2016-17 Excalibur

• G-001 base_print_run=unnumbered carries basis=assumed, NOT sourced, and the distinction is the whole point. NO source consulted states in words whether 2016-17 Panini Excalibur base cards carry a serial number; cardboardconnection's product page says nothing either way. What supports the value is a WITHIN-PRODUCT POSITIVE DISCRIMINATIVE TEST: the same two renderers (Checklist.cfm and Gallery.cfm on this product's own sids) print no token on any of the parent's 200 base rows while printing SN49 on all 200 rows of Duke (sid 154260), SN199 on Marquis (154258), SN149 on Prince (154259), SN10 on King (154261) and SN1 on Emperor (154255). Per EXTRACTION_SPEC 2.1 that is good evidence and still not a source, so the basis is assumed. RESOLVE BY: finding a Panini sell sheet, a blog.paniniamerica.net first-look post, or a marketplace listing for a base card that states the numbering in words.
• G-002 release_date=2017-07-05 IS SINGLE-SOURCE AND THE SECOND SOURCE IS EXPLICITLY AN ESTIMATE. tcdb's parent page renders the unqualified line "Release Date: July 5, 2017" with the word Estimated appearing nowhere ( https://www.tcdb.com/ViewSet.cfm/sid/149101/2016-17-Panini-Excalibur ). cardboardconnection's PRODUCT page ( https://www.cardboardconnection.com/2016-17-panini-excalibur-basketball ) renders "Estimated Release Date: 7/7/2017". Per EXTRACTION_SPEC 3.2 a source that labels its own date an estimate has told you it does not know, so the two-day difference is NOT recorded as value_contested and cc is NOT recorded as corroboration. NOTE FOR THE SCOPE MEMO: memo section 1c lists Excalibur under "smaller divergences" as "tcdb Jul 5 / cc index Jul 07" and warns the cc INDEX date carries a health warning. I opened the cc PRODUCT page as instructed and it gives the same 7/7 - so for this product the cc index and the cc product page AGREE, unlike Prestige where they differed by a month and three days. That is a new datapoint about cc, and it does not rescue the figure, because the product page labels it an estimate. RESOLVE BY: a dated first-party or retailer announcement.
• G-003 FOUR SET-SCOPE TIERS ARE PROVABLY SERIAL NUMBERED WITH A PER-CARD VARIABLE RUN AND THE SCHEMA CANNOT EXPRESS IT, so all four carry print_run=UNKNOWN / serial_numbered=UNKNOWN and the real values are here. This is EXTRACTION_SPEC 2.3 and CAV-041's expressiveness gap, now recurring on a sixth line. Each was enumerated over EVERY row of its sid, never sampled, and each was confirmed on a second differently-varied endpoint with a direct named-player question. THE CENSUS: (1) Apprentice Shield Prime, sid 154264, Total Cards 30, all 30 rows read: MEM+SN10 on 29 and MEM+SN3 on Jamal Murray, card 6. (2) Apprentice Signature Shield Prime, sid 154274, Total Cards 32, all 32 rows read: AU+MEM+SN10 on 30 and AU+MEM+SN5 on Buddy Hield, card 5, and Jakob Poeltl, card 8. (3) Armory Prime, sid 154276, Total Cards 34, all 34 rows read: MEM+SN10 on 32 and MEM+SN5 on Patrick Beverley, card 11, and Jamal Crawford, card 31. (4) Knights Cloak Prime, sid 154355, Total Cards 20, all 20 rows read: MEM+SN25 on 19 and MEM+SN10 on Karl-Anthony Towns, card 15. A modal token was NOT written in any of the four. All four are SET scope, where UNKNOWN is legal; had any been base scope, CAV-041 says there would have been no legal row at all and the tier would have had to be deleted. RESOLVE BY: a print_run_basis column on parallels, or a sibling table holding per-card runs.
• G-004 THE BASE RUN IS STRUCTURALLY TWO-PART AND NO SOURCE NAMES THE PARTS, SO card_tier IS EMPTY ON ALL 200 ROWS AND NO base_card_tiers.csv IS WRITTEN. Cards 1-180 are then-current players grouped by franchise; cards 181-200 are retired legends (181 Alex English, 182 Allen Iverson, 183 Artis Gilmore, 184 Shaquille O'Neal, 185 Grant Hill, 186 Scottie Pippen, 187 David Robinson, 188 Dave Cowens, 189 George Gervin, 190 Hakeem Olajuwon, 191 John Havlicek, 192 Jerry Lucas, 193 Lenny Wilkens, 194 John Stockton, 195 Wilt Chamberlain, 196 Patrick Ewing, 197 Dominique Wilkins, 198 Karl Malone, 199 Gary Payton, 200 Charles Oakley), read off Gallery.cfm pages 19 and 20. Neither tcdb nor cardboardconnection ( https://www.cardboardconnection.com/2016-17-panini-excalibur-basketball ) gives that block a name - I asked cc directly and it carries none - so writing a tier label would be invention, and per CAV-043 a card_tier string that does not match a tier_name loads tier_id NULL and gates green, which makes a guess here invisible rather than merely wrong. The 20 legends ARE all base cards: they render no AU/MEM flag and sit inside the parent's own 200. RESOLVE BY: a source naming the subset, e.g. a Panini sell sheet or a first-look post.
• G-005 FOUR SET COUNTS WHERE SOURCES DISAGREE, WITH WHAT I WROTE AND WHY. sets.csv has no value_contested column, so the disagreements live here. (a) Apprentice Signature Shield - tcdb header 30, tcdb RENDERED ROWS 31 (my own count), cardboardconnection 32. THREE different figures. I wrote 31, the only one I verified myself by counting rendered rows. Unresolved which unit the 30 and the 32 are counting. (b) Apprentice Signatures - base page states and renders 26; its own Holo Gold parallel (154272) states and renders 27 including #3 Kris Dunn; cc says 27. I wrote 27. (c) Calligraphy - base page states and renders 24; Holo Gold (154329) and Gunmetal (154330) each state and render 25 including CAL-DB Devin Booker; cc says 25. I wrote 25. (d) Manuscripts - base page states and renders 35; Holo Gold (154358) and Gunmetal (154357) each state and render 36 including M-KP Kristaps Porzingis; cc says 36. I wrote 36. ⚠ (b), (c) and (d) are the SAME DEFECT THREE TIMES on one release: tcdb's base page for an autograph set is exactly ONE ROW SHORT of its own parallels and of cc, and in each case the missing player is identifiable by name. That is CHASE_CHECKLIST_SPEC 1d (the parallel page recovering a card missing from the base page) reproducing three times in one product, which is worth recording as a pattern rather than three coincidences. RESOLVE BY: a third independent roster per set. ⚠ (a) IS THE WEAKEST FIGURE IN THIS RELEASE AND THE PARALLEL-PAGE TEST DOES NOT RESCUE IT, BECAUSE THE TWO PAGES ARE NON-NESTED. I enumerated both rosters by card number. Apprentice Signature Shield (sid 154273) renders 31 rows: 1, 2, 3, 5, 7-23, 25-31, 33-35. Apprentice Signature Shield Prime (sid 154274, Total Cards 32) renders 32 rows: 1-5, 7-18, 20-23, 25-35. So the base page HAS 19 (Malachi Richardson) which Prime lacks, while Prime HAS 4 (Kris Dunn) and 32 (Tyler Ulis) which the base page lacks. NEITHER PAGE IS A SUPERSET OF THE OTHER and their union is 33 distinct numbers - a fourth figure. That is CHASE_CHECKLIST_SPEC 1c-septies (tiers are not nested), so the 1d recovery logic I used on Apprentice Signatures, Calligraphy and Manuscripts - where the parallel page was a strict superset by exactly one named player - CANNOT be applied here, and cardboardconnection's 32 is as plausible as my 31. I wrote 31 because it is the only figure I verified by counting rendered rows myself, not because it is better evidenced. RESOLVE THIS ONE FIRST.
• G-006 CRUSADE CAMO IS A tcdb-ONLY TIER AND ITS PRINT RUN RESTS ON ONE LINEAGE. tcdb's Inserts.cfm lists EIGHT Crusade sids (Black, Blue, Camo, Gold, Orange, Purple, Red, Silver) while cardboardconnection's Crusade sentence names SEVEN: "Choices include base Silver along with Blue (#/149), Red (#/99), Purple (#/49), Orange (#/25), Gold (#/10), and Black (1/1) versions." - Camo appears nowhere on cc. I wrote Camo as unnumbered off a full 100-row enumeration on two endpoints under two sorts plus a within-set positive discriminative test, so the value is evidenced, but cc's silence is not corroboration of it and this row is source_confidence=single. ⚠ A Camo tier is often a retail-channel tier on Panini products, which would make it a channel question too; nothing sourced says so here and channel is UNKNOWN. RESOLVE BY: a second lineage that carries a Camo tier for this product.
• G-007 SEVEN SETS HAVE A UNIFORM PRINT RUN OF THEIR OWN AND sets.csv HAS NOWHERE TO PUT IT (EXTRACTION_SPEC 2.2: there is no print_run column here and card_count must not be used for one). The figures, each from a full enumeration of every row of the set's own sid: Apprentice Shield /149 (MEM+SN149 on all 35 of 35 rows, sid 154263); Apprentice Signatures /199 (AU+SN199 on all 26 rendered rows, sid 154267); Armory /99 (MEM+SN99 on all 38 of 38, sid 154275); Calligraphy /149 (AU+SN149 on all 24 rendered rows, sid 154328); Emblem /99 (MEM+SN99 on all 25 of 25, sid 154342); Manuscripts /149 (AU+SN149 on all 35 rendered rows, sid 154356); Team USA /99 (MEM+SN99 on all 5 of 5 tiles, Gallery.cfm, sid 154371 - see G-012, Checklist.cfm dropped it). Three of these are independently corroborated in words by cardboardconnection ( https://www.cardboardconnection.com/2016-17-panini-excalibur-basketball ): "Armory Relics - 38 cards, #/99", "Emblem Relics - 25 cards, #/99" and "Team USA Relics ... 5 cards. Serial numbered #/99." NONE of the seven was written into a parallels.csv row, because a run that applies to the whole set is not a parallel tier. RESOLVE BY: a set-level print_run column, or a documented convention for this shape.
• G-008 NO MARKETPLACE CORROBORATION WAS ATTEMPTED ON THIS RELEASE, so no print run, count or card number here has been checked against a card in hand. EXTRACTION_SPEC 3.4 requires marketplace corroboration before writing plain integers onto an autograph set; this build wrote no card numbers for any autograph set (insert_cards.csv is a separate pass and was not touched), so the requirement is not breached - but it does mean every serial figure in this release rests on tcdb's flag column plus, on some tiers, cardboardconnection. 2016-17 is the Simmons/Ingram/Hield/Brown/Murray rookie year so the raw material exists. The scope memo section 9 records that marketplace corroboration was never tested on this season at all, so there is still no per-season verdict either way. RESOLVE BY: goldin.co lot titles or free-text eBay seller titles on a Duke, a Crusade Blue and an Armory Prime.
• G-009 parallel_family IS DERIVED AND I WROTE WHAT THE DERIVATION RETURNS, INCLUDING SIX VALUES THAT ARE WRONG. The values were produced by AST-lifting FAMILIES and family_of out of pipeline/fix_schema.py so the repo keeps one definition; NOTHING was added to FAMILIES (CAV-039). TWO NEW ORDERED-SUBSTRING DEFECTS, both of the Pink/Ink, Cosmos/Cosmo and Metallized/Meta class, now the fourth and fifth instances: (1) "Apprentice" CONTAINS "ice", so family_of returns Ice for Apprentice Shield Prime, Apprentice Signature Shield Prime and Apprentice Signatures Holo Gold. The third of those is doubly wrong - Ice sits earlier in the ordered list than Holo, so the bug also OVERRIDES a match that would have been right. (2) "Gunmetal" CONTAINS "Meta", so family_of returns Meta for Calligraphy Gunmetal, Manuscripts Gunmetal and Signature Knights Gunmetal. Meta is a real Spectra finish on about 50 corpus rows, so this silently merges three Excalibur rows into it. Note the existing guard does not help: ('Metallized','other') was added before ('Meta','Meta') for exactly this bug and Gunmetal was simply the next word nobody thought of. I did NOT repair either, per the brief and CAV-039, and per CAV-019 the corpus-wide dry-run that caveat demands has not been run. THE THEMED LADDER: the nine base-set peerage tiers (Baron, Count, Viscount, Marquis, Duke, Prince, King, Emperor, Lord) all fall through to "other" - FAMILIES has no peerage pattern, which is the expected outcome and is not a defect. FAMILIES DOES carry Knight, Crusade, Camo and Storm, so Knight in Shining Armor Blue/Purple and Knights Cloak Prime land on Knight, and the Crusade ladder lands on Crusade - EXCEPT Crusade Camo, which lands on Camo because Camo precedes Crusade in the list, splitting one visual family across two values. Storm the Castle Blue/Purple land on Storm. RESOLVE BY: an owner decision on the ordered-substring list, with the corpus-wide dry-run CAV-019 requires.
• G-010 NO FLOOR IS CLAIMED FOR THE EXCALIBUR LINE AND NONE MAY BE INFERRED FROM THIS RELEASE. This is the catalogue's first Excalibur release and the line is OLDER than 2016-17: specs/2016-17_SCOPE_MEMO.md section 3a records Excalibur present in tcdb's 2015-16 index at sid 126959 and its 2014-15 index at sid 114825, so at least two seasons sit below this one and the depth beneath THEM is unmeasured. Section 8a decision 2 admitted the line while stating that the admission opens a backfill of unmeasured depth. Depth was NOT chased, as instructed. ⚠ AND THE STUDIO LESSON CUTS THE OTHER WAY TOO (memo 3b): absence from an adjacent index would not have been a floor either, so nothing short of a positive statement in words could establish one. I found no such statement. INCIDENTAL DEPTH FINDING, recorded and not chased: cardboardconnection's Kaboom! copy reads "Carrying over the top option from the 2015-16 product", which is an independent first-hand confirmation from a second lineage that a 2015-16 Panini Excalibur product exists - the memo's evidence for that was a tcdb index row alone. It still says nothing about how deep the line runs. RESOLVE BY: researching the line's own history once, as a separate task.
• G-011 THE SID BLOCK'S INTERIOR GAPS ARE ONLY PARTLY PROBED. BOTH BOUNDARIES ARE PROVEN: 154252 is "2017 Bowman Platinum", Total Cards 100, BASEBALL 2017 (foreign, and it is also the probe required by CHASE_CHECKLIST_SPEC below the bottom of the listing); 154373 is "2016-17 Panini Preferred", Total Cards 298, Basketball 2016-17 - a DIFFERENT 2016-17 NBA PRODUCT IMMEDIATELY ABOVE THE BLOCK TOP, exactly the hazard the brief names, and its figures match scope memo section 1b. Excalibur's own 62 subset sids run 154253-154261 (9 Parallel Sets) and 154263-154372 (53 Insert Sets), so 58 sids inside that span belong to something else. PROBED AND FOREIGN: 154262 "2015 Grandstand Erie SeaWolves" (baseball), 154265 "2017 Bowman Platinum - Ice" (baseball), 154300 "2017 Bowman Platinum - Bowman's Next Generation Green" (baseball) - so the 48-sid gap 154277-154324 is a Bowman Platinum block interleaved INSIDE the Excalibur range. PROBED AND NON-EXISTENT: 154348 renders the tcdb HOME PAGE ("Cards Listed: 35,440,084", Random Card of the Day), the documented signature, reproduced on this season. NOT PROBED: 154266, 154268-154271, 154277-154299, 154301-154324, 154347, 154349. Nothing is known to be hiding there - the Inserts.cfm section headings reconcile exactly with my own counts ("Insert Sets (53)" against 53, "Parallel Sets (9)" against 9) - but CHASE_CHECKLIST_SPEC makes gap probing mandatory and 58 of them are unprobed. RESOLVE BY: ViewSet.cfm probes on the remaining sids.
• G-012 CAV-042 FIRED THREE TIMES ON THIS RELEASE, IN THREE DIFFERENT SHAPES, AND EACH WOULD HAVE WRITTEN A FALSE VALUE. Recorded because two of the three shapes are not in the caveat as written. (1) SAME Gallery.cfm URL, sort varied: Gallery.cfm/sid/154335/...-Crusade-Blue?PageIndex=8 under the default sort reported the flag column EMPTY on all 10 tiles and then SELF-CONTRADICTED inside the same response ("SN149 present in all cards... though the actual card tile display shows EMPTY"); the identical URL with &sort=0 returned SN149 on all ten. This is the 2026-09-30 extension of CAV-042 reproducing on a new line. (2) NEW SHAPE - SAME Checklist.cfm URL, sort varied, and the loss spared only the FIRST row: Checklist.cfm/sid/154334/...-Crusade-Black under the default sort rendered SN1 on card 1 and an EMPTY flag column on cards 2-79 - 78 of 79 rows lost their token. Read at face value that is a per-card variable tier and would have been written UNKNOWN/UNKNOWN: a false 'we did not resolve this' on a tier that tcdb's own Inserts.cfm Notes calls "Serial numbered to 1" and cardboardconnection calls "Black (1/1)". The same URL with &sort=0 returned SN1 on all 70 rows it rendered. (3) NEW SHAPE - ENDPOINT varied, and the loss took the MEM flag as well as the serial token, across a whole set FAMILY: Checklist.cfm?sort=0 on both Team USA (154371) and Team USA Prime (154372) renders the literal token "USA" and nothing else on all 5 rows of each, while Gallery.cfm?PageIndex=1 on the same two sids renders "MEM, SN99" and "MEM, SN10" respectively, and cardboardconnection independently calls the family "Team USA Relics ... Serial numbered #/99". Writing the Checklist reading would have put unnumbered/no on a provably /99 relic set and a provably /10 tier. NOTHING WAS WRITTEN FROM ANY OF THE THREE DISCARDED READINGS. Also measured, in the opposite direction: Gallery.cfm dropped the MEM flag on Knights Cloak rows 1-10 that Checklist.cfm renders - so the loss runs both ways on this release, as the caveat says. RESOLVE BY: nothing to resolve; this is a record for the caveat file.
• G-013 Errors.cfm ON THE PARENT RETURNED ITS USELESS SILENCE AND I CONCLUDED NOTHING FROM IT (CAV-020). https://www.tcdb.com/Errors.cfm/sid/149101/2016-17-Panini-Excalibur renders the heading "Errors / Variations" and the sentence "No errors or variations have been added for this set yet." That endpoint returns the identical sentence on sets that provably DO have errata, so it discriminates nothing. I am recording the rows it returned (none) as evidence of nothing, and the claim that this release has no variation set is NOT made. What does survive independently: base_set_size 200, the base run's contiguity, and the absence of any numbered block past the base run, all of which rest on a contiguous enumeration of 200 distinct numbers across two endpoints, not on this endpoint. Note that a b-suffixed variation layer would render as duplicate base numbers that my enumeration found none of - 200 rows over 200 distinct numbers - which is evidence, but Gallery.cfm strips b suffixes, so it is not conclusive. RESOLVE BY: a Checklist.cfm pass looking specifically for b/c suffixes.
• G-014 CHANNEL IS UNKNOWN ON 42 OF 44 PARALLEL ROWS AND THE PRODUCT-LEVEL HOBBY FACT WAS DELIBERATELY NOT INHERITED ONTO THEM. cardboardconnection ( https://www.cardboardconnection.com/2016-17-panini-excalibur-basketball ) states "Hobby boxes include two autographs and one relic" and gives the configuration "8 cards per pack, 20 packs per box, 16 boxes per case" - so a hobby configuration provably exists for this product. That is a statement about BOXES, not about which tier reaches collectors through which channel, and EXTRACTION_SPEC 2.3 is explicit that declining to inherit a product-level hobby note onto tiers it does not cover is the right call: channel=UNKNOWN, channel_basis EMPTY, reasoning here. THE TWO EXCEPTIONS: cc's PARALLEL CARDS line reads "PARALLEL CARDS: Baron, Count, Emperor, Lord (Retail), Viscount (Retail), Marquis #/199, Prince #/149, Duke #/49, King #/10." - it marks Lord and Viscount, and only those two, "(Retail)". Those two rows carry channel=retail with channel_basis=sourced-presence, not sourced-exclusive, because the source PLACES them in retail without using the word exclusive, and per CAV-010 a single named channel value already asserts exclusivity - a stronger claim than the evidence. channel_confidence=single. ⚠ Note the consequence: a retail-marked tier plus a hobby box configuration means this product had both, and no source consulted says which of the other seven base tiers sat in which. RESOLVE BY: a Panini sell sheet or retailer listing naming the configurations per tier.
• G-015 insert_cards.csv WAS NOT WRITTEN - card-level insert and autograph checklists are a separate pass under CHASE_CHECKLIST_SPEC and EXTRACTION_SPEC says do not write that file. What this build learned that the chase pass should start from, so it is not rediscovered: TWO SETS ARE LETTER-CODED and the rest appear plain-numbered. Calligraphy uses the prefix CAL- (CAL-AI Allen Iverson, CAL-ZL Zach LaVine) and Manuscripts uses M- (M-AC Antoine Carr, M-ZR Zach Randolph) - both read off tcdb's own Checklist.cfm rows, and both prefixes are real, not aggregator renumbering, because the parallel pages print the identical codes. TWO GENUINE CODE COLLISIONS, both expected per the spec and both to be written verbatim and flagged rather than dropped: CAL-DS sits on Dennis Scott AND Damon Stoudamire, and M-DW sits on Dwyane Wade AND Dominique Wilkins. FOUR SETS HAVE NON-CONTIGUOUS CARD NUMBERS - Apprentice Shield Prime (1, 2, 6, 8, 10, 11, ...), Apprentice Signature Shield Prime, Emblem Prime (2, 3, 4, 5, 7, 8, 10, ...) and Team USA (1, 2, 5, 6, 9) - because each is a shorter-roster parallel that keeps its base set's numbers. RESOLVE BY: running the chase pass.
• G-016 cardboardconnection IS THE ONLY EDITORIAL LINEAGE THIS RELEASE RESTS ON AND ITS INDEPENDENCE FROM tcdb WAS NOT ESTABLISHED. checklistcenter, checklistinsider, sportscardchecklist and waxstat were not swept; per CAV-037 a constructed checklistcenter URL that 404s is evidence of nothing, so no path was constructed, and per CAV-040 sportscardchecklist's per-product indexes can be contaminated with another product's sets. beckett.com was not consulted - confirmed 302-redirecting to a maintenance page across seventeen products (CAV-033). blog.paniniamerica.net was not swept for this product, which is a real omission: the scope memo section 7.6 found it first-party, reachable and paying twice on exactly the questions the open web is worst at, and it is where a base-card numbering statement (G-001) or a name for the legends subset (G-004) would most plausibly live. WHERE cc AND tcdb AGREE THEY MAY BE ONE LINEAGE, NOT TWO: they match exactly on the Crusade ladder (149/99/49/25/10/1), on four base-tier figures (199/149/49/10), on Armory /99, Emblem /99 and Team USA /99, and on eight set counts. Both plausibly descend from Panini's sell sheet, so every source_confidence=multi row in this release should be read as two statements of possibly one opinion. WHERE THEY DIVERGE IN CHECKABLE DIRECTIONS, which is the only real independence evidence I have: cc carries a Retail marking on Lord and Viscount that tcdb does not carry at all; cc names a Crusade set that tcdb has no sid for; tcdb carries a Crusade Camo tier that cc does not name; and they disagree on four set counts (G-005). RESOLVE BY: sweeping a genuine second editorial lineage and re-reading every multi row.
• G-017 RUN THE GAUNTLET CARD 24 CARRIES THE FLAG TOKEN UER AND I COULD NOT LEARN WHAT THE ERROR IS. The token appears on the base set (sid 154359, card 24, which Checklist.cfm identifies as Giannis Antetokounmpo) and again on Run the Gauntlet Red (sid 154361, card 24, rendered as "SN99, UER"). UER is recorded as a literal string per EXTRACTION_SPEC 3.2 and no gloss was accepted for it - the fetch layer is documented to invent expansions for tcdb flag tokens. The endpoint that would normally resolve it, https://www.tcdb.com/Errors.cfm/sid/149101/2016-17-Panini-Excalibur , returns "No errors or variations have been added for this set yet." - so the parent carries no errata rows and per CAV-020 that silence means nothing. It is worth noting the token did NOT appear on Run the Gauntlet Holo Gold, so it is not uniform across the family. This matters beyond curiosity because UER is the flag that surfaced the 2020-21 Hoops Premium non-integer card-number trap (EXTRACTION_SPEC 2.4); nothing of that shape appears in THIS release's base run, which enumerates as 200 contiguous integers. RESOLVE BY: Errors.cfm on sid 154359 itself, or a marketplace listing describing the error.

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