The LMASTER Mod

A mod that restores content the PAL release cut or left broken, and offers a few improvements. Each restoration and improvement is a separate patch_disc.py switch first, so it can be built and tested in PCSX2 on its own. The mod comes in three editions, LMASTER lite, full and max, made of the groups below.

The switch levels

The switches come in three levels, and --mod takes a name of any level:

Editions: lmaster-lite (also lmaster), lmaster-full and lmaster-max, each a list of groups.
Groups: features that each work alone. Every switch is in exactly one group.
Switches: one feature each. Parts that need each other are one switch (the third tier's five parts are third-tier), and machinery is no switch: third-tier turns on mod-saves itself.

Group	Switches	Lite	Full	Max
`training-fixes`	camp-types, camp-cities, camp-santiago, camp-clubs, drill-names, salesman-items	yes	yes	yes
`injury-fixes`	injury-value, injury-row	yes	yes	yes
`club-fixes`	sponsor-negotiation, season-status, pr-cycle, mobile-site, overseas-pop, cup-last16	yes	yes	yes
`match-fixes`	rival-manager, press-tendency, national-cups, vpf-styles, vpf-positions	yes	yes	yes
`text-fixes`	retail-text (below)	yes	yes	yes
`leagues`	league-balance, third-tier	–	yes	yes
`restorations`	injury-lines	–	yes	yes
`names`	real-names, own-stadium (below)	–	yes	yes
`stadiums`	special-stadium, real-grounds, sponsor-boards (below)	–	yes	yes
`graphics`	realism (below)	–	yes	yes
`saves`	mod-saves	–	yes	yes
`harder-economy`	facility-tiers, wider-tickets, dearer-ads, sponsor-fees, tv-fees, season-demand (`ECONOMY.md`)	–	–	yes
`balance`	injury-odds, overseas-bases, transfer-fees, fee-tiers, slow-reputation, fanbase, merch-costs, pr-costs, camp-costs, tour-costs (below)	–	–	yes
`test-tools`	skip-tutorial, launcher	–	–	–

Which edition a fix goes in (Nexxus's rule, 2026-11-01): a bug fix that can't break a retail career already in progress goes in lite; one that may break such a save stays in full, whose careers go to their own folder. Lite keeps the retail save folders, so a career must carry on under it as on the retail disc.

LMASTER full is the fixes, the third tier, the balanced league schedules, the restored injury lines, the real names, the real grounds and the modded disc's own saves; LMASTER max adds the two difficulty groups, harder-economy and balance (Nexxus's split, 2026-10-27). Until then lmaster-full held what lmaster-max holds now, so the test reports before that date that name a full disc mean today's max.

Full and max always include mod-saves, so their careers and VS data go to their own memory-card folders and never mix with the retail game's (SAVE_FORMAT.md). Full's careers go to BESLES-54151-Fnnn and max's to -Mnnn: mod-saves takes M whenever a difficulty switch is on, since a career from a disc whose economy and injuries work differently isn't one to load on another. The VS data of both goes to BESLES-54151-Vnnn, as a VS team is the same on either. G and C stay the retail game's career and VS folders. The tutorial skip and the developers' launcher are test tools, in no edition.

patch_disc.py switches prints the tree. The editor's Build disc window (File > Build disc) draws it: a button for each edition ticks its groups (leaving the test tools as they are), each group has a box that ticks all its switches, and the switches show under a group when it is opened. The box of a partly ticked group shows as partly ticked, the window names the edition and the save letters the choice gets, and the build command names the highest levels that fit (patch_disc.mod_names: --mod lmaster-max,skip-tutorial, or lmaster-full,balance and single switches when a group is partly ticked). A disc built with the level names is byte for byte the one the switch lists built (checked on LMAST-test-full-germany.iso, 2026-10-27), and a max disc differs from the earlier full build in the VS letter alone (one byte, checked on LMAST-test-max-germany.iso). A test disc of each edition (full and max need --copies and --dat for the third tier):

python SRC/patch_disc.py patch "Let's Make a Soccer Team! (Europe, Australia) (En,Fr,De,Es,It).iso" lmaster-lite.iso --mod lmaster-lite
python SRC/patch_disc.py patch "Let's Make a Soccer Team! (Europe, Australia) (En,Fr,De,Es,It).iso" lmaster-full.iso --mod lmaster-full --copies --dat DAT
python SRC/patch_disc.py patch "Let's Make a Soccer Team! (Europe, Australia) (En,Fr,De,Es,It).iso" lmaster-max.iso --mod lmaster-max --copies --dat DAT

Rules for an entry:

Show from the game code or data that the content exists and is cut or broken, with addresses, as the other docs do.
Keep the fix as small as possible and say what it changes.
Test it in PCSX2 before it counts as restored.
A restoration that needs new text (a translation into the other EU languages, say) says so.
Compare with the Japanese release, the earlier build (see JAPANESE_RELEASE.md). Content it has and PAL lacks was cut for PAL; content missing from both was cut before release.

Status

Content	Kind	Evidence	Fix	Status
Main sponsor negotiation	cut (stub check)	`SPONSOR_NEGOTIATION.md`	`--mod sponsor-negotiation`	restored, tested in PCSX2
Debug menu	developers' switch	`RootMainSeq.sqb`	`--mod launcher`	works, tested in PCSX2. A tool, not part of the mod
Kind-6 injuries starting at 1600	broken data (dropped digit?), in both releases	`INJURIES.md`: `0x532d70` holds 1600 where 16001 fits	`--mod injury-value`: the u16 at `0x532d70` → 16001	built; how `calcKegav` uses the table is confirmed, but the effect isn't checked in the game (kind-6 injuries are rare and random)
Injury rows past fatigue 1000	bug (out-of-bounds read)	`INJURIES.md`: row 11 for ages 35+	`--mod injury-row`: the row clamped to 10 in `judgeKegaKind`'s own words (`0x21a08c`, 6 words)	built; checked in a MIPS simulator over 18,018 cases (only the six out-of-bounds ones change), not in the game (rare)
"Two months" and "nine months" injury lines	cut before both releases (the Japanese code is the same)	`INJURIES.md`: `plPinfo_KegaRecoverDaysChno` never returns chno 3 or 6; French, German, Italian and Spanish hold "See Let's Make revised text 0425"	`--mod injury-lines`: the heavy ladder (`0x216e14`, 15 words) gets thresholds 4,000 (two months) and 12,500 (nine months), and messages 4 and 7 of category 200 get text in the four languages, in the placeholders' bytes (`INJURIES.md`). It adds new text and shows lines the released game never did, so it is in `--mod restorations` (LMASTER full and max), not lite	built (thresholds chosen 2026-10-24); not checked in the game
The travelling salesman's items	unused boost and mismatched boosts, in PAL (the Japanese release isn't checked)	`EVSDATABIN_FORMAT.md`: flag 336's boost to the attack patterns (abilities 59–63) is paired with no event, so it never runs, and the special training coach's follow-up (event 190) only rests the squad; the Football Theory Headphones' text promises mental growth but they raise kick strength, tackle and contact strength, and the Lead-Spike Boots' promises power but they raise only crosses	`--mod salesman-items` (Nexxus's choice, 2026-10-31): event 190 gets handler type 34, so it sets flag 336 (`SIMPRG.REL 0x1d8cc0`), and event 188 ("the temporary coach's training has waned") clears it 8 turns later (`0x1d8cf0`, and the free pair of check type 58 at `0x1d92a8`), as it does the lecturer's 335; the boost jump table's entries for flags 333 and 334 (`SLES 0x5313c8`) trade places, so the spikes get the strength boost, and the headphones' case becomes new code over the crosses boost (`0x2105d0`, 33 words): 40% of the experience of abilities 26–32 for every squad player. The event table is edited in `EVENT/EVSDATABIN_EVENT.BIN` and its copies in `PRELOAD/SIMFILE0`–`6.PAC`. It fixes boosts that don't match their items and a coach whose training does nothing, and no retail career in progress breaks on it, so it is in `--mod training-fixes`, LMASTER lite (Nexxus's choice, 2026-11-01; it was in `restorations` before)	built 2026-10-31. The new code was run in a small MIPS simulator on a made-up squad (154 `pwkGUtl_AddExp` calls, each 40% of the right experience), and a whole LMASTER full build carries the change in all eight copies of the event table. Not checked in the game: the coach comes from season 8
Training camps train as the next camp type	bug (off by one), in both releases	`CAMP_FORMAT.md`: `CAMP_LIST` numbers the types 1–87, the training table and the theme check read them as 0–86; Parma, a goalkeeper camp, trains no keeping	`--mod camp-types`: every type lowered by one in `CAMP_LIST.TBB`	tested in PCSX2: one keeper camp raised all three keepers' saving, aerial ability and GK aptitude most (559–617 experience), then catching and rushing out (444–488), and left pace and acceleration alone, the fixed row's pattern (`CAMP_FORMAT.md`)
Seven English camps in the wrong town	PAL left the camp tables on the Japanese city numbers when it renumbered the English cities	`CAMP_FORMAT.md`: Liverpool's camp is in Scunthorpe, Manchester's in West London, and five more	`--mod camp-cities`: the seven city ids in `CAMP_LIST.TBB` and `CAMP_EXPLANE.TBB`	tested in PCSX2 (Nexxus's report): with low popularity the England list shows Southampton, Blackburn and Sheffield, not Basingstoke, Newport and Oakham. The other four need more popularity (Liverpool a sponsor) and weren't seen; the switch moves them the same way
Santiago's camp description	bug (wrong city id), in both releases	`CAMP_FORMAT.md`: `CAMP_EXPLANE` keys message 6146 to city 764, the camp is 774; the lookup (`SIMPRG.REL 0xa4a80`) finds nothing	`--mod camp-santiago`: the description's city 764 → 774 in `CAMP_EXPLANE.TBB`	tested in PCSX2 (Nexxus's report): Santiago's camp shows "The Chilean capital is a hub of South American finance …"
Canberra's last two camp clubs	bug (loop bound), in both releases	`CAMP_FORMAT.md`: `SIMPRG.REL 0x4c8f8` reads 735 of `CAMP_LIST` table 1's 737 pairs, so Fukuoka and Kanagawa never appear	`--mod camp-clubs`: the bound `0x2de` → `0x2e0` at `SIMPRG.REL 0x4c9d4` (one instruction; whole disc image)	tested in PCSX2 (Nexxus's report): Canberra's camp lists South Melbourne, Aichi, Fukuoka and Kanagawa
Drills 75–86 under the next drill's name	broken text (PAL: names one message early, an added "Jackknife")	`TRAINING_FORMAT.md`: practice n's name is message 971 + n of category 801; the Japanese names (both discs) and the training rows have a dummy at 74 that no coach teaches, PAL's names in all five languages start one early, so the drill named "Pass training." trains shooting	`--mod drill-names`: messages 1045–1056 → 1046–1057 in `801_1`–`6.mbb` of `MES.PAC` and their copies in `PRELOAD/SIMFILE1`–`6.PAC` (entry 28); the jackknife name goes to the dummy's 1045, so the files keep their sizes. No new text; a `MES.PAC` target gets the fix in its own bytes	tested in PCSX2 (Nexxus's report): coach R.Bloemzaad's drill 76 (short and long pass, cross) reads "Dribble training" on the retail disc and "Pass training." on `LMAST-test-drills.iso`
Play styles of an imported Virtua Pro Football player	bug (a compare function that subtracts pointers)	`PMPLAYER_FORMAT.md`: the import's qsort compare (`SAVEPRG.REL 0x3df00`) never reads the ratings, so slot t keeps style t when rated 70+, and 18 of the 22 VPF styles never come in; G.Binns, rated 86 in a style that doesn't count, came in with none	`--mod vpf-styles`: the compare's 4 words load both ratings and return b − a through the tail of the executable's `strcmp`, and one more rates style 0 ("none") at 0 in the sorted copy so it never takes a slot (`PMPLAYER_FORMAT.md`)	restored, tested in PCSX2: an edited V. Drake came in with four styles in slots 1–4 where the retail import gives one
Retail text: wrong leagues in two Game Over conditions, the Spanish intro's club count, "AL-Hyde"	broken text, in PAL	Retail text fixes	`--mod retail-text` (lite): the messages' own text in `MES.PAC`, and the name's `STATIONMES` copies	tested in PCSX2 with the real names on `LMAST-test-names.iso` (Nexxus's report, 2026-10-30)
Made-up league, competition and club names	licensing (the retail game names unlicensed clubs after their cities or grounds)	Real names	`--mod real-names` (full and max): `SRC/realnames.csv`'s names, the stadiums' too, in `MES.PAC`'s name lists and the English mentions, with `--copies`; `--mod own-stadium` names your stadium after you	tested in PCSX2 (Nexxus's report, 2026-10-30, `LMAST-test-names.iso`): the names, the stadium names and the new grounds work. The 75 non-English mentions are hand edits (`SRC/realnames_phrases.csv`), checked by `realnames.py info` but not seen in the game. `own-stadium`, the official names and the nations' names tested too (2026-10-31, below)
Crest presets 4–11 in three tables	broken data (a missing byte)	`EMBLEM_DIR.md`: tables 93, 101, 105 of `EDIT_EMBLEM.TBB`	add the missing `04 00` key	candidate. Check first whether the crest editor shows the gap
The rival's manager by style	bug (wrong register), in both releases (Japanese `0x25e8e8`)	`TEAMINIT_FORMAT.md`: the loop at `0x25e6d0` finds the record for the rival's style, but its load reads record 0, so the rival always gets H.Raphanel (898), never I.Ricardo (1,970), L.Alfano (1,247) or L.Nobili (1,246)	one instruction at `0x25e6d8`: `lw $s3, 8($a0)` (`8c930008`) becomes `lw $s3, 4($v1)` (`8c730004`). The rival's style is already the opposite of yours (`0x5531e0`), so its manager then matches it. `--mod rival-manager`	fixed, tested in PCSX2 (Nexxus's report): in a new Teamwork career in England, the rival SC Dunstable's Information screen showed a different policy area (the big blue area on the right, which follows the manager) from the unpatched game. The game never shows the rival manager's name. The other three styles weren't tried
The press tendency counts only as 0 or 1	bug (a constant), in both releases	`PBDATA_FORMAT.md`: both press readers clamp a player's level to a per-side cap, which a career match leaves at its constructor's 2 (`GAMEPRG.REL 0x74320`), and each reader turns 2 into the cap 1 with a `movz` in the delay slot after setting 4, so levels 2–4 act as 1	`movz $s0, $a1, $v0` (`0x9a474`, `00a2800a`) and `movz $s0, $a3, $v0` (`0x9a52c`, `00e2800a`) become `nop`, which keeps the cap 4 for all five levels. `--mod press-tendency` (lite)	built. A first version patched the cap's setter `0x74380`, which a career match never runs: over PINE in a German league match its patched word was in memory and both caps still read 2 (2026-10-31). Checked over PINE in a German league match on `LMAST-test-season2.iso` (2026-10-31): both readers' words read 0 (`nop`) in memory, with the caps at `02` as expected. Whether high-press sides play differently is a judgement left to Nexxus (section 7)
The World National Cup final on the last turn of the year	bug (schedule data)	the knockout UID 122 plays turns 92–95, and 95 is the year's last turn (`SCHEDULE_FORMAT.md`), so the season ends on the turn of the final and its results can't be viewed in the Competition menu (Nexxus's report). The group UIDs 114–121 play 89–91. The Euro National Cup ends on 94	move both national cups back one turn (`schedule.py shift`, masks and start turns; the end turn stays 95): World Cup groups 88–90 and knockout 91–94, Euro groups 88–90 and knockout 91–93. No league or cup plays after turn 87, and the two cups are in different years of the cycle. The national-team calendar (`SCHEDULE_LIST.TBB`) moves with them: match days 88–94 and 88–93, release on the final's turn, and the call-up on 88 with the first group game (87 has the second-division play-off finals). `--mod national-cups` writes `SCHEDULE_SYSTEM.PAC` and its copy in `PRELOAD/STATIONFILE.PAC`, and `SCHEDULE_LIST.TBB` and its copies in `PRELOAD/SIMLOCALMEM0`–`6.PAC`. With a league of another size (`leaguesize.py build`, patched with `--copies`), it moves the cups in the league build's `SCHEDULE_SYSTEM.PAC` instead: a rebuilt league keeps its first and last game day, so it never reaches turn 88. Checked with England at 22 clubs: one build holds both, and `schedule.py info` finds no new problems	tested in PCSX2 (Nexxus's reports): the Schedule screen's globes follow the new calendar (Euro year 2007-08: turns 88–93, week 4 empty; World Cup year 2009-10: turns 88–94, week 4 late empty). The World Cup's results after its final aren't seen, since that year is the fourth season of a career, but are taken to work: in the retail game the Euro National Cup's final is on turn 94 and its results can be viewed in the Competition menu after it (Nexxus's report). The moved World Cup final is on that same turn, and both cups' knockouts are built the same way (an end turn of 95 and the same flag); the bug came only from a final on the last turn
Season-ticket demand by club status	bug (branch order), in both releases (Japanese `0x2728f0`)	`ECONOMY.md`: `pwkUnkei_GetExpectSeat` has factors 0.5–1.0 for status bands of 10,000, but its first test takes every status over 10,000, so no club gets more than 0.6	`--mod season-status`: the ladder (`0x272840`, 31 words) rewritten as a straight run of the intended bands	restored, tested in PCSX2: a top-status club put 88,000 season tickets on sale and sold 72%, the fixed band's value (43% on retail; `ECONOMY.md`). Makes big clubs richer, so with `--mod harder-economy` it softens `season-demand`
PR events and the business cycle	bug (branch order), in both releases (Japanese `0x20e7f0`)	`ECONOMY.md`: each event holds +20, +10, −5, −10 for business cycles 0, 1, 3 and 4, but `plClubEvent_Execute`'s ladder reads byte 2 for 0–1, byte 4 for 2 and byte 3 for 3–4, so the boom's +20 never applies and an average economy takes 10% off every event (41% success instead of 51%)	`--mod pr-cycle`: the ladder's first 11 words (`0x20dfb0`) read the byte of each cycle, none for cycle 2	built 2026-10-25; not checked in the game
PR event 37, Renew Mobile Site	cut (one store clears it), in both releases (Japanese `0x20f108`)	`CLUBEVENT_FORMAT.md`: its three records are complete (from 2010–11, popularity 30,001, Jan–Aug, £50,000, +200 home and overseas popularity) and its mail texts are in `MES.PAC` (563: 11233–11234, the report 1290–1291), but `plClubEvent_GetList` clears its flag after every check (`sw $zero, 0xa4($sp)` at `0x20e868`)	`--mod mobile-site`: that store becomes a nop, so the event follows its own record like Renew Home Page	built 2026-10-26. Tested in PCSX2 (Nexxus's report, 2026-10-31, `LMAST-test-season.iso` with event 37 offered from the start): Renew Mobile Site is in the PR activities list, can be booked, and the month-end mail reports it
Overseas popularity ten times as fast with a rank-2 player	broken data (a misplaced decimal point), in both releases (Japanese `0x550bf8`)	`ECONOMY.md`: the overseas factor by your best player's rank from that region (`0x554ec8`) runs 1.02, 1.05, 10.8, 1.11, … in steps of 0.03, so a best player of rank 2 multiplies the monthly, match and title changes there by 10.8	`--mod overseas-pop`: the float at `0x554ed0` becomes 1.08	built 2026-10-26, confirmed from the code (the disassembled patch reads as intended). No in-game test: player rank and popularity are hidden, so a rank-2 player can't be picked out on screen
No European Club Cup popularity for the last 16	bug (compares the wrong competition), in both releases (Japanese `0x268aec`)	`ECONOMY.md`: `0x268858` gives the 4th points column (European League 120, European Club Cup 60) to places 8–15 only for compe 34 and to places 16–31 only for 35, so the European Club Cup pays its last-32 losers and not its last-16 ones	`--mod cup-last16`: places 8–15 get the column for both cups and 16–31 none. `xori $v1, $v1, 0x22` at `0x26894c` becomes `sltiu $v1, $v1, 0x24` (1 for 34 and 35), `xori $v1, $v1, 0x23` at `0x26896c` becomes `move $v1, $zero`, and the shared `movz` at `0x268970` becomes `movn`	filed by Nexxus and built 2026-10-26, confirmed from the code (the disassembled patch reads as intended). No in-game test: the points are hidden, and a last-16 exit takes a long run
475 combination records that never load	unused data (and a repeat check that only works for pattern 1)	`BPB_FORMAT.md`: `0x14c3f0` takes a team's records only from three 12-entry tables (`GAMEPRG.REL 0x28c7f0`, `0x28c820`, `0x28c850`) and from `0x24a5e8`[pattern − 1], the first of each attack pattern's 16. So 15 of each 16 (450 records, same script as the first, slightly different areas, no distance limits) and 25 basic records in no table never load. 22 of those 25 could start; 0, 1 and 18 have block A kind 1, which never triggers	Basic records (data only): swap entries of the three tables, same size, at their file offsets in `DLL/GAMEPRG.REL` (a `disc:` target). The duplicate 12 in set 0 is a free slot. More than 12 a set needs bigger tables: no free bytes follow them (strings do), so a new table elsewhere in the image, its address in the `lui`/`addiu` pairs at `0x14c484`/`0x14c488`, `0x14c4fc`/`0x14c504` and `0x14c56c`/`0x14c574` (local relocations, so only the immediates change), and the shuffle's count 12 (`addiu $a2, $zero, 0xc` at `0x14c468`, `0x14c4f4`, `0x14c564`). Pattern variants (code): a random 0–15 added to `0x24a5e8`[p − 1] in both branches of `0x14c3f0` (`fcEuroRand_Get0toI`, as `0x14c6f8` calls it), which needs room for the extra call: rewrite the branch's three repeated loads as one. A data-only fallback points `0x24a5e8`[p − 1] at another fixed variant. Repeat check: compare with the record (`0x24a5e8`[p − 1]), not p + 60, at `0x14c718`. Each change needs its Japanese addresses through `gamever.py`	candidate. Check first whether the extra records play sensibly: their areas are untested, and only the first of each 16 has distance limits. The effect is hard to see in a match, so a test needs a way to tell which combination ran (PINE: the controller, the team's `+0x1788`, keeps the running entry at `+0x6c0` (`0xc1af0`); that entry's `+0x80` points at the combination, whose `+0x4c` is the record)
32 dialogue categories	cut before both releases (no reference in the Japanese code either)	`EVSDATABIN_FORMAT.md`: derby interviews, a facilities tour, notices	needs triggers in the event system and English text	research
The manager's match policy	bug? (reads the wrong field), in both releases (Japanese `GAMEPRG.REL 0xd8f44`, `0xf34e8`)	`PBDATA_FORMAT.md`: nothing reads the match policy (`PlTeamData +0x4889`); the match's tactic shifts by period and score (`GAMEPRG.REL 0xd96e8`) and what looks like its substitution tables (`0xf3cb0`) read the selection policy (`+0x4888`) instead, in groups that fit the match policy's names (act boldly and change tactics a lot together, "rarely change" in the mildest group)	two instructions: the offset `0x4888` → `0x4889` in `lbu $s2, 0x4888($v0)` at `GAMEPRG.REL 0xd970c` (`90524888` → `90524889`) and `lbu $v1, 0x4888($sp)` at `0xf3cb0` (`93a34888` → `93a34889`), through `gamever.py` for the Japanese ones	candidate. Check first in the game that the selection policy changes a computer club's in-match tactics or substitutions (`TODO.md` section 7)
Tactics of a much stronger club	bug? (a missing case), in both releases (Japanese `0x22f46c`)	`PBDATA_FORMAT.md`: `plTeam_SetStyleTeamParam` picks a column of slider steps (`0x534438`) from the strength ratio; above 1.19 it gives column 0, the same as up to 0.79, and column 4 (the opposite steps) is never used	one instruction at `0x22ef8c`: `movn $v1, $zero, $v0` (`0002180b`) becomes `addu $v1, $v1, $v0` (`00621821`), so the column is 3 + (ratio > 1.19). `$v0` is overwritten by the next instruction	candidate. Check first which end of a slider (0–6) is attacking: only then is it clear that column 4 was meant for the stronger club
Goalkeeping aptitude of imported VPF players	bug (unset fields left at 0), in PAL (the Japanese release has no import)	`PMPLAYER_FORMAT.md`, found by Ohayoghurt: the import stores only the first of a player's three positions (`SAVEPRG.REL 0x3ebc4`, `0x3ebd0`); the other two stay 0, the goalkeeper's cell, instead of 13 (none), and each gives the goalkeeper cell one more level, so every imported outfield player has goalkeeping level 2	`--mod vpf-positions`: the main-position picker the import runs just before (`0x3ddc8`, its only caller `0x3ebb4`) loses its 9-case jump table (`0x547c0`): the same cells are worked out in a few words (3i/2 for an even index i, 3(i − 1)/2 + 1 for an odd one, plus 1 when the player leans right), and the words saved write 13 to positions 2 and 3 of the edit work (`+0x24`, `+0x28`) and the `PlPinfo` (`+0x1b8`, `+0x1bc`), through the caller's `$s1` and `$s5`. 19 words at `0x3de48`; the two relocated words of the table's address (`0x3de58`, `0x3de60`) stay as they are. Players imported before the fix keep the extra level in the save	restored, tested in PCSX2 (Nexxus's report, 2026-11-01): on `LMAST-test-vpfpos.iso` (this switch alone) G.Binns, a forward, came in with only his forward cell lit and the goalkeeper box dark. The old and new picker were also run in a small MIPS simulator for all 9 indices and both leanings: the same cell every time, and all four positions set to 13
Coach's training plan by policy	broken data (rows out of step with the names), in both releases (Japanese `SIMPRG.REL 0x224d70`)	`PBDATA_FORMAT.md`: the plan has 6 rows (`SIMPRG.REL 0x228cc0`) for 8 training policies; man-marking (3) gets the physical row, physical (4) the formation row, and formation, condition and match practice (5–7) the match row	needs two more rows (a man-marking one and a condition one, written new) and the clamp `slti $v0, $t1, 6` at `SIMPRG.REL 0xf0850` raised to 8. No free bytes follow the table (other data does), so the 8 rows need room elsewhere and the table's `lui`/`addiu` at `0xf06a8`/`0xf06b0` pointed there. Six rows can't fit eight names whatever their order	candidate. Only the player's own club uses the plan (computer clubs train without it), so decide first whether new rows are worth it

Improvements

Not restorations: changes to data that works but plays badly. They follow the same rules (evidence, smallest change, tested in PCSX2) and stay separate switches, so the restorations can be used without them (LMASTER lite). LMASTER full has the league schedules, and LMASTER max all of them. --mod harder-economy and --mod balance are the difficulty groups, in max only; balance is the switches built 2026-10-24 to 26 with the values Nexxus chose (the groups' lists are in the switch levels; injury-lines is in Status, and in --mod restorations).

Change	Evidence	Fix	Status
Better home and away runs in the league schedules	`SCHEDULE_FORMAT.md`: the disc has one schedule per size, and some send a club to one venue up to 7 times running (the 22-club leagues, UIDs 11, 30, 40, 48) or 9 (the 26-club English second division with your club, UID 1); the 8-club VS league (UIDs 154–155) reaches 5	`--mod league-balance` rebuilds those seven UIDs with `schedule.balance_pack` (`set_league` at the same size): no more than 2 running, same clubs, game days and turns, and game bit `w0` 8 kept on its day. Only `PARAM/SCHEDULE_COMPETITION.PAC` changes, by 7 entries that keep their size, so its `.HED` and the `STATIONFILE` copy stay. With a league of another size (`leaguesize.py build`) it balances the build's pack, skipping the leagues the build already rebuilt. Checked with England at 22 and France at 18 plus national-cups in one patch. An improvement, so not in LMASTER lite	tested in PCSX2 (Nexxus's report): in a new England career on `LMAST-test-leagues.iso` (with the tutorial skip), the Schedule screen shows the club's league games alternating `AHAHAHAHAHAH…`, and each game's details screen agrees with it, where retail's first 16 in the English second division (UID 1, slot 24 or 25) are `HHAHAAHHAHHHHHAA` or `AAAAHHAAAAAAAHHH`. So the game takes its fixtures from the rebuilt pack. Not played to the end of a season, and the other rebuilt leagues (the 22-club ones, VS) weren't looked at; they are built the same way. `python SRC/schedule.py league DAT/PARAM` shows what it rebuilds
A harder economy that punishes mistakes: dearer facilities, a wider ticket price range, dearer advertising, and the like	`ECONOMY.md`, confirmed: a build charges the facility record's cost (`PLRESOURCECOMMON.PAC` entry 0) × the business cycle (75–125%) × any sponsor discount; the plan screen's limits are constants in its set-up (`SIMPRG.REL 0x3eda0`); the ad budget and ticket price set attendance through a 21 × 11 table that only covers prices 60–300 and budgets up to 30,000,000, and the budget adds up to 50 supporter popularity a month; the salary limits, £10,000–£12,000,000 (`TEAMINIT_FORMAT.md`)	`--mod harder-economy`, six switches with the values Nexxus chose (2026-10-23). The cost side: `facility-tiers` (`plrcommon.facility_tiers`, data): build costs and upkeep ×2 for all equipment and adverts; grounds (sites) ×2 and ×4 (their upkeep is never charged); club houses: the starting one's upkeep ×2 (£25,000 → £50,000), the next ×2 (£15M, £150,000), the top ×4 (£60M, £600,000); stadiums: 15,000–25,000 ×2 (£15M, £45,000 a stand), 30,000–40,000 ×3 (£52.5M, £124,500), 50,000–60,000 cost ×5 and upkeep ×4 (£150M, £290,000), 110,000 cost ×3.5 and upkeep ×6 (£350M, close to the record's limit of about £358M; £600,000). The stand costs within a stadium aren't traced, so they stay. `wider-tickets` (code and table): both ticket screens go to 480 (£80), and the ticket demand table is respaced over 60–480 in steps of 21: up to 300 the old curve, above it crowds fall as (300 ÷ price)^1.47, so £80 draws half of £50's crowd. `dearer-ads` (code): the ad budget buys half as much per pound (popularity budget ÷ 1,200,000, a demand column per 6,000,000) and goes up to 60,000,000 (£10M), so the full effect costs twice as much. The income side, for the late game (chosen 2026-10-23): `sponsor-fees` and `tv-fees` (tables in `SIMPRG.REL`) keep a sponsor fee or TV offer up to £2M and halve the part above (£45M → £23.5M, the £50M TV offer → £26M; G006's TV £31.2M → £19.1M a season); `season-demand` (code) halves season-ticket demand (G006 would sell about half its 36,000, £20.5M → about £10.5M). The season-ticket status bug is fixed separately, by the restoration `--mod season-status` (Status)	tested in PCSX2 (Nexxus's report, `LMAST-test-economy.iso` with `--mod harder-economy`): the facility window showed the tiered costs, the plan screen and the other competitions' ticket screen went to £80 and the ad budget to £10M, and a home match at the new prices drew a crowd; "all seems fine". The three income switches were then tested on the same disc with all six (Nexxus's report: sponsor fees, TV offers and season-ticket sales as on the test sheet; "all fine"). Not yet played through a whole season. The scripts named after season tickets, spectators and popularity are never run (`SQB_FORMAT.md`); the C code above is what runs. Played through a season with every max switch (Nexxus's report, 2026-10-27, a German career on an LMASTER max build): the fanbase grows much more slowly, as the ad budget's halved popularity (`dearer-ads`) predicts, and the game feels much harder economically
Long injuries more common	`INJURIES.md`, confirmed: a heavy match injury is six months or more 0% of the time for a rested player and 39% at full fatigue, so rotation all but rules them out	`--mod injury-odds`: kinds 4–8 doubled in every row of both kind tables, to at least 2% (practice) and 5% (match) of a row, taken from kind 3 and then kind 0 so light match injuries stay as they were; a heavy match injury is then six months or longer 5% of the time rested and 78% at full fatigue (`INJURIES.md`)	built; not checked in the game
Overseas bases: dearer, each pound buying half as much	`PLRESOURCESIM_FORMAT.md`, confirmed: one cost per region sets the set-up, the investment per level and the upkeep; the level (0–3) scales the region's popularity change by 0.6–1.2	`--mod overseas-bases`: the 13 costs doubled in `PLRESOURCESIM.PAC` and its `SIMLOCALMEM0`–`6` copies, so the investment per level and the upkeep double too; what a level does stays as in retail (Nexxus's choice, 2026-10-26: like `dearer-ads`)	built; not checked in the game
Reputation harder to raise	`SAVE_FORMAT.md`, confirmed: the club status (your club rank is status >> 11) rises after each match (win and draw factors at `0x399668`), at final places (points at `0x555188`, `0x555260`, `0x555288`) and by 500 each year end (`0x26e2c4`)	`--mod slow-reputation`: every gain halved, losses as before (Nexxus's choice, 2026-10-26)	tested in PCSX2 (save states, 2026-10-27): a league win gained 47 status where retail gives 95 (`SAVE_FORMAT.md`). The club status is hidden in game (Nexxus's report), so it is read from saves or save states. The year-end and final-place halves aren't seen yet
Transfer fees by player rank: lesser players dearer, the best up to £250M	`ECONOMY.md`, confirmed: the player's rank enters the fee only through table A (`0x398478`, 16 ranks × 6 club-rank bands), which only the fee function reads	`--mod fee-tiers`: table A's rows ×2 for ranks 0–5, ×1.5 for 6–11, ×2 for 12–14 and ×3 for rank 15 (Nexxus's choice, 2026-10-26); with `transfer-fees` the best player at a top club reaches about £246M	tested in PCSX2 at the low end (Nexxus's report, 2026-10-27, a German career on an LMASTER max build): lesser players' fees about 4× retail, with `transfer-fees`; a fee above £120M not yet seen
Transfer fees higher, up to £250M	`ECONOMY.md`, confirmed: one base (1,350,000.0, `0x247288`) starts every fee you pay and every offer you get; price kind 1 caps them at £120M	`--mod transfer-fees`: the base 2,700,000.0 and the cap 15,000,000 (£250M) in all four currencies (`0x24a1c8`–`0x24a21c`)	tested in PCSX2 (Nexxus's report, 2026-10-25): fees doubled; a save state holds the 15,000,000 cap in all four currencies. A fee above £120M, and any limit of the transfer screens, not yet seen (`ECONOMY.md`)
A bigger fanbase counts for more	`ECONOMY.md`, confirmed: merchandise starts from a base by supporters, a flat table (12.5× from the smallest fanbase to the biggest); crowds and season tickets weigh popularity against a fixed part; TV offers vary only 0.9–1.15 by popularity	`--mod fanbase`: the merchandise bases ×0.5 at the bottom to ×3 at the top (`SIMPRG.REL 0x1d9ec8`), popularity weighted 195 for 130 and the fixed part 1,500 for 3,000 in crowds and season tickets, and the TV factor 0.75–1.25	merchandise income much lower (Nexxus's report, 2026-10-27, a German career on an LMASTER max build), as the halved bases of a small fanbase and `merch-costs` predict; the top of the scale, crowds and TV offers not yet compared. Goods prices stay, and their costs are `merch-costs` (below)
Dearer PR events	`ECONOMY.md`, confirmed: an event costs its record's `+0x14` × 300, £5,000–£500,000	`--mod pr-costs`: × 600 (`0x20deec`), so the booking and the refund both double	tested in PCSX2 (Nexxus's report, 2026-10-27, a German career on an LMASTER max build): PR events cost double
Dearer camps and tours	`ECONOMY.md`, confirmed: rates by camp level and region, and by tour region, × the squad × 300; a tour earns £1.2M–£6.3M against under £1M	`--mod camp-costs`: the camp rates × 2 (`SIMPRG.REL 0x1ce4e8`); `--mod tour-costs`: the tour rates × 3 (`0x1ce6c0`), so a tour where the club is little known about breaks even. Tour income unchanged	built; not checked in the game
Dearer merchandise stock, smaller margins	`ECONOMY.md`, confirmed: a unit of stock costs 30% of its price, less a bulk discount of up to 40%, paid when ordered (payment 22), so goods sell at margins of 70–82%	`--mod merch-costs`: the unit cost 50% (`SIMPRG.REL 0xbcce4`) and the discounts halved, down to 80% for the biggest lot (`0x1d10d8`); margins 50–60%	tested in PCSX2 (Nexxus's report, 2026-10-25): orders cost half the price a unit, the biggest lot gets 20% off, margins 50% (`ECONOMY.md`)

Ideas for the harder economy (Nexxus's), with where each lives. All are built now: the first three as --mod harder-economy, the fourth as --mod season-status and the rest in --mod balance (above):

Idea	Where it lives	Known
Dearer facilities and their upkeep	`PLRESOURCECOMMON.PAC` entry 0 facility records: build cost, a tenth-size second cost at `+0x1c`, upkeep (`ECONOMY.md`)	confirmed: costs and upkeep are data, so this is a data edit; the business cycle and sponsor discounts still scale the build cost
A wider ticket price range	the plan screen's limits (`SIMPRG.REL 0x3ef94`, `0x3ef98`) and the ticket demand table (`0x556250`, rows by price, thresholds at `0x556120`) (`ECONOMY.md`)	confirmed: both are fixed at 60–300 (£10–£50). Widening the screen alone makes every price past either end draw the same crowd as that end, so the table needs new rows as well
Dearer advertising	the ad budget's limit (`SIMPRG.REL 0x3eec0`) and what it buys: ticket demand up to 30,000,000 and budget ÷ 600,000 supporter popularity a month (`ECONOMY.md`)	confirmed: payment 17 is a twelfth of the budget a month, so a dearer ad means a smaller effect per pound (change the 600,000 divisor at `0x268ce0` and the table's 3,000,000 step at `0x2719e0` and `0x271a88`), not a higher price
Season tickets for top clubs	the status factor in `pwkUnkei_GetExpectSeat` (`0x272838`–`0x2728dc`)	confirmed bug: every club status above 10,000 gets 0.6, so the 0.7–1.0 steps for higher status are never reached (`ECONOMY.md`). Built as the restoration `--mod season-status` (above)
Dearer overseas bases, investment in them and their upkeep	`PLRESOURCESIM.PAC` entry 1: set-up cost per region and level rates (`_GetOverseasBranchEstablishCapital`, `0x232d88`; `PLRESOURCESIM_FORMAT.md`), payments 18–19	built as `--mod overseas-bases` (above)
Merchandise: dearer to stock, bigger margins	income 7 (Merchandise), `pwkGd_GoodsMonthlySales` (`ECONOMY.md`)	built as `--mod merch-costs` (dearer stock, margins 50–60%) and `--mod fanbase` (sales by fanbase), above (`ECONOMY.md`); the shop prices (shop list `+0x8`, `0x241a00`) stay
Reputation harder to raise	your club rank is the club status >> 11 (`SAVE_FORMAT.md`)	built as `--mod slow-reputation` (above)
Lesser players dearer, the best up to £250,000,000 (a top tier, a "football icon")	transfer fees (income 9, payments 14) and the price limits (`TEAMINIT_FORMAT.md`)	built as `--mod transfer-fees` and `--mod fee-tiers` (above)

Built as --mod third-tier (Nexxus's idea, design agreed 2026-10-26): a third tier and a harsher end.

The design:

Every nation has a third tier that isn't played on screen, in all six leagues, whether you play there or not.
Only the second division's last club goes down each season, and one third-tier club, chosen at random, comes up.
In your own nation only, two clubs start in the third tier, sent down from the second division at the start, so the second division with your club and the rival in it has its real size (England 24). The other nations' second divisions don't take your club and the rival, so they keep their clubs, and their third tiers start with their three real clubs (Nexxus's correction, 2026-10-27). One club down and one up each season keeps every tier at its starting size.
A player who finishes last in the second division loses at once, after one season, so they never end up in the third tier.
Third-tier clubs should still show in the other clubs' club info and in the world rankings, if the game allows it.

What is known so far:

The game over. Confirmed (SIMPRG.REL 0x15d958): from 2007, with your club in the second division, the game ends when your final rank in each of the last two seasons is past the division's line, which 0x15d8f0 takes from the down place slots of its REGULATION.TBB row (the club count less the slot's count). The warning before it (0x15da60) uses this season's live rank and last season's. patch_disc.py --mod third-tier (built as one-season-end) makes both decide on one season (not checked in the game). With the down count at 1, the line is the last place.
A third tier was planned. Empirical: every second division's REGULATION.TBB rows, own-nation and other-nation versions (UIDs 1 and 2, 11–12, 20–21, 29–30, 39–40, 48–49), send their bottom two "down" to a competition the disc has no schedule for: 50 England, 51 France, 52 Germany, 53 Italy, 54 Spain, 55 Netherlands. The past records' seed table (PLRRSRC_INITTEAMDATA.TBB table 1) has 56 competition rows, of which the game reads 49 (INITTEAM_FORMAT.md). So the six third tiers look like a cut feature, and competitions 50–55 are where they would go.
The third tiers are named and have clubs. Competitions 50–55 are "English League 1", "French League Division 3", "Third German League", "Italian League C", "Spanish League Division 3" and "Third Dutch League" (message category 100009). Their UIDs, 142–147 (first-promote, turns 93–95), are the tutorial's promotion playoff: your club (team 1) against three clubs of that nation's third tier (empirical, schedule.py entry DAT/PARAM 142): England's 442–444 are Gillingham, Nottingham and Rotherham, France's 445–447 Angers, Valence and Niort, Germany's 448–450 Essen, Oberhausen and Trier, Italy's from 451 Genova and Venezia. Clubs 442–459 have club records (PLRESOURCESIM.PAC entry 3) but no squads (OTEAMMEMBER.TBB covers 3–441).
Past records stop at 48. Confirmed: pwkRec_GetCompePastRecord (0x253590) refuses competitions from 49 (slti 0x31); a record is 0x98 bytes at save block 5 +0x260 + compe × 0x98 (team ids from +0x8, 32 of them). So competitions 50–55 keep no record a LAST_RANK slot could read.
No room for more records. Confirmed: the 49 past records end at block 5 +0x1f78, and the non-resident clubs' league records begin at +0x203c (0x24d330: up to 146 records of 0x1c8 bytes, found by bsearch), so 196 bytes lie between: not enough for six more records.
Clubs 442–459 can't join a league as they are. Confirmed: a non-resident club (442 on) gets a temporary league record and its 25 players from the database (0x24d6b0: ids 0x63f7 + 25 × (club − 442), 25,591 onward); for such a club pwkOteam_GetPointer (0x24b788) returns club 3's record (pwkOteam_Team2Otindex maps only 3–441, and −1 becomes record 1), so code that updates a league club's block 2 record (main league, rank) would change club 3's. In the retail layout block 2 has no room for more records.
The plan: grow the save (Nexxus's choice, 2026-10-26). LMASTER full saves are separate already (mod-saves), so the third tier may break save compatibility. Confirmed for the approach: the blocks' sizes come from the table at 0x533ac8, and PlPworkTask (0x21d6f0) allocates their sum from heap 6 in a shuffled order, so all code reaches a block through get(i); block 2's serializers loop a fixed 440 records (SAVEPRG.REL 0x29548, slti 0x1b8), and pwkOteam_Team2Otindex (club − 3 < 439) can map clubs 442–459 to records 440–457 by its bound alone. The phases:
1. The save: block 2 gains 18 records for clubs 442–459, and block 5 seven past records for competitions 49–55 at its end, with pwkRec_GetCompePastRecord sending them there; the serializers read and write the new space (save.py follows, as it runs them). Built (as save-expand, now part of --mod third-tier), with an LMASTER code segment in the executable for the new code (SAVE_FORMAT.md); tested in PCSX2: a converted career loads, saves and loads again (Nexxus's report).
2. Clubs 442–459 as full clubs: the game's limits 440 (0x1b8), 441 (0x1b9) and 442 (0x1ba), each checked; their records and squads at the start. The club info screen and the world rankings should then list them. The 206 sites are sorted in CLUB_LIMITS.md, about 65 of them about clubs. The order (Nexxus's choice to go on carefully, 2026-10-26): (a) the resident checks and the club loops, with squads for 442–459 added to OTEAMMEMBER.TBB; (b) the buffers and arrays sized for 441 clubs (ClubRank, the club-rank list, the schedule's club lists, the transfer AI), each read and resized; each step a switch of its own, tested on a disc. Step (a) is built (as tier-clubs) (CLUB_LIMITS.md), tested in PCSX2 (a new career and the tutorial's playoff). Step (b) is built (as tier-lists) (CLUB_LIMITS.md): the club-rank list, the transfer AI's lists, the offers' lists and the world ranking hold 459 clubs; tested in PCSX2 so far: the All Clubs Ranking lists every club (Nexxus's report).
3. The schedules: six hidden third-tier leagues every season (three real clubs, and in your nation the two sent down), the second divisions' last club down and the tier's winner up, two clubs sent down at the start in your nation, REGULATION.TBB's down count 1, and the one-season game over. Nexxus's choices (2026-10-27): the third tiers get new UIDs, 165–170 and their own-nation versions 171–176 (UID 164 is the default rules, UID_LIMITS.md), so the tutorial's playoff (UIDs 142–147) and VS mode's leagues (148–163) stay as they are; every table and loop sized for 164 schedules was found first, as in phase 2. The club that comes up is the tier's winner (LAST_RANK rank 1 of its competition), which changes each season as the games are simulated. Built (as tier-uids), (a) the leagues and (b) the movements (UID_LIMITS.md); tested in PCSX2: a German career's season end moves the second division's last club down and the tier's winner up, in your nation and the others, after the fix that keeps the tiers' results (UID_LIMITS.md). The rival goes down like any club (Nexxus's choice, 2026-10-27). Step (b): the movements, and your nation's tier at 5 clubs (an own-nation version of each tier, as the second divisions have). Found for (b), confirmed:
  - ScheCallback_ProcPromotion (0x1131c8), once at a career's start, puts your club and the rival in the first two empty places of your second division's past record (the nations' second divisions in a 6-entry table at 0x51ad70, nations 1–6 to competitions 1, 7, 12, 17, 23, 28); that is how they become ranks 25 and 26 in England.
  - The past records are seeded for competitions 0–48 only (0x253200, slti 0x31); the seed table's rows 49–55 are empty.
  - CScheEuro::calcCompetetionFinish (0x205fa8, slti 0x31) writes no record from competition 49 on. Past it, the yearly honours (pwkRec_GetCompeConventionRecord*, 0x253af8–0x253c58) exist for competitions 0–48 only and return null past them, so a third tier must write its table and skip the honours and the title effects.
  - Every own-nation version of a domestic UID (second division, cup qualifying, cup, league cup) differs from its other-nations version only by your club's and the rival's places, the cups chaining on through NOW_RANK of their own versions. With your second division at its real size, each becomes a copy of its other-nations version, its UID references moved to the own ones.
  The plan for (b): each season a second division takes its own ranks but the last plus its tier's rank 1, and a tier its own ranks from 2 plus the second division's last. Season 1 is seeded so that holds: a second division's seeded last place holds a tier club, and the tier's rank 1 that last club. In your nation, ProcPromotion instead places your club and the rival in the second division's ranks S − 2 and S − 1 and sends the two below them down to your nation's 5-club tier. A tier club in a seeded last place plays that nation's cups in season 1, as relegated clubs do in later seasons.
4. Two seasons played in PCSX2: done for a German career (the tiers' winners up, the last clubs down).
The five switches it was built as are one switch now, --mod third-tier (Nexxus's rule: switches that need each other are one feature), in --mod leagues (LMASTER full and max); it turns on mod-saves itself and needs --copies and --dat. The machinery under it is no switch: the LMASTER code segment (code_segment_words) and the save-growing wrappers (save_grow_words) are there for any later feature, and the old names are taken as third-tier. A full LMASTER build was compared entry by entry with one without the third tier: only the third tier's own files and entries differ, national-cups and league-balance applied on top of its schedule packs, and the switches whose PRELOAD edits share its rebuilt packs (drill-names, injury-lines, facility-tiers) land as entry edits of the rebuilt packs.
Other Club Info lists a nation's clubs from a loop that stopped at 441; third-tier runs it to 459, so the tier's clubs are listed (tested in PCSX2, Nexxus's report, 2026-10-27; CLUB_LIMITS.md). What is left to trace is in UID_LIMITS.md.

The text

--mod third-tier rewrites the messages that state the rules the tier changes, in English, French, German, Italian and Spanish (Nexxus's choice, 2026-10-27, from a reviewed draft): 136 messages, 27 in each language and one name. With the tier your second division has its real size, two clubs fewer than retail's own version, its last club goes down to the tier, and finishing last ends the game after one season (retail: two seasons past a line).

Messages (`MES.PAC` category:id)	What changes
487:1100–1105, the advisor's intro to your division	the club count and matches (England 24 and 46, France 20 and 38, Germany 18 and 34, Italy and Spain 22 and 42, the Netherlands 20 and 38), and a last line: finishing bottom is game over
487:1200–1205, Game Over conditions (help)	point 2: finishing last (24th, 20th, 18th, 22nd, 22nd, 20th) at the end of a season
30003:1, the game over screen's conditions	point 2: finishing last in the division at the end of a season
35680:0, the advisor's warning	finishing bottom this season, not twice
487:902, the game over message	"we finished bottom of the league", not "nothing for the past two seasons"
30001:101, 107, 112, 117, 123, 128, the second divisions' rules screens	a Relegation section: the club placed last goes down to the tier, by name
832:433, 436, 442, 438, 440, 445, the second divisions' season-start news	a closing sentence: the bottom club goes down to the tier

Which message belongs to which nation (empirical, from the data): the intros hold the retail sizes in the order England, France, Germany, Italy, Spain, Netherlands (26, 22, 20, 24, 24, 22); the Game Over texts follow it, and their Italian versions name each nation, so 1204 is Spain and 1205 the Netherlands, where the other languages repeat Italy (1204) and Spain (1205). The rules screens go by their promotion rules, and the news by the competition in each news record's date rule (EVSDATABIN_FORMAT.md): competitions 1, 7, 12, 17, 23 and 28, the second divisions. The code that picks the intros and Game Over texts isn't traced (no script or code site names their ids).

Retail slips in the same messages are fixed on the way: "216 equipos" (26) in the Spanish text of England's intro, a stray "Vereine" in the German text of Italy's, "41 partite" (42) in the Italian text of the Netherlands', the Spanish name of the Dutch tier, "Cuarta Liga Holandesa" (fourth, 11:1147), the Spanish game over screen's point 3, cut off at "es menor" (less than 8), and the nations of 1204 and 1205.

patch_disc.py keeps the edits as short replacements and its own sentences (tier_text_edits), applied to MES.PAC by a target hook (target_tier_text): the disc's own file, or one a target writes (the editor's Text tab), which keeps its own edits. Every grown file stays in its slot, so no entry moves; --copies carries category 832 to its copies in PRELOAD/NEWS1–5.PAC and 11_5 to STATIONMES5.PAC. Checked on LMAST-test-max-germany.iso (2026-10-27): the 136 messages read back as built, every other message equals the build without the text, and the five NEWS copies of 832 equal MES.PAC's; a --mod lmaster-full build with a MES.PAC target keeps the target's edit and gets the text. Tested in PCSX2 (Nexxus's report, 2026-10-31, LMAST-test-season.iso, a German career in English): the advisor's intro gives 18 clubs, 34 matches and the game over at the bottom, the help's Game Over conditions say last place (18th), the 2nd German League's rules screen has a Relegation section and the season-start news says the bottom club goes down. The other languages aren't seen.

The economy switches (tickets, adverts, fees, costs, reputation) need no text: no message states their old limits.

Retail text fixes

Messages the retail game itself gets wrong, fixed in LMASTER lite by --mod retail-text (group text-fixes; built 2026-10-30, tested in PCSX2 with the real names, below). The third tier's text, which states its own rules, is separate (The text).

Message	Retail	Fix
Game Over conditions, Spain (`MES.PAC` 487:1204), point 2	English names "Italian League B" (and says "23th"), German "Italienische Liga B"; French, Italian and Spanish are right	Spain's second division, 23rd place, in English and German
Game Over conditions, the Netherlands (487:1205), point 2	English "23th in Spanish League Division 2", French "division espagnole", German "23." and "Spanischen Liga 2", Spanish "2ª Liga Alemana"; only Italian is right	the Dutch second division, 21st place, in those four languages
Spanish intro to the division (487:1100)	"Hay 216 equipos"	26
Team 435's name (categories 3, 4 and 100003, ids 2435 and 3435)	"AL-Hyde"	"Al-Hyde" (Nexxus's report), in all five languages

The point-2 fixes follow the Italian texts, which name each nation's own league and place. Category 487 has no PRELOAD copy, so its files grow inside their slots in MES.PAC; the name keeps its size, so the switch patches its copies in PRELOAD/STATIONMES1-5.PAC byte for byte, and lite needs no --copies. A MES.PAC target gets the fixes in its own bytes. With third-tier, the tier's text rewrites point 2 again and finds the Spanish fix in place; with real-names, team 435 becomes Al-Sadd. The Japanese release is left alone.

Other retail typos seen while making the name table, not fixed: "Repulic of Ireland" (category 3) and "Motherwwell" (category 100003).

Real names

--mod real-names (group names, in LMASTER full and max; built and tested in PCSX2 on 2026-10-30) gives every league, competition, club and national team its real name of the game's era where the retail game uses a made-up one: Highbury as Arsenal, City as Manchester City, the European League as the Champions League (Nexxus's request, 2026-10-30). It also names each club's stadium as of 2005-06, the game's season, without sponsors where the ground has another name (Westfalenstadion, not Signal Iduna Park; Nexxus's choice, 2026-10-30). The club names match the ones LMAST Realism shows on its logos and kits, so text and textures agree. The game lets the player name their own club and themself with at most 18 characters (Nexxus's report), so the club names keep to 18 too. Like the licensed Spanish and Dutch clubs, the English, French and German ones use their official names as far as 18 characters allow (Arsenal FC, Olympique Lyonnais, Borussia Dortmund), abbreviated where they don't fit (Eintr. Frankfurt, Paris St-Germain; Nexxus's request, 2026-10-31), and so do every other club (AS Roma, AFC Ajax, AC Sparta Prague, CA Boca Juniors). The Spanish clubs' names, licensed ones included, have no dots (FC Barcelona, Real Madrid CF, RCD Espanyol; Nexxus's choice). The international competitions carry their governing body, as their official names do: UEFA Champions League, UEFA Cup, UEFA Intertoto Cup, UEFA Super Cup, UEFA EURO, FIFA World Cup and their qualifiers (Nexxus's request, 2026-10-31). The Standings menu's "European League Ranking" is a ranking of clubs, not the competition, so it reads "UEFA Ranking" (English, Italian, Spanish; Nexxus's report: it had read "Champions League Ranking"). The nations and national teams take FIFA's 2005-06 names (Nexxus's request, 2026-10-31): Serbia-Montenegro, United States, Republic of Ireland, United Arab Emirates (the team; the nation list, whose names run to 19 characters, says UAE), Côte d'Ivoire, Suriname, Trinidad and Tobago, FYR Macedonia and Cape Verde Islands; China PR, Korea Republic and Korea DPR already were. The nations are categories 6 and 7 (5000 + PlNati, the same names twice), which realnames.csv rows of category 6 write to both. National team names run to 22 characters in the game's own lists, so they aren't held to the clubs' 18.

The switch. SRC/realnames.csv holds only the real names (no game text): 1069 rows, keyed by category (3 teams, 10 stadiums, 9 competitions, 11 stages, 6 nations) and id, with the English, French, German, Italian and Spanish names and a team's three-letter form. realnames.py writes them into:

Category	Ids	What
3, 4, 100003	2000, 3000, 2000 + team	the team's name, in each language the row gives
10	11000 + team	the club's stadium (clubs 3–459). The game's is "<club> <word>" ("Highbury Stadium", "Highbury Stade"); a row gives the ground's own name, the same in every language, and a club with a real name and no row gets "<real name> <word>"
5	4000 + team	the three-letter form, in every language: the name's first three letters in capitals, as the game forms them, or an override where two clubs would share one (MCI, MUN, SHW, SHU)
9	10000 + competition	the competition's name; in capitals where the game has it so (French, Italian and Spanish)
100009	the same	the English name in capitals, in every language, as the game has it
11	1000 + stage	the stage or group's name
100011	the same	the same name where it repeats 11, or the English one where it holds English (in four languages run together, "European LeagueGroup A")

Stadiums (2026-10-30). 382 of the 457 clubs have a stadium row, the grounds they played at in 2005-06 (Nexxus's choice, 2026-10-30, after a first list of 2006-07): Highbury for Arsenal, not the Emirates Stadium; the Stadio delle Alpi for Juventus and Torino, not the Stadio Olimpico di Torino they moved to in 2006; the Alkmaarderhout for AZ, not the DSB Stadion; La Condomina for Real Murcia, not the Nueva Condomina; the Ciudad Deportiva for Real Madrid Castilla, not the Alfredo Di Stéfano. Sponsor names give way to the ground's own (Westfalenstadion, Arena AufSchalke, Volksparkstadion, Son Moix, El Sadar, Stadion Letná); where a ground had no other name (Reebok Stadium, JJB Stadium, Ricoh Arena, KC Stadium, Walkers Stadium, Britannia Stadium, Arke Stadion, Polman Stadion, Rosada Stadion, Mandemakers Stadion, Kras Stadion, Finnair Stadium, Home Depot Center, Fukuda Denshi Arena) the row is left out. Names that belong to the club or its owner rather than a sponsor stay (Madejski Stadium, BayArena, Volkswagen Arena, Philips Stadion, MSV-Arena, Wacker-Arena). The other 75 clubs, mostly the small and far-away ones whose 2005-06 ground isn't certain or needs letters cp850 lacks (Bežigrad, Andrův stadion), keep "<club> Stadium". The game's names are at most 30 characters ("Athletic Club de Bilbao Stadio"), and so are the rows; info lists any longer. The rows are suggestions for Nexxus to check.

--mod own-stadium (group names): your club's stadium is named after you, "<your name> Stadium", where the game used your city (Nexxus's request, 2026-10-30). Confirmed: Msg::VarBuf_MakeMyStadium (0x11eaf8) builds the name into variable slot 9, which message 11001 of category 10 (team 1, "USE Stadium") shows, as snprintf("%s %s") of slot 3 and the word "Stadium" (GetString type 0x28, message 0x52, in each language: Stade, Stadion, Stadio, Estadio). Slot 3 holds the city (Nexxus's report: the stadium defaults to "<City> Stadium"); slot 0 holds the name typed at the start (SetOwnerName_Keyboard, the {var:1:7} of the messages). The switch changes the slot, addiu $a0, $zero, 3 → 0 at 0x11eafc (24040003 → 24040000). The buffer is 33 bytes, so an 18-character name and " Estadio" fit. The rival's stadium (VarBuf_MakeRivalStadium, slot 10) is built the same way from the rival club's name and is left alone.

A new career builds the stadium name before you type your own, so it came out as " Stadium" (Nexxus's report, 2026-10-31; over PINE slot 0 held the name and slot 9 " Stadium"). A load rebuilds it from the saved slot 0 (SAVEPRG.REL 0x32f24, after restoring slots 0 and 1), but a new career doesn't, so the switch also ends SetOwnerName_Keyboard (0x11ec58) with a jump to VarBuf_MakeMyStadium: its jr $ra at 0x11ec74 (03e00008) becomes j 0x11eaf8 (08047abe), with $ra already restored and the frame popped in the delay slot (Japanese build: 0x11e6f4, jumping to 0x11e580). Tested in PCSX2 (Nexxus's report, 2026-10-31, LMAST-test-season4.iso): a new career shows "<your name> Stadium" straight away, and so does a loaded one.

Letters the European text's encoding (cp850) has no code for lose their accent ("Zeljeznicar", "Besiktas"), as the game's own names do. A competition's real name also replaces its game name in the English messages that mention it (26: the tournament rules, the news, the advisor's lines), matched in any case and outside {tags}; a qualifying round's mention takes its competition's name ("we'll qualify for the Champions League"). The other languages' 75 mentions are edited by hand, in SRC/realnames_phrases.csv: their game names are often plain phrases ("monter en première division"), and a proper name in their place needs its article and gender changed ("nel campionato italiano B" → "in Serie B"), which a swap can't do. Each row is one phrase of one message and its replacement, written against the game's text. In a build, retail-text and the third tier rewrite point 2 of the Game Over conditions first, so a row whose phrase isn't there is skipped, and the Netherlands' French conditions (487:1205, which the retail text calls the Spanish second division) have a row for each wording, both giving the Eerste Divisie; info marks a message that none of its rows matches in the game's own text. 63 messages get real names; 12 keep a generic phrase (an empty replacement), such as "première division" or "erste Liga" where any nation's top flight is meant, the World Club Championship's own name, and two European-cup wins that may be any cup (German 35557:0 "Europapokalsieg", Spanish 832:378 "Copa Europea").

The league rules' "European League Rank/position" (30001:100 to 127) is the nation's place in the European ranking (563:1009: "country X is ranked Nth in Europe… we'll qualify… if we're in the top 4"), not the Champions League, so the hand edits name it "UEFA Rank" in English before the automatic swap runs, and "ranking UEFA", "clasificación UEFA" in Italian and Spanish. The Italian "prima divisione olandese" (487:5) is the Eerste Divisie, the Dutch second tier, not the Eredivisie. Club names in other messages are left, since many are city names.

Every file still fits its slot in MES.PAC, which keeps its size; the name lists grow by about 550 bytes in each language, so their STATIONMES copies grow. The switch brings MES.PAC as a whole file and needs --copies (and --dat), like third-tier, whose text it follows: the tier's added sentences name the tiers' leagues, which then read "League One" and so on. PAL only.

Checked on LMAST-test-names.iso (LMASTER full with skip-tutorial, built 2026-10-30), read back from the image: Arsenal, Manchester City and MUN in MES.PAC, the Champions League in each language, "Arsenal" and "Stella Rossa" in the English and Italian STATIONMES copies the game loads, and "League One" in the third tier's rules. Tested in PCSX2 (Nexxus's report, 2026-10-30): the names showed as intended apart from two swapped pairs, since corrected (below). The career started in England although the club was made in Germany, because the disc was built without --skip-nation germany; that is the test switch, not the names. On the rebuilt disc (with --skip-nation germany) the names, the stadium names and the new grounds work (Nexxus's report, 2026-10-30).

The official club names, the UEFA and FIFA competition names, the "UEFA Ranking" label and the stadium names of every club were then tested in PCSX2 on LMAST-test-season3.iso (Nexxus's report, 2026-10-31, a German career loaded from the memory card): Arsenal FC, Bayern Munich, Juventus FC, FC Barcelona, Real Madrid CF and SK Rapid Wien read in full; the UEFA Champions League, UEFA Intertoto Cup, UEFA EURO and FIFA World Cup show; Standings has "UEFA Ranking" with "Here are the UEFA rankings."; and grounds without a row show the club's real name ("Bolton Wanderers Stadium"). On LMAST-test-season4.iso (the same day) the 2005-06 nation and national team names show, "United Arab Emirates" and "Republic of Ireland" fit wherever team names appear, and the nation list says UAE. The other languages aren't seen yet.

python SRC/realnames.py info DAT/MESSAGE/MES.PAC       # the names against the disc
python SRC/realnames.py import names/realnames.csv     # from the filled name table
python SRC/realnames.py apply DAT/MESSAGE/MES.PAC out.PAC

How the names were chosen:

The name table (groundwork, 2026-10-30). python SRC/mbb.py names DAT/MESSAGE/MES.PAC names/realnames.csv lists the 776 names to consider: 457 clubs (teams 3–459), 83 national teams (460–542), 72 competitions (category 9) and 164 competition stages and groups (category 11). Each row has the game's English name and length, an empty column for the real name with a length formula, the French, German, Italian and Spanish names, every copy (categories 4, 100003, 100009, 100011 and the short name in 5), the stadium named after a club (category 10, " Stadium") and the other messages that mention it. names/ is ignored, since the file holds the game's text, and the command won't overwrite a table that exists. What it shows:

Every game club name already fits 18 characters (the longest is 18). 113 competition and stage names are longer, so the limit can only be a guide for those.
80 club names differ by language (French "Real Madrid Cast."), and most competition names are translated, so each language needs its own real name, not only the English one.
Category 100003 differs from 3 in 7 names (spellings such as "Albinoleffe" and the typo "Motherwwell").
142 names appear in other messages too: "European League" in 15 (the tournament rules among them), the nations in news and supporter lists. Those messages need the same rename.
The PRELOAD/STATIONMES<lang>.PAC copies of categories 1 and 3–11 have little room to grow: 1,840 free bytes for English, 644 for French, 1,072 for German, 232 for Italian and 1,008 for Spanish (python SRC/preload.py info DAT). Longer names past that need the pack rebuilt (--copies does it), which the Italian copy will.

LMAST Realism holds no text list of names: its logos are named by texture hash, and its kit folders use the game's made-up names. Its league folders (ENG/Premier League and so on) each hold that league's or region's real clubs, so the table's last column, "LMAST Realism logo shows", was read from the logos by sight (contact sheets of each folder) and paired with the game's clubs of the same league or region. Where a game name could fit two crests, the club's kit colours decided (uniform.py show: Glasgow City wears green and white, so it is Celtic). The English folders are the 2005/06 clubs; the game's starting divisions are a season older (Crystal Palace, Norwich and Southampton in the top flight), which doesn't change any name.

Style (Nexxus's choice, 2026-10-30): short names, like the licensed clubs of Italy, Spain and the Netherlands ("Juventus", "Atlético Madrid", "PSV Eindhoven"), never shortened past the name a club is known by ("Real Madrid", not "R. Madrid"). The table's real name column holds a suggestion only where the game's name isn't how the club is known (Highbury → Arsenal, Gelsenkirchen → Schalke 04, La Boca → Boca Juniors); an empty cell keeps the game's name, and the licensed names stay. 340 suggestions and 220 logo readings are in; Nexxus then gave the names of ten clubs (Sheva → Beitar Jerusalem, North Shore → Auckland City and others), and the three "Club Cups": Coppa Italia Dilettanti, Copa Federación de España and KNVB Amateur Cup, the amateur cups (the first two are 23 and 25 characters). Nexxus confirmed the pairings that were in doubt (Velika as Radnik V. Gorica, Kutaisi as Torpedo Kutaisi, Kulubu as Bursaspor). Two pairs first went the wrong way round and were swapped after Nexxus saw them in the game (2026-10-30): the game's Principat is FC Encamp, and Diaconsa, whose Realism crest is FC Rangers' (1981), is FC Rangers; Baki is Inter Baku and Baku is Qarabagh. The World Club Championship is a made-up tournament and keeps its name. AL-Hyde is Al-Sadd (Sadd Sports Club, Nexxus's report): "Al-Sadd" in full and max, while lite only fixes its casing to "Al-Hyde" (Retail text fixes). That leaves Doha, which had the Sadd crest: the only Qatari crest left in Realism's Middle East folder is Al-Gharafa's, which Nexxus confirmed. No cell is left in doubt. Four game names end in a stray space ("Baku ", "Moskva ", "Újpesti ", "Ararat "), which their real names drop. 27 suggestions, all competition or stage names, are over 18 characters, as many of the game's own are.

The other languages. The table ends with "real fr", "real de", "real it" and "real es" columns (mbb.py names adds them), filled from the English suggestions: 1,350 cells. The game's own non-English names are descriptive translations ("Coupe d'Angleterre", "Englischer Pokal"), so the real names are mostly the same proper names in every language: the national leagues and cups (FA Cup, Bundesliga, Copa del Rey) stay untranslated. The UEFA and FIFA competitions take each language's usual name ("Ligue des champions", "UEFA-Pokal", "Coppa UEFA", "Copa de la UEFA"; the German "Weltmeisterschaft" and "Europameisterschaft" are kept). Stage names follow the game's own wording and order in each language ("Tour de …", "Gruppe A", "Gruppo A …", "Final …"), and category 9 stays in capitals in French, Italian and Spanish as the game has it. Clubs copy the English suggestion, except where a language has its own common form ("Étoile Rouge", "Roter Stern", "Sparta Prag", "CSKA Moscou", "1860 München"); every club suggestion fits 18 characters in every language.

Stadiums

Group stadiums (LMASTER full and max, Nexxus's request, 2026-10-30). Its switches change how big grounds are; the names of the grounds are in Real names.

How a ground's size works (confirmed). Two tables hold capacities:

Table	Whose	What
`PLRESOURCECOMMON.PAC` table 0.7 (`PLRESOURCECOMMON_FORMAT.md`)	your club's	9 stadium types with up to three stand levels each: types 0–1 8,000 (the start), 2–5 15,000/20,000/25,000, 6 30,000/35,000/40,000, 7 50,000/55,000/60,000, 8 (the Special stadium) 110,000 at every level. Gate money and the crowd use these
`STADIUM_DATA.TBB` (`INITTEAM_FORMAT.md`)	the 119 match stadiums	roof, level and capacity in thousands for each stadium id; a computer club's record names one id

In a match at your ground, PlGiTask::InitStadium (0x2131c0) turns your stadium type, league and the stand, roof and lights you have built into a stadium id through CONV_INFO_BUILD (plTeam_GetStadiumDataIndex, STADIUM_DIR.md), and SetMyAudience (0x212d9c) divides your crowd by that id's STADIUM_DATA capacity to fill the stands. The ids your builds reach are 0–101 (each level's ids hold that level's capacities: 0–5 8,000; 6–35 15,000–25,000; 36–65 30,000–40,000; 66–95 50,000–60,000; 96–101 110,000), so their capacities must follow your ladder. Ids 102–118 are only used by computer clubs. Computer clubs use 89 ids, 219 clubs on ids your builds share and 238 on the other 17, with 14 different capacities.

--mod special-stadium. The Special stadium (type 8) holds 100,000 at each level, not 110,000 (Nexxus's choice of 90,000 or 100,000). plrcommon.special_stadium writes the three capacity words of record 8 (+0x38, +0x48, +0x58), and ids 96–101, the match stadiums of a level-4 ground, go from 110 to 100 thousand so a full house still fills the stands. Computer clubs on ids 96–100 (Chelsea, Arsenal, Monaco, Bayern, Dortmund, Leverkusen, AC Milan, Inter, Barcelona, Boca Juniors) hold 100,000 too. The switch writes only the changed bytes, like facility-tiers, which writes the same records' costs (it now does the same): on a scratch disc with both, record 8 cost £350M and held 100,000, and the STATIONFILE copy and the seven SIMFILE copies of STADIUM_DATA matched. An editor target (Facilities, Nations & stadiums) takes the change in its own bytes. Not yet tested in the game.

--mod real-grounds. Each computer club plays at a ground of its real 2005-06 capacity (the game's season; Nexxus's choice, after a first build with 2006-07 figures), rounded to 5,000 (Nexxus's choice, closer to the retail game's few sizes than a figure to the thousand), and clubs of one size are spread over the stadium ids of that size, so grounds of the same size look different (Nexxus's request, 2026-10-30). SRC/grounds.csv holds each club's ground and real capacity in thousands (suggestions, from memory: lower-division and far-away figures are approximate, which the rounding mostly absorbs; Old Trafford 68,200, the Allianz Arena 66,000 and the Olimpico 82,300 are their 2005-06 sizes). initteam.py does the rest:

Looks. Ids 0–101 keep your ladder's capacities, and their models follow those sizes. The ladder's models look small for their numbers at the top (Nexxus's report on the first build: Chelsea, Manchester United and Bayern), so a ground's look follows its size, but the ho04a bowl is Camp Nou's alone, and the retail game's own away grounds (aw01a, aw03a) are used again, so more clubs have a ground of their own (Nexxus's choices, 2026-10-31):

Band	Look	Ids
8,000–35,000	the ladder's models (`ho00`–`ho02a`)	the ladder's, 6–18 a band
20,000, 25,000, 30,000	`aw01a`, their own look	103, 105, 107
40,000, under 40,000 real	the ladder's `ho02a`	the ladder's 40,000 ids
40,000, 45,000	`ho03a` with roof (and lights at 45,000)	102, 104 and 106
40,000, 45,000, 50,000	`aw03a`, their own look (the rival's model)	112, 113, 115
50,000–60,000	`ho03a`	the ladder's 66–95
65,000–90,000	the `aw04b` bowl	108 (65), 109 and 116 (70), 110 (75), 111 (80), 117 (85), 118 (90)
100,000	the `ho04a` bowl	the Special stadium's 96–101

The club-only ids 102–118 take the look of a ground (its BUILD_STADIUM row: model, parts, crowd set, but not its advert boards, which stay the id's own; stadium.copy_rows), its roof and the level of its model: 102 row 69 (ho03a, roof), 104 and 106 rows 75 and 85 (roof and lights), 108, 110 and 117 row 116 and 109 and 111 row 118 (aw04b); 103, 105, 107, 112, 113, 115, 116 and 118 keep their own. A look is copied only from a row the switch leaves alone, so applying it twice changes nothing. GetCapacityFromID adds 500 seats to 103, 105 and 107. Stadium 114 is the rival's ground in every league (TEAM_INIT_DATA table 7, TEAMINIT_FORMAT.md): it keeps its look, boards and 42,000, and no other club is put on it, so the rival's ground stays its own (Nexxus's request); other grounds may use its aw03a model (Nexxus's choice).

Clubs. As many grounds as possible, each close to its real size (Nexxus's request): a ground may take any band within 3,000 of its real capacity, or its nearest band (the bigger on a tie) if none is; under 12,500 the 8,000 grounds. The ladder's own 40,000 grounds, with the smaller ho02a look, only go to grounds really under 40,000 (Highbury, 38,400). Each ground takes the least-used of its choices, nearer bands first on a tie, the grounds with the fewest choices first; clubs naming the same ground (AC Milan and Inter, Lazio and Roma, Juventus and Torino, Bayern and 1860) share one. 90,000 (Azadi; Maracanã's 88,000) goes to 85,000, since no id is left for 90,000–95,000. Camp Nou alone is at 100,000, which is why the switch turns on special-stadium, and it is the only club on the ho04a bowl.
Result. The clubs use 96 of the 118 ids (89 in retail; the rival's 114 apart). By model: ho04a 1 club, aw04b 27, aw03a 10, aw01a 16, ho03a 50 and the smaller ladder models the rest. The mean distance from the real capacity, over the clubs above 12,500, is 1,200 seats, the largest 3,000.

The club records (PLRESOURCESIM.PAC entry 3, +0x06) and STADIUM_DATA go in with their PRELOAD copies, and BUILD_STADIUM.TBB (no copies), only the changed bytes, next to overseas-bases and special-stadium, which write the same files. A PLRESOURCESIM.PAC or STADIUM_DATA.TBB from the editor takes the change in its own bytes (target hooks, as for special-stadium); where its own edit replaced what the switch expects, the editor's value stays. A BUILD_STADIUM.TBB from the editor's Stadiums tab takes the looks row by row: rows whose look the editor changed keep it (STADIUM_DIR.md). Tested in PCSX2 (Nexxus's report, 2026-10-30, the first build): the new grounds work; the biggest looked too small. With the looks by size above, on LMAST-test-stadiums.iso (2026-10-30): "Stadiums are good" (Nexxus's report). With the looks of 2026-10-31 above, on LMAST-test-season3.iso (Nexxus's report, 2026-10-31): only Camp Nou uses the ho04a bowl, San Siro, the Bernabéu and the Westfalenstadion use the aw04b bowl, some smaller clubs play at the retail away grounds, and the rival's ground is still its own.

--mod sponsor-boards. Every stadium's 12 advert boards are drawn afresh from all 205 sponsor logos (Nexxus's request, 2026-10-30: more variety of sponsors at the grounds). A computer club's ground shows the boards of its BUILD_STADIUM row, bytes 116–127 (InitStadium clears the match's own 12 at 0x2136a8); value v is logo v − 1 of SPONSOR_TEXTURE_M, 1–205 (Adverts). The retail rows use 119 of them, 12 different ones a row. The switch draws each row's boards afresh, a fixed random draw per stadium (stadium.sponsor_boards), so every build is the same. How many different sponsors a ground shows grows with its size (Nexxus's request, 2026-10-31): 3 up to 8,000, then one more for each step of 15,000, 20,000, 25,000 ... 50,000, and all 12 boards different from 55,000 (sponsor_count, sizes as real-grounds gives them); each drawn sponsor is on at least one board and the rest repeat them. 202 of the 205 logos appear. The rival's ground (114) keeps its own boards. The adidas board every ground shows (Nexxus's LMAST Realism makes it Nexxus Drako Multimedia) isn't one of the 12: it is ADT_TEXCMN entry 212, which no row names (they use 1–200); the match request's +0x10 moves a row's boards up one and puts it first (0x1cd618). InitStadium sets +0x10 at every computer club's home game except in Spain (the request in a match), so it stays on those grounds (Nexxus's request), and the switch's draws never pick it (value 206; v − 1 is clamped to 205, so any value from 206 on would be adidas too). It writes only those bytes, so it sits beside real-grounds. An edited BUILD_STADIUM.TBB keeps the boards of the rows the editor changed and takes the draws in the rest (STADIUM_DIR.md). Clubs that share a stadium id still share its boards; boards of each club's own would need code in the match set-up. Tested in PCSX2 with the grounds (Nexxus's report, 2026-10-30), and with the counts by size on LMAST-test-season4.iso (Nexxus's report, 2026-10-31): small grounds repeat a few sponsors (3 at the smallest), big grounds show 12 different ones, and the rival's boards are unchanged.

python SRC/plrcommon.py show DAT/PARAM 0.7          # your stadium types
python SRC/initteam.py stadiums DAT/PARAM           # the 119 match stadiums
python SRC/stadium.py info DAT/STADIUM DAT/PARAM    # CONV_INFO_BUILD's ids by level

LMAST Realism on the disc

--mod realism (group graphics, LMASTER full and max; Nexxus's choice, 2026-11-07) puts the logos and match pictures of LMAST Realism, Nexxus's own PCSX2 texture pack, onto the disc, so they show without PCSX2's texture replacement. It is one switch, in full and max and not lite, because it is an improvement and no fix.

python SRC/patch_disc.py patch <retail iso> out.iso --mod realism     --realism <LMAST-Realism>/textures/SLES-54151/replacements --copies --dat DAT

The pack's files are not in this repository. Without --realism the switch looks for a clone beside it, ../LMAST-Realism (Nexxus's suggestion, 2026-11-07), and --realism picks another folder: the pack's repository, its textures folder or the replacements folder itself. The editor's Build disc window has a box for it, filled in with the clone when it is there and changeable. A group or edition that holds the switch, such as lmaster-full, leaves it out when there is no pack and says so; naming --mod realism without one is an error. The pack is read as it stands each time, so a new commit of it only needs a new build. PAL only. It writes seven whole packs (0SYSTEM/EMBLEM_TEXTURE.PAC, FLAG_TEXTURE.PAC, MATCH_TEXTURE.PAC, MINIEMBLEM_TEXTURE.PAC, MINIMATCH_TEXTURE.PAC, SPONSOR_TEXTURE_M.PAC and STADIUM/ADT_TEXCMN.PAC), and --copies carries them to the PRELOAD copies. The work is texmap.py apply (TEXMAP.md):

the pack's pictures whose texture and palette match a disc texture exactly (781 textures) go into their packs with svr.py's writer;
the original crests of the licensed Italian, Spanish and Dutch clubs that the pack doesn't replace get a transparent border of 2 pixels (the picture is scaled down by 4), as Realism's logos have (Nexxus's request and choice, 2026-11-07; the mini-crest sheets already had it);
the kits stay Realism-only (Nexxus, 2026-11-07): a kit part has up to 38 palettes in the pack and the disc keeps one, so they are not put on the disc.

Built, and the files checked: a disc built with the switch equals one built from texmap.py apply's files with --copies. Tested in PCSX2 (Nexxus's report, 2026-11-07, LMAST-test-realism.iso): the disc shows the pack's pictures and works; the mini logos first came out half transparent and were fixed (TEXMAP.md). Not done: the CSP and ACROBATA packs hold 58 more of the pack's pictures (the league logos, the intro), and they have no writer.

Planned

Ideas Nexxus has agreed on (2026-10-27), not started:

Possible, for LMASTER full and max (Nexxus's idea, 2026-11-01), not started:

The Japanese in-game player editor. The Japanese release lets you make an edit player in the game; the PAL release has no editor and imports one from Virtua Pro Football instead (JAPANESE_RELEASE.md). Restoring it would need: where the Japanese editor's screens live (its overlay and SIMPRG.REL/SAVEPRG.REL entry points), whether the PAL code still holds them unreachable or they were cut, and the English text for its screens. Nothing is traced yet.

Still unknown

The in-game checks still to make are in each row's status above, and TESTING.md and TODO.md section 8 hold them as tasks. The open questions behind the candidates:

Whether the 475 unused combination records play sensibly, and how to tell in a match which combination ran (BPB_FORMAT.md).
Whether the match's two readers of the selection policy were meant to read the match policy, and which end of a tactics slider is attacking (PBDATA_FORMAT.md).
Whether the three short EDIT_EMBLEM layer tables lose visible parts in the crest editor (EMBLEM_DIR.md).
Which code picks the advisor's intro and the Game Over texts that the third tier rewrites (The text).
Why Renew Mobile Site was disabled (CLUBEVENT_FORMAT.md).

Not candidates

EVENT/EVENTDATA_TURN.TBB, 0SYSTEM/SCHEDULE.TBB and the CVS/ files are development leftovers. The game has no code that reads them.
Language slot 6 of MES.PAC is an unused near-copy of English.
CSE/GP_PRACTICEICON.CSP is empty in PAL but not in the Japanese release. PAL loads the per-language GP_PRACTICEICON0–6.CSP, which hold the same icons.
The Japanese PLRESOURCESIM.PAC has no free-agent list: PAL added it.

Navigation Menu

Repository navigation

LMASTER_MOD.md

LMASTER_MOD.md

The LMASTER Mod

The switch levels

Status

Improvements

The text

Retail text fixes

Real names

Stadiums

LMAST Realism on the disc

Planned

Still unknown

Not candidates

FilesExpand file tree

LMASTER_MOD.md

Latest commit

History

LMASTER_MOD.md

File metadata and controls

The LMASTER Mod

The switch levels

Status

Improvements

The text

Retail text fixes

Real names

Stadiums

LMAST Realism on the disc

Planned

Still unknown

Not candidates