not print runs.
• G-004 PINK PRIZMS MARKETING HAS THREE DIFFERENT PRINT RUNS INSIDE ONE tcdb sid, AND THE SCHEMA CANNOT CARRY THEM. tcdb sid 169716 renders 'Total Cards: 300' and covers the whole base run, but its serial token changes with the base tier: 'SN10' on every rendered row of Checklist.cfm?PageIndex=1&sort=0 (rows 1-77, the Concourse block), 'SN4' on every rendered row of ?PageIndex=2 (rows 101-170, Premier Level) and 'SN1' on every rendered row of ?PageIndex=3 (rows 201-283, Courtside). A second, wholly independent source states the same three figures against the same three tiers: cardboardconnection prints 'Pink #/10' on its Base Concourse parallel line, 'Pink #/4' on its Base Premier Level parallel line and 'Pink 1/1' on its Base Courtside parallel line (https://www.cardboardconnection.com/2017-18-panini-select-basketball-nba-cards). So the run is 10 / 4 / 1 by tier on two independent sources, and the parallels.csv row nevertheless carries print_run=UNKNOWN because one row cannot hold three figures and splitting it into three block rows is blocked by the unique index: block rows have scope_set_id NULL unconditionally, so three rows named 'Pink Prizms Marketing' would collide, and neither source prints a longer per-tier name to disambiguate with. This is the EXTRACTION_SPEC section 2.3 expressiveness gap in its per-block form, the same shape caveats.csv CAV-036 records for chronicles-bkb-2017-18, now reproduced on Select. THE FIGURES ARE HERE SO THEY ARE NOT LOST: Concourse 1-100 = 10, Premier Level 101-200 = 4, Courtside 201-300 = 1.
• G-005 PINK PRIZMS MARKETING - THE TWO SOURCES DISAGREE ABOUT HOW IT REACHED COLLECTORS, AND THE CHANNEL VOCABULARY HAS NO VALUE FOR EITHER READING. tcdb files sid 169716 under a 'Promo Sets (1)' heading, separate from its 'Parallel Sets (18)' heading, with the set-level note 'Cards were distributed as promos at the 2018 National Convention' (https://www.tcdb.com/Inserts.cfm/sid/166727/2017-18-Panini-Select). cardboardconnection lists Pink among the ordinary pack-inserted PRIZM PARALLELS of all three base tiers, with no distribution note at all (https://www.cardboardconnection.com/2017-18-panini-select-basketball-nba-cards). A convention give-away is not hobby, retail, blaster, mega, h2, choice, fast-break, FOTL, multi, asia, international, china, lucky-envelopes, panini-direct, factory-set, fanatics-exclusive, FDI or lunar, so channel is UNKNOWN with an EMPTY channel_basis rather than forced into a near-miss value (EXTRACTION_SPEC section 2.3). The row is kept in parallels.csv rather than dropped, because both sources agree the cards exist and carry the base numbering.
• G-006 THREE AUTOGRAPH SETS HAVE A PER-CARD VARIABLE PRINT RUN THAT NOTHING IN THE SCHEMA CAN HOLD. Enumerated on Checklist.cfm?PageIndex=1&sort=0, each rendering every row of its set: Signatures (sid 169108, 28 of 28 rows) renders three distinct serial tokens - SN49, SN99 and SN149. In Flight Signatures (sid 169133, 25 of 25 rows) renders SN60, SN149 and SN199. Autographed Memorabilia (sid 169147, 31 of 31 rows) renders SN50, SN99 and SN149. sets.csv has no print_run column (EXTRACTION_SPEC section 2.2) and these are not parallel tiers, so the figures live here and in each set's notes. A modal token would have been a confident false value on all three. Rookie Signatures (SN199 on all 36) and Rookie Jersey Autographs (SN199 throughout) are by contrast uniform - so 2 of 5 serial-bearing insert/auto sets on this release are uniform and 3 of 5 are variable. PER-TOKEN COUNTS ARE NOT STABLE ACROSS FETCHES AND ARE NOT RECORDED AS FACT: this build read Signatures as SN49 x8 / SN99 x10 / SN149 x10 and In Flight Signatures as SN60 x9 / SN149 x8 / SN199 x8, while the 2017-18 Phase 0 probe read the same two sets as SN49 x8 / SN99 x12 / SN149 x8 and SN60 x7 / SN149 x7 / SN199 x11. Both readings sum to the set size and the DISTINCT token set is identical in both, which is the part the conclusion rests on. The two agree exactly on Autographed Memorabilia (SN99 x8, SN50 x4, SN149 x19).
• G-007 TWO SET COUNTS DISAGREE BETWEEN tcdb AND cardboardconnection, BOTH BY EXACTLY ONE, AND tcdb'S FIGURE WAS WRITTEN. Autographed Memorabilia: tcdb sid 169147 renders 'Total Cards: 31' and I counted 31 rendered rows; cardboardconnection heads the set 'Autographed Memorabilia Set Checklist - 32 cards'. In Flight Signatures: tcdb sid 169133 renders 'Total Cards: 25' and I counted 25 rendered rows; cardboardconnection heads the set 'In Flight Signatures Set Checklist - 24 cards'. The other twelve sets agree exactly between the two sources. tcdb's figures were written because they are the ones I enumerated row by row, but CHASE_CHECKLIST_SPEC warns that a tcdb header agreeing with its own rows only proves tcdb is internally consistent, so this is recorded rather than settled. No third source was reachable: checklistcenter was not consulted for this release in this build.
• G-008 TWO SETS HAVE A UNIFORM PRINT RUN OF THEIR OWN THAT sets.csv CANNOT CARRY. Rookie Signatures (sid 169119) renders 'AU' and the serial token 'SN199' on all 36 of its 36 rows, and Rookie Jersey Autographs (sid 169124) renders 'AU', 'MEM' and 'SN199' throughout its 34 rows. That is a property of the SET, not of a parallel tier, and EXTRACTION_SPEC section 2.2 forbids putting it in card_count and gives sets.csv no print_run column. It is recorded in each set's notes in words and here: Rookie Signatures #/199, Rookie Jersey Autographs #/199. Single-source (tcdb row-level tokens); cardboardconnection states print runs for these sets' PARALLELS but not for the sets themselves.
• G-009 PREFIX COLLISION ACROSS TWO PRODUCTS IN ONE SEASON. Select Swatches (sid 169113) codes its cards SS- - lowest 'SS-BIG Brandon Ingram', highest 'SS-ZRD Zach Randolph', read from https://www.tcdb.com/Checklist.cfm/sid/169113/2017-18-Panini-Select-Select-Swatches?PageIndex=1&sort=0 - and 2017-18 Panini Prizm uses SS- for Sensational Signatures in the same season. Any resolver keyed on the card-number prefix alone will cross-hit between the two products. (set_name, card_number) is the only safe key. No card-level rows are written by this build, so nothing here is wrong yet; this is a warning for the CHASE_CHECKLIST_SPEC pass that will write insert_cards.csv.
• G-010 ONE SET NAME DIFFERS BETWEEN THE TWO SOURCES. tcdb's sid 169098 heading is '2017-18 Panini Select - Sparks'; cardboardconnection heads the same set 'Sparks Relics Set Checklist - 10 cards'. Both agree it is 10 cards and both agree its parallels are Purple, Copper, Tie-Dye, Gold and Black. tcdb's form was written because it is what the sid's own heading prints and because the Inserts.cfm row that enumerates it prints 'Sparks'. Nothing was renamed; this records the alternative form so a resolver matching cardboardconnection's string can find it.
• G-011 THE TASK PROMPT ATTRIBUTES A QUOTE TO cardboardconnection THAT I COULD NOT REPRODUCE. The prompt states that cardboardconnection says Autographed Memorabilia runs '#/25 to #/199'. Two targeted verbatim asks against the live page - one for every sentence containing any of 'Concourse', 'Premier', 'Courtside', 'Prizm', 'parallel', '#/', 'numbered', 'Zebra', 'Scope', 'Silver', 'Tri-Color', 'Tie-Dye', and one for every sentence naming any of the fourteen insert or autograph sets - returned no such string (https://www.cardboardconnection.com/2017-18-panini-select-basketball-nba-cards). What the page does print for that set is 'PARALLEL CARDS: Purple #/65 or #/35, Tie-Dye #/25 or less, Gold #/10, Black 1/1.', which is about the set's PARALLELS and not about the set itself, and which contains a ceiling ('#/25 or less') and an unresolved alternative ('#/65 or #/35'). The underlying observation in the prompt still holds and is recorded in gaps.csv G-006: the base Autographed Memorabilia rows render SN50, SN99 and SN149, and no figure any source states for the family matches all three. The claim is not disproved - a summarising fetch's silence is weak evidence and the string may sit in page furniture I did not reach - but it is not written into any value in this release, and no side of the disagreement has been picked.
• G-012 COVERAGE OF THE SERIAL-TOKEN ENUMERATION ON THE BASE PARALLELS, STATED EXACTLY. Every one of the 19 base-run parallel sids was probed with one Checklist.cfm?PageIndex=N&sort=0 fetch per 100-card block it covers, so a per-block difference in print run would have been caught - and on Pink Prizms Marketing it WAS caught (G-004). Within a block, tcdb's fetch layer truncates, so between roughly 40 and 100 of each block's 100 rows were actually seen; the unseen rows sit at the top of each block's number range. Two parallels are enumerated 100 percent: White Prizms (all 100 rows, Checklist plus Gallery pages 9 and 10) and Copper Prizms and Purple Prizms Die Cut were additionally topped up with their last gallery page. The residual risk is a single-card exception inside a block that no fetch rendered. Closing it would cost roughly two more Gallery.cfm fetches per block, about 60 fetches. cardboardconnection independently states a flat figure per tier for every one of these parallels, and agrees with tcdb's token on all 14 tokens tcdb annotates, which is why the residual is recorded rather than treated as blocking.
• G-013 ROOKIE FLAGS ARE TAKEN FROM tcdb'S RC TOKEN AND tcdb IS NOT CONSISTENT ACROSS THE THREE TIERS. Forty-eight of the 300 base rows render 'RC' - 23 in Concourse, 20 in Premier Level and only 5 in Courtside. The flag column is live in both directions on every page used - each rendered page carries RC on some rows and not others - which is what licenses writing rookie_flag=no rather than UNKNOWN on the unflagged rows. But the same 2017 draftee can be flagged in one tier and not another: De'Aaron Fox is RC at 49 (Concourse) and 175 (Premier Level) and unflagged at 217 (Courtside); Lonzo Ball is RC at 28 and 161 and unflagged at 241; Lauri Markkanen is RC at 8 and unflagged at 233; Markelle Fultz is RC at 68 and 195 and unflagged at 275. Written as rendered. Whether the card itself carries an RC logo in the Courtside tier is unverified.
• G-014 NO VARIATION LAYER WAS LOOKED FOR AND ITS ABSENCE IS NOT ESTABLISHED. Errors.cfm was not fetched for this release. caveats.csv CAV-020 records that its silence is worthless anyway - it returns the same 'no errors or variations' sentence on a set that provably has UER rows - so nothing was lost by skipping it, but nothing was gained either. What IS established is narrower and rests on a contiguous enumeration: the 300 base numbers are contiguous 1-300 with none unissued and nothing is filed past card 300. A b-suffixed variation layer would carry DUPLICATE base numbers and would therefore not surface in that enumeration at all, and Gallery.cfm strips the b suffix, so this build cannot say whether one exists.
• G-015 SOURCE INDEPENDENCE, STATED PLAINLY. Every structural value in this release rests on tcdb, cardboardconnection, or both. checklistcenter was not consulted; a constructed URL was not tried because caveats.csv CAV-037 records that a 404 on a constructed checklistcenter path is evidence of nothing, and no search was spent on it. Beckett is unreachable for 2017-18. No marketplace listing was fetched for this release, because no plain integer was written onto an autograph set and no card number was written for any insert or autograph set at all. Where this file says 'two independent sources', it means tcdb and cardboardconnection, which the 2017-18 scope memo section 7.6 and CAV-037 find are NOT interchangeable and are plausibly one lineage descending from Panini's sell sheet. Treat every 'multi' in parallels.csv with that in mind. The one place they are genuinely independent in a checkable direction is the cardboardconnection page's per-tier parallel lines, which predict tcdb's roster sizes exactly on all four parallels whose Total Cards I measured - White 100, Scope 200, Silver 300, Zebra 300 - a prediction a copy of tcdb could not make.
• G-016 FOUR SET-SCOPE PARALLEL TIERS HAVE A PER-CARD VARIABLE PRINT RUN AND CARRY UNKNOWN/UNKNOWN. All 56 set-scope parallel sids were enumerated in full - each Checklist.cfm?PageIndex=1&sort=0 fetch rendered a row count equal to that page's own Total Cards - so these are complete enumerations, not samples. Four tiers disagree with themselves: Autographed Memorabilia Purple (sid 169150, 31 rows) renders THREE tokens, SN65, SN35 and SN43. Autographed Memorabilia Tie-Dye (sid 169151, 27 rows) renders FIVE, SN25 on most rows plus SN21, SN11, SN7 and SN4. In Flight Signatures Neon Green Prizms (sid 169136, 24 rows) renders SN65 and SN35. Signatures Neon Green Prizms (sid 169111, 29 rows) renders SN65 and SN35. The other 52 are uniform. TWO OF THE FOUR ARE ALSO A SOURCE DISAGREEMENT and are flagged value_contested=yes: cardboardconnection prints 'Purple #/65 or #/35' for Autographed Memorabilia Purple, which omits SN43 entirely, and 'Tie-Dye #/25 or less' for Autographed Memorabilia Tie-Dye, which is a CEILING and not a print run - EXTRACTION_SPEC section 2.3 says a ceiling is not a print run and this catalogue used to write the ceiling anyway, so a reader comparing this row against the older Contenders Ticket Stub 77, Season Ticket Stub 50 and Elite Red/Aspirations/Status 99 rows will find the same evidence shape carrying opposite values. THE OTHER TWO ARE NOT CONTESTED: cardboardconnection prints 'Neon Green #/65 or #/35' for both Signatures and In Flight Signatures, which is exactly the split tcdb renders, and prints an unhedged 'Neon Green #/65' for Rookie Signatures, whose 36 rows are indeed uniform SN65. That is cardboardconnection hedging precisely where the data is split and not hedging where it is not, on three sibling tiers - the strongest source-calibration result in this build.
• G-017 AUTOGRAPHED MEMORABILIA: tcdb'S BASE PAGE IS ONE CARD SHORT AND THE MISSING CARD IS NAMED. sets.csv records card_count=32 against tcdb's base-page header of 31. Evidence: sid 169147 lists 31 card codes (AM-ADR ... AM-ZZP) and its own Black parallel sid 169148 lists 32; diffed code by code the two are identical except that the Black page also carries 'AM-LMA LaMarcus Aldridge'. cardboardconnection independently heads the set 'Autographed Memorabilia Set Checklist - 32 cards' (https://www.cardboardconnection.com/2017-18-panini-select-basketball-nba-cards). This is the CHASE_CHECKLIST_SPEC parallel-page recovery case, and also its 'the most complete tier can be the scarcest one' case - Black is the 1/1 tier and is the only one of the four parallels that renders 32 (Gold 28, Purple 31, Tie-Dye 27). WHAT IS STILL OPEN: the base AM-LMA card's own serial number. The base page does not render the row at all, so its print run is unknown even though its Black parallel is SN1. No card-level rows are written by this build; the CHASE_CHECKLIST_SPEC pass should source AM-LMA from the Black page and note the provenance.
• G-018 PARALLEL ROSTER SIZES ARE NOT UNIFORM WITHIN A SET, AND parallels.csv HAS NOWHERE TO SAY SO. Measured Total Cards per tier: Autographed Memorabilia base 31 / Black 32 / Gold 28 / Purple 31 / Tie-Dye 27; Rookie Jersey Autographs base 34 / Black 34 / Gold 35 / Purple 33 / Tie-Dye 34; Throwback Memorabilia base 40 / Black 40 / Copper 40 / Gold 37 / Purple 40 / Tie-Dye 37; X-Factor Memorabilia base 30 / Black 30 / Copper 30 / Gold 28 / Purple 30 / Tie-Dye 28; In Flight Signatures base 25 / Black 23 / Gold 24 / Neon Green 24 / Tie-Dye 25; Signatures base 28 / Black 28 / Gold 28 / Neon Green 29 / Tie-Dye 28. cardboardconnection documents the mechanism for some of them in words - 'Purple #/99 (No Giles/Bell*)', 'Tie-Dye #/25 (No Leaf*)', 'Gold #/10 (No Giles/Bell*)', 'Black 1/1 (No Hollis-Jefferson*)' (https://www.cardboardconnection.com/2017-18-panini-select-basketball-nba-cards) - so certain players are omitted from certain tiers. parallels.csv has no card_count column, so these numbers are recorded here and in each row's visual_notes. Note the consequence for anyone sizing a tier: Rookie Jersey Autographs Gold renders 35 rows against a 34-card base, and Signatures Neon Green Prizms renders 29 against a 28-card base - a parallel can be LARGER than its base page as well as smaller.

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