OpenFUT/docs/CARD_TAXONOMY.md

FIFA 17 FUT — Card Taxonomy (single source of truth)

Status: verified 2026-08-12. This document supersedes every earlier scattered taxonomy claim in the repo (see "Superseded taxonomy" at the bottom).

Evidence labels: OBSERVED = read directly from a shipped table or decompiled function; INFERRED = derived from OBSERVED facts; HYPOTHESIS = plausible, not yet proven. UNKNOWN is a valid answer and is stated as such.

Provenance

• Authoritative tables: 10.10.0.105:/home/alex/Documents/OpenFUT/fifa17-recon/data/tables/, copied byte-identical into this repo at fifa17-recon/data/tables/. SHA-256 manifest + verification: docs/evidence/fifa17-recon/table-hashes.sha256 (36 files: 31 fcc_*.json + 5 staff tables; combined hash-of-hashes 10f239add919089354d8dbff873fc9737b0a0f80f6ac41b1aa2a096c0ec8d331).
• Each table file is a DECODED dump {table, source, rowcount, schema[], rows[]}; card instances are in rows[], one object per card. Row counts and id ranges below were re-extracted from the in-repo .120 copies.
• Semantic (subtype→kind) labels are decompiler-derived, carried in fifa17-recon/tools/fut_consumables.py (BY_SUBTYPE, _DOC), generated by tools/build_consumables.py from FUN_1800d8330, FUN_18013f4d0, FUN_1801bfac0, FUN_1801aa230, FUN_180048780 + the fcc tables.
• Staff family selection: fifa17-recon/docs/CARD_SYSTEM.md (2026-08-04/05 blocks) and fifa17-recon/tools/fut_staff.py / fut_coaches.py.

Family selector (OBSERVED, binary)

Before registering an item, FUN_180141660 merges the client's local DB. It switches on record +0x4c (cardtype), which FUN_1800d8330 derives from JSON atom 0x6c cardsubtypeid alone:

cardsubtypeid 0..3 -> cardtype 1   players
              4    -> cardtype 2   managercards
              5    -> cardtype 3   headcoachcards
              6    -> cardtype 10  gkcoachcards
              7    -> cardtype 5   physiocards
              8    -> cardtype 4   fitnesscoachcards
              0x1e,0x1f,0x91..0x96,0xe7..0xe9,0xec -> cardtype 9  club items + misc
              absent (default 342) -> cardtype 0   no merge

Key: carddbid == resourceId for non-player families (queried RAW, no mask); players are queried by playerid = resourceId & 0xffffff (atom 0x287), and assetId (atom 0x23) is never read by the merge.

Consumable & staff families — evidence table (OBSERVED, .120 tables)

Family	Table	Rows	carddbid range	cardsubtype set	cardassetid (ART)
Contracts	fcc_contractcards.json	13	5001001–5001013	{201, 202}	{7, 8}
Fitness + Healing	fcc_healingcards.json	27	5002001–5002030	{211,212,213,215,216,217,218,219,220}	{9, 10}
Training (6 sub-families)	fcc_trainingcards.json	143	5003001–5003159	51-57, 61-67, 91-110, 121-136, 250-273, 300-340	{1,3,32,34,35,50,51}
Misc	fcc_misccards.json	42	5004001–5004042	{231, 232, 233, 236}	{43, 44, 45, 46}
Badges	fcc_badgecards.json	656	6000000–6000656	(none in row)	{39}
Stadiums	fcc_stadium.json	78	6200000–6200077	(none in row)	{36}
Kits	fcc_kitcards.json	1482	6300000–6400654	(none in row)	{35}
League logos	fcc_leaguelogos.json	44	8010000–8010044	(none in row)	{40}
League logo stickers	fcc_leaguelogostickers.json	39	8010000–8010039	(none in row)	{40}
Balls	fcc_balls.json	42	8120194–8120236	(none in row)	{37}
Managers	managercards.json	417	1000001–1001552	(staff; by cardsubtypeid 4)	(assetid col)
Head coaches	headcoachcards.json	124	2000004–2000328	(staff; subtype 5)	—
GK coaches	gkcoachcards.json	121	9000001–9000324	(staff; subtype 6)	—
Physios	physiocards.json	51	4000002–4000259	(staff; subtype 7)	—
Fitness coaches	fitnesscoachcards.json	115	3000019–3000328	(staff; subtype 8)	—

Notes:

• Contracts (5001xxx) — 201 = player_contract, 202 = manager_contract (OBSERVED, BY_SUBTYPE).
• Fitness + Healing (5002xxx) — subtype 214 is a valid enum value but ships ZERO rows (OBSERVED: 214 absent from rows[]). Enum split (OBSERVED, BY_SUBTYPE): healing = 211-218, player_fitness = 219, squad_fitness = 220. The +0x58 rareflag == 1 trap flips subtype 219 to squad-fitness.
• Training (5003xxx) — SIX sub-families, all OBSERVED in rows[] and labelled in BY_SUBTYPE:
  • gk_training 51-57 (7)
  • player_training 61-67 (7)
  • position_mod 91-110 (20)
  • formation_mod 121-136 (16)
  • chem/play styles 250-273 (24): player_playstyle 250-268 (19) + gk_playstyle 269-273 (5)
  • manager_league 300-341 in the client enum (42), but the shipped table contains only 300-340 (41 rows) — 341 is defined, not shipped.
  • Client enum also defines vestigial/unshipped ranges NOT in the table: manager_formation_mod 71-86, and DEAD_ZONE 58-60,68-70,87-90,111-120,203-210 (28). These have no rows.
• Misc (5004xxx) — subtypes {231,232,233,236}; cardassetid ART 43-46. cardsubtype 231 = 0xe7 anchors the 0xe7..0xe9 block to misc via FUN_1800d8330.
• Club items (badges/stadiums/kits/logos/balls) carry no cardsubtype in the row; they are keyed by carddbid range and distinguished on the wire by ART cardassetid (badges 39, stadium 36, kits 35, logos 40, balls 37). Kits live at 6300000–6400654 and are NOT badges (badges are 6000xxx). The client-side subtype→family map (which of 0x1e,0x1f,0x91..0x96 is ball vs stadium vs badge vs kit) is UNKNOWN — it is in none of the dumped tables and cardtype 9 has no miss-fill arm (see CARD_SYSTEM.md, "Club items", 2026-08-05).

Staff — subtype/cardtype selectors (OBSERVED, binary)

cardsubtypeid	cardtype	Family	Table	Rows	carddbid range
4	2	Manager	managercards.json	417	1000001–1001552
5	3	Head coach	headcoachcards.json	124	2000004–2000328
6	10	GK coach	gkcoachcards.json	121	9000001–9000324
7	5	Physio	physiocards.json	51	4000002–4000259
8	4	Fitness coach	fitnesscoachcards.json	115	3000019–3000328

cardsubtypeid -> rec+0x50 is the only family selector; FUN_1800d8330 maps it to rec+0x4c cardtype. Manager merge = FUN_1801356c0 (queries managercards by carddbid raw); coach merges live in the respective tables (fut_coaches.py). All five render live with real photos/bonuses, zero "DB Error" (CARD_SYSTEM.md, 2026-08-05).

Players (context, out of taxonomy scope here)

Players use cardsubtypeid 0..3 -> cardtype 1; merged by playerid = resourceId & 0xffffff against the local players table. Identity/name/nation/team come from the client DB; rating/position/attributes come from our item JSON. See CARD_SYSTEM.md and the migration work in openfut-adapter-fifa17.

Superseded taxonomy (report only — no other files edited)

The earlier working taxonomy (in prior agent-session notes / long-term memory, and partially in the original TablesTaxonomy scout output) got four things wrong. Corrected here:

1. Chemistry / play styles are 250-273, NOT 91-136. 91-110 = position_mod, 121-136 = formation_mod. (OBSERVED in fcc_trainingcards + BY_SUBTYPE.)
2. 6300xxx/6400xxx are KITS, not badges. Badges are 6000xxx. (OBSERVED.)
3. 5004xxx misc cards exist (subtypes 231/232/233/236, ART 43-46). Previously omitted. (OBSERVED, fcc_misccards.)
4. 8010xxx league logos exist (+ stickers), ART 40. Previously omitted. (OBSERVED, fcc_leaguelogos.)

Repo blast-radius of the superseded values: NONE. A repo-wide grep (docs/, openfut-adapter-fifa17/, openfut-core/, fifa17-recon/) for the wrong claims (91-136 chem styles; 6300/6400xxx as badges) returns zero hits. fifa17-recon/docs/CARD_SYSTEM.md and plan-2026-08-06-card-subsystem.md already use the correct 250..273 for chem styles and the correct staff subtypes 4-8; CARD_SYSTEM.md is a mechanism log, not a taxonomy, and asserts none of the wrong ranges. The superseded values therefore live only in non-repo agent memory, which should be corrected to point at this document.