- Generated by
1.9.8
|
Monopoly — Portage moderne 0.1.0
Portage moderne C/C++23 du Monopoly de 1999
|
Continuation from a58c16f; no gameplay, rule or sequence-clock changes. The existing recovered procedural Blender model remains the authoring source. All screenshots and runtime logs below are generated under build/polish-continuation-20261001/ and are not redistributed licensed assets.
725edd7; evidence build/city-hoist-qualification.json.menu-pass01.jpg exposed an empty background and rough text.menu-pass02.jpg, load-menu-pass02.jpg and file-menu-pass02.jpg retain authored positions, entry animation and hit areas.menu-pass03-aa.jpg, player-select-pass03.jpg and token-picker-pass03.jpg were inspected. The latter still exposes legacy token thumbnails/preview needing qualification.player-select-pass04-contrast-before.jpg, player-select-pass05.jpg, escape-modal-pass05.jpg. All dynamic player names remain original leaves.deed-pass01.jpg shows a real Vermont Avenue purchase: $923 → $823. Full Chance/Community bodies come from the decoded USA idle faces; occluded words are corroborated by associated read-only narration comments. Only idle face owners are replaced; transitions retain original artwork and clocks. Cached deeds, cards and property tiles check context before substitution.A real two-player game ran 10:16:44–10:39:18 UTC (22 min 34 s), with dice, modern car/Dog movement, AI purchases, third-double jail, paying jail, rent, passing GO, Community Chest and saving/reloading. The saved cash state Dog $927 / human $923 was reproduced after restart. Subsequent real turns bought Vermont and States Avenue; another save retained Dog $227 / human $683. Settled 1920×1080 title samples are 59.9–60.1 FPS; this is sampled telemetry, not a continuous minimum-FPS guarantee. Native window captures are DPI-scaled.
The 180-second Windows default-render-endpoint loopback recorded 48 kHz stereo, nonzero signal in all 178 reported intervals, RMS −28.08 dBFS, peak −1.33 dBFS, no nonfinite samples or timestamp errors, and 17 discontinuity flags. The Realtek USB speaker endpoint was unmuted at volume 1; immediately afterwards Monopoly was the only active unmuted session observed. Evidence: audio-real-session/audio-output-qualification.json and windows-output.wav. This establishes software output; physical speaker audibility still requires human confirmation. The old unsupported WAVE error was fixed in a58c16f.
Run the real game with:
Optional assets/modern/presentation/menu_city.png copies the actual GPU probe image build/procedural-play-20261001/scene-06/scene.png. Stage it through build/modern-assets/presentation/menu_city.png. Missing, malformed or oversized previews retain the branded flat backdrop. The UI still uses its original logical coordinates; decorative widescreen palette must not expand input rectangles. Unsupported locale/context/artwork keeps the exact retail assets. French Paris DAT banks remain absent.
Focused qualification uses MenuSkin, World2D GPU, FontRuntime, EuropeanDeed, MeshGPUResources and ModernGltfMesh targets, without repeated full suites. The locked Source tree remains dc0b23ed3aed178721143760e68da91d87a3884d.
token-picker-pass03.jpg against token-picker-pass06.jpg and token-picker-hat-pass06.jpg. F11 was exercised on the real picker at desktop 4096x2160, then returned to 1920x1080. Settled title samples were 59.6-60.0 FPS.pack_token_turntables.py accepts only complete eleven-token/28-frame captures and publishes transactionally. Runtime PNG loading accepts only bounded 768x640 images. One frame pack now decodes asynchronously; original retail frames remain visible while pending. A bounded LRU keeps the live backdrop when old animated derivatives retire. Cold selection still has upload/artwork costs; no continuous minimum-FPS guarantee is claimed. token-picker-pass07-async.jpg records the real result.--dump-textured-board-256 path reads all eight actual licensed USA bitmap files and preserves the old 128 export path. The procedural board was regenerated from the recovered authoring model, whose SHA256 remained unchanged. Vertex/triangle counts and bounds match; material deduplication gives 21 rather than 22 batches. Compare real 1920x1080 GPU images ../board-256-gpu-20261001/before/modern-scene-probe.png and ../board-256-gpu-20261001/after/modern-scene-probe.png: corner/icon text is sharper, with the same city framing and geometry. Fenced 100-frame probe measurements were 114.50 versus 114.96 FPS; these are probe, not full-game FPS.Recreate a token capture with MonopolyModernSceneRenderProbe <asset-root> <shader-root> <existing-output-directory> --token-turntable <slug>, then package all eleven capture directories using pack_token_turntables.py --capture-root <all-captures> --output modern/build/modern-assets/presentation/tokens --asset-root modern/build/modern-assets. The normal app build stages optional presentation files. Missing or malformed packs keep the exact original preview/thumbnail.
Focused MenuSkin, TokenPreview and World2D GPU tests cover strict frame IDs, whole-pack failure, asynchronous selection changes, cache retention, explicit CNK-bound override, nonidentity transforms, invalid-rectangle atomic rejection, and complete retail framebuffer restoration. No full suite was repeated.
board256-trade-pass08.jpg is the real restored game at exactly the previous Tennessee Avenue state: Dog $222, human $683,1920x1080,60.0 FPS. Compare the previous resumed-ui-milestone.jpg: the licensed Community Chest graphic is crisper and Trade now matches the modern HUD. Its strict settled-only exception clips artwork at x130 while preserving the actual x132 input boundary; moving press frames retain retail artwork. Measured real DAT dimensions and context cache fallback have focused tests. The next AI turn visibly constructed houses on the orange group; no rule or timing change was made.
The actual 1920x1080 game now uses matching teal/brass Trade shells, canonical player rails and readable offer controls. A pending purchase deed is presented in the unused right-hand Trade panel instead of hiding instructions. This is limited to the current zero-origin raw purchase bitmap at priority1002; hover priority1003, explicit CNK bounds, sequence clocks and gameplay remain authored. Compare trade-pass09-after.jpg and trade-pass10-deed-fit.jpg in the existing continuation capture directory. Settled real-game title:60.1 FPS.
Score text uses blended3x actual font rasterization within its original logical 184x32 footprint, preserving authoritative cash interpolation. Score icons use the same eleven actual GPU token thumbnails as the picker. Native bitmap update restores original geometry and filtering; unsupported context keeps retail.
Auction panel extents were corrected using actual DAT measurements, including 201x90 and200x92 variants. The flat-stage experiment was rejected after GPU inspection: it removed the floor beneath the animated auctioneer. The complete retail stage is retained until a complete modern replacement exists, while bid plates/player panels remain modern. auction-pass10-stage-restored.jpg shows the real scene at60.1 FPS, before the AI's successful$5 Park Place bid.
Compare status-pass10-before.jpg and status-pass11-panels.jpg: same real Park Place state, Dog$67/human$683. Player panels and calculator frames now match the HUD and retain source alpha, dynamic text and canonical player rails. The old bottom tabs/calculator keys and pending-deed placement in Portfolio remain the next visible work. Title samples near menu transitions still dip; these screenshots do not establish a continuous minimum60 FPS guarantee.
Targeted RuntimeBitmapSurface, ScoreTextPlayback, MenuSkin, IBarSkin and real World2D GPU tests passed. GPU qualification covers purchase relocation, hover priority isolation, nonzero-origin/explicit-bounds fallback, unchanged clocks and exact retail restoration after rasterization failure. No global suite was repeated. The procedural Blender authoring file and Source tree stay unchanged.
status-pass12-tabs-first.jpg exposed oversized clipped captions. After the measured sizing/frame correction, status-pass13-tabs-sized.jpg shows readable Players/Turn tabs and headings, blended 3x dynamic status text and a pending purchase deed inside the Portfolio board viewport. Player balances, panel positions, sequence clocks, history limits and input rectangles are preserved. Only qualified USA presentation opts into the new text path; retail remains the fallback. Targeted MenuSkin and StatsTextPlayback tests passed.
The environment decoder now uses half the available hardware threads, capped at eight, with caller-ordered publication and contained per-asset failures. Injected loaders stay serial unless explicitly enabled. Three real asset runs per setting measured median CPU decode times of 5148.89/1896.62/1409.71 ms for 1/4/8 workers, with peak working sets of 370.7/383.2/383.6 MiB. All 19 ordered geometry/material signatures matched. The actual app measured environment publication at 1389.77 ms and total modern publication at 1646.35 ms, versus 5123.04/5378.17 ms sequentially. These are CPU phase timings, not total launch or GPU fence timings. Focused failure/order/worker-bound tests passed.
The fresh 1920x1080 GPU probe in ../environment-parallel-gpu-20261001/after is byte-identical to the prior board-256 PPM. Its 100 fenced warm frames measured 204.84 FPS; this is an offscreen probe, not a claim about continuous gameplay. sample_game_fps.ps1 records actual window-title telemetry for an explicitly identified process. A 120-second real-turn sample measured median 60.0, p5 59.9 and minimum 52.5 FPS (237 samples); 95.78% reached 59.5 FPS. Cold Portfolio opening still dipped to 22.6 FPS and remains a performance target. F11 entered 4096x2160 desktop fullscreen (59.9 FPS title sample) and restored 1920x1080. status13-fullscreen-real.jpg records that actual fullscreen view.
Physical sound output was captured from the unmuted Realtek default endpoint: 180 seconds, 48 kHz stereo, RMS -28.08 dBFS and peak -1.33 dBFS. This proves software output, not that a person physically heard the speakers. The actual match also continued through AI auctions/building, GO and Community Chest, and saved the progressed state in slot q. Missing French DAT files remain real blockers; none were fabricated. Source and the recovered Blender file remain unchanged. Calculator keys and Bank/Deeds panels remain further visual work.
Actual GPU inspection rejected the first numeric-key mapping: production roots are 1..9,0, but bitmap leaves are 0..9. The corrected exact mapping now skins all 20 idle/pressed states. Selected Bank/Deeds controls also required measured terminal/return frames. Unknown intermediate frames retain authored artwork. calculator-pass14-first.jpg versus calculator-pass15-after.jpg, and bank-pass14-before.jpg versus bank-pass15-after.jpg, record real 1920x1080 comparisons in the continuation directory. Settled title samples were 60 FPS.
deeds-pass14-first.jpg versus deeds-pass15-after.jpg records the removal of the heavy gray grid frame. Canonical deed icons and game-derived values remain separate. Values, Trade names/instructions/cash now use optional blended 3x actual fonts inside their native logical footprints. Cache/queue/clock/font restoration and exact retail fallback tests passed. Deed values cache the actual slot0+size8+weight500 glyph settings, avoiding irrelevant caller refresh. deeds-owner-sort-pass15.jpg confirms a real owner sort with unchanged holdings.
Railroad/utility deed templates had been omitted from full-card repainting. The six measured USA fronts now recover their actual train/bulb/water artwork from original pixels, convert white to coverage on cream, and enlarge only the title above the unchanged rent rows. Unknown IDs, dimensions, colored or invalid art keep the whole original deed. railroad-deed-before.jpg versus railroad-deed-pass15-after.jpg shows the actual Pennsylvania Railroad fix. The real electric/water runtime views still need individual qualification.
MonopolyModernMenuSkinBenchmark <runtime-data-root> loads 47 real DAT fixtures, checks root/leaf ownership against actual decoded production sequences, rejects fallback, and records deterministic pixel hashes plus cold/text/cache CPU costs. All47 passed before raster optimization. The separate isolated shell benchmark proved horizontal-triplet equivalence but measured only about9% gain in Debug; its helper is not yet enabled in the game and no game-FPS gain is claimed.
A further ten-minute real-turn/UI telemetry sample recorded1183 observations: median60.0, p5 54.8, minimum39.3 FPS. This included concurrent targeted builds and diagnostics; it is not an isolated performance guarantee. Actual play continued through built-property rent, doubles/prison, a $50 jail payment and Pennsylvania Railroad; distinct save r preserves Dog$42/human$663 with purchase pending. Existing p/q saves remain. Function-icon tiles and cold-menu costs are still further visible/performance work. Source and authoring Blender unchanged.
The optimized production shell evaluates each aligned horizontal triplet once, with direct RGBA stores and per-output-row gradients/source alpha. Photo and token artwork retain their original path; unaligned widths use scalar rendering. All 47 actual DAT fixtures retain their complete pixel hashes and focused MenuSkin tests pass. Debug cold medians improve from 28.3096 to 12.47055 ms for the 800x225 panel, 13.9123 to 6.1663 ms for the 400x225 panel, and 6.76275 to 2.8208 ms for the calculator panel. These are CPU construction timings, not a game-FPS guarantee. Logs: menu-skin-real-pass15-before-opt.log and menu-skin-real-pass15-after-opt.log.
water-deed-pass16.jpg qualifies Water Works artwork in the actual saved game at 1920x1080, with unchanged printed rents. Physical audibility remains unconfirmed by a person; Windows endpoint loopback proves software output.
The real app now runs the optimized path. calculator-functions-pass16-after.jpg shows all eight function tiles with readable two-line meanings instead of metal buttons and dated symbols. Exact idle/pressed owners, dimensions, alpha and whole-font-failure fallback are tested; input rectangles and existing hover explanations are preserved. A real Escape during the opening movie now reached the main menu without the previously observed Quit prompt; the input filter only suppresses repeated Escape key-down, preserving first press and key-up. No deterministic repeated-input runtime test was performed.
The Electric Company was individually qualified via the separated Portfolio grid hover: electric-deed-pass16-qualified.jpg. A five-minute actual-turn sample after optimization (no concurrent builds) recorded592 title observations: median60.0, p5 58.4, minimum54.3 FPS; 87.5% were at least59.5 FPS. This included menu departure, the railroad purchase, AI movement, Chance sending the human to jail, and failed jail rolls. It does not establish a continuous60 FPS minimum.
Actual before/after captures: deed-detail-panel-pass17-before.jpg and deed-detail-panel-pass17-after.jpg, plus miniatures-grid-pass17-after.jpg. The 56 measured normal/mortgaged Patterns miniatures use authentic property names, immutable canonical costs and group colors in their exact36x42 footprint. Mortgaged cards retain a separate red treatment. Source alpha, hit rectangles, owner bars and full-size deed content remain unchanged. The gray detail frame and black value bars now follow the teal/brass UI, with optional3x blended labels/values at the original logical400x235 footprint. Focused MenuSkin, IBarSkin, FloaterText, DeedValueText and FontRuntime tests pass, including exact native fallback pixels, saved-font invalidation and clock/root stability.
GPU inspection exposed a warm-grid slowdown near49 FPS. Saved font inspection now avoids TTF size/style mutations on cache hits, and miniature cache keys use retained immutable source identity instead of scanning all1512 alpha samples per card per frame. Runtime cadence after these changes remains to be measured. The enlarged fullscreen capture deeds-fullscreen-pass17-values-qualified.jpg confirms180 for St. James; temporary grid diagnostics also matched all28 metrics to their canonical costs. No rule or price correction was necessary.
After the constant-time miniature cache correction, deed-detail-cache-fixed- pass17.jpg and water-deed-cache-fixed-pass17.jpg show the actual revised UI. A90-second fixed-hover sample recorded178 title observations: median58.9, p5 58.1, minimum56.1 FPS in Debug. This improves the observed roughly49 FPS regression but remains below a strict sustained60 target; Release qualification and continued real-game testing remain outstanding.
Actual GPU passes18-23 qualify authored house bevel edges, filtered action captions, unobstructed Options headings and selected music, modern Help chrome, visible Quick Help/credits, and complete help lines after native-metric fitting. Captures are in build/polish-continuation-20261001/: actionbar-pass19-after.jpg, options-text-pass20-after.jpg, help-menu-pass21-after.jpg, credits-layer-pass22-after.jpg, and quick-help-pass23-native-fit-{first,next}.jpg. Matched house closeups are recorded in PROCEDURAL_PLAY.md. The original Help and credits panel priority10 was covered by an older opaque backdrop at the same priority. Only the qualified modern presentation uses foreground11; retail priority, buttons, native metrics, pagination and sequence clocks remain. Blended glyph coverage fits the original measured line extent; text never changes wrapping or hit rectangles. Focused menu, IBar and Options tests pass.
The launcher now prefers an already-built/staged Release and otherwise uses Debug. It copies Debug saves only when Release/savegame does not exist; existing Release saves remain intact. A real launcher invocation selected Release and reloaded save r; isolated migration copied all six save files with identical SHA256 values. Ordinary builds do not run Blender.
Release startup CPU measurements recorded138-152ms for the19 environment items and221-235ms for modern asset publication, versus roughly1.4/1.67s in Debug. These exclude GPU completion and time spent at the menu. A38-minute Release session exercised actual purchases, auction handling, AI turns, jail, GO and save/reload; this included menu inspection and is not an uninterrupted full-game stress test. Its five-minute title sample measured median60.0, p5 59.9 and minimum43.7 FPS; two GPU diagnostics overlapped that sample, so isolated Release qualification follows. Physical human audibility remains unconfirmed; the Windows output loopback qualification is documented above. Isolated Release checks (no builds or other GPU captures) now record178 fixed Electric Company hover samples: median60.0, p5 60.0, minimum59.0 FPS,98.876% at least59.5 FPS. A five-minute real-turn sample records590 observations: median60.0, p5 59.9, minimum57.9,95.763% at least59.5 FPS. The window framebuffer is1920x1080; title telemetry is not individual frame-time measurement and does not establish a strict continuous60 FPS minimum. Actual play continued through AI Chance movement, human failed jail rolls, $50 exit payment and built-property payments. The user also played and confirmed audible game sound from the PC, then confirmed saving before the next qualified relaunch.
Board/Options/Status/Trade labels now use the same optional3x blended action path with22 measured settled normal/disabled fixtures and moving/native fallback. The St. Charles idle Chance card is exactly400x239, so its descriptor now uses 239 instead of240; the other31 measured idle cards retain400x240. Unknown and FaceIn roots keep native assets. The optional native-cursor experiment passes focused GPU fixtures, but the real game still showed the glowing pointer. That unqualified patch is withheld from this milestone. Navigation captures are navigation-pass24-after.jpg and navigation-board-pass24-after.jpg; F11 returned to1920x1080 after fullscreen.
Matched production GPU captures are build/face-in-polish-20261001/st-charles- {t12,t24}-{before,after}.png. The probe starts at0 before advancing to12/24; earlier paper-only capture filenames incorrectly implied advancement while both were clock0. Those older pictures remain a valid matched paper comparison, but not a progressed-animation proof. The corrected comparisons retain the same actual clock, layers, transforms, priority and silhouettes.
A first paper-only pass exposed orange mascot fills on cream. The refinement recolors visible warm paper/art pixels together. The exact33 paper payloads and 1151 measured root/leaf/extent/origin sprite records qualify;60 neutral sprites retain their exact original assets. Alpha, transparent RGB and dark/grayscale ink remain byte-identical. Unknown IDs, shapes, palettes or context keep retail. --card-face-inspect <absolute-runtime-data-root> regenerates the sprite table; --card-face-qualify checks all32 actual roots and all1211 sprite payloads. The FaceIn slot path processes each leaf independently instead of choosing a largest shell that would erase separate animated characters. No clock changes. Focused IBar/World2D CPU and actual GPU checks pass.
The actual cursor diagnostic run reports standard=1, exact bitmap131875, priority65535 and suppressed=1. The optional native pointer path also configures a freshly-created renderer immediately. Temporary diagnostics are removed; focused GPU tests preserve independent hover pixels, native fallback and clocks. The red pointer marker still seen in desktop click captures is not proof that the suppressed game bitmap remains. No claim is made about that desktop marker.
Further shadow trials remain rejected: explicit caster world-depth reconstruction is pixel-identical to baseline; half-texel caster slope bias leaves wall bands. Canonical shader artifacts and source were restored after these diagnostics.
The actual licensed Mono01.hlp export contains55 topics and no bitmap resources. Compiled production styling preserves all94577 original HTML bytes and adds2147 bytes of local CSS and viewport metadata. Unknown/malformed/oversized exports keep their original presentation. Atomic publication precedes browser dispatch.
Actual1920x1080 browser before/after views were inspected; desktop and390x844 navigation remain usable without horizontal overflow. Proof is in build/help-theme-pass27/proof.json; desktop captures are browser-before.jpg and browser-after.jpg. Those desktop images include browser chrome and are scaled by the capture tool; the browser viewport qualification is separate. Edge automatic translation visible in these captures is not game localization.
Focused Windows tests pass for exact HTML preservation and single bounded asynchronous browser dispatch, including cancellation, timeout and errors. The actual in-game button opened the themed full-help page at21:44:50 local time in pass33. Its distinct export-617983628067200-1/contents.html contains 96724 bytes; build/polish-continuation-20261001/pass33-full-help-real-button.png records the inspected browser page. No new global gameplay pause is introduced.
A focused60-frame CPU benchmark with real Arial reduced SelectPlayer sync mean from1.71205ms to0.011325ms after caching unchanged profile name bitmaps. The warm path produces zero name rasterizations and preserves shared font slots, immutable asset ownership and advancing sequence clocks. EnterName remains unchanged (0.03416ms baseline). This is CPU sync time, not loading time or a measured game FPS gain. Logs: pass28-setup-baseline-cpu.log, pass28-setup-cache-tests.log in build/polish-continuation-20261001/. Focused regression tests and independent cache-invalidation review pass; fresh in-game qualification remains pending.
The13 measured profile owner/leaf pairs retain source alpha and fixed native logical corners. Five incoming/outgoing96x110 UAP poses previously remained metal; exact leaf IDs and immutable origin26,25 now qualify. The97x110 idle path remains unchanged. Unknown provenance/shape/origin keeps retail.
The actual seven SelectCity owner/leaf pairs now use the modern teal/brass palette and supersampled caption. Native city-name field, independent arrows, fade alpha, hitboxes and sequence clocks remain unchanged. Matched production GPU images (800x600 isolated UI fixtures, not whole-game1080p qualification) are build/menu-panel-polish-20261001/{profile-in-t1,profile-out-t4, city-idle-t0,city-in-t1,city-out-t1}-{before,after}.png.
The real six-player game exposed lost idle-card illustrations. Chance15's exact native Morris illustration and printed caption are recovered together on cream, with original ink/alpha/placement intact. GPU proof is chance15-native-before.png versus chance15-ink-after.png in that directory; the earlier actual text-only in-game image is build/polish-continuation-20261001/six-player-chance-real.jpg. The32-card native audit sheet showed this artwork-loss issue affected other cards as well; the complete recovery is qualified in pass32 below. No new artwork was invented. Focused menu/IBar CPU tests and full actual-resource World2D GPU tests pass.
All28 actual USA front deeds retain identical output/native extent and FNV pixel hashes after reusing accepted line rasters during wrapping. Font callbacks fall 20127→15280 (-24.08%); aggregate Debug cold CPU11804.7→8962.4ms (-24.08%). This is CPU raster time, not a measured game FPS gain. Logs are build/ibar-deed-wrap-{before,after}.log; font state restoration still qualifies.
The isolated six-player first10-minute window-title sample has1180 FPS values, median60.0, p559.0,94.237% at least59.5; brief first-generation dips reach25.7. This does not establish a continuous60FPS minimum. No GPU/build work overlapped that sample. Later qualification builds/GPU work are outside its scope.
The SelectCity panel now retains its actual native skyline inside the measured rectangle x7..201, y12..79. The three exact F03/F04/F05 leaves supply the silhouette and intermediate edge coverage; only its presentation palette changes. The baked caption and arrows are outside that mask. Native alpha/fades, logical corners, clocks, city-name field and independent arrow owners remain unchanged.
Pass33 raises the modern caption limit from12 to18 native pixels, centered in the measured y88..108 band above the name well at y112..129. Focused tests verify the full18-pixel glyph extent without clipping. Actual seven owner/leaf CPU checks and matched production GPU captures pass; the inspected skyline and caption retain City identity. Captures remain under build/menu-panel-polish-20261001/; the smaller-caption comparison is retained as city-idle-skyline-smalltype-before.png. Logs are pass32-city-skyline-real-tests.log, pass33-city-caption-cpu.log and pass33-city-caption-gpu.log in build/polish-continuation-20261001/.
All32 actual Chance and Community Chest idle cards now retain their native illustrations and printed captions together on the modern paper. CPU checks qualify every actual owner/leaf and exact ink/alpha; GPU checks verify fixed logical placement, original clocks/priorities, cached compatible presentation and exact retail asset/framebuffer fallback outside the supported context. The 400x239 St. Charles exception remains measured; the other31 cards are400x240. No illustration was invented or replaced with a text-only reconstruction.
The complete matched GPU sheets are build/menu-panel-polish-20261001/idle-cards32-native-before.png and idle-cards32-ink-after.png, ordered Chance0..15 then Community0..15. These are isolated production fixtures, not proof that every card occurred in the live six-player session. Logs: pass32-idle-cards-real-tests.log and pass32-all-cards-real-gpu.log in build/polish-continuation-20261001/.
Modern text can reuse at most32 private font faces without changing the active retail face or its saved settings. Focused real-Arial tests qualify the bound, exact glyph pixels, style/color behavior and failure handling. Across all28 actual USA front deeds, aggregate Debug cold CPU time is11804.7ms originally,8962.4ms with accepted wrap-raster reuse, then3055.4ms with private faces. All28 native extents and FNV pixel hashes are identical across the three runs. This is CPU raster cost, not a measured game FPS improvement. Evidence: build/ibar-deed-wrap-before.log, build/ibar-deed-wrap-after.log, build/ibar-deed-private-font-after.log and build/polish-continuation-20261001/pass32-font-tests.log.
The actual SelectPlayer menu has a separate60-second window-title sample with 118 values: median60.0 FPS, p5 59.6, minimum46.3 during transition, and98.305% at least59.5 FPS. It does not establish an uninterrupted60 FPS minimum or individual frame times. The actual menu capture and sample are pass33-select-player-real.png, pass33-select-player-warm-fps.csv and pass33-select-player-warm-fps.summary.log in that qualification directory.
The real one-human/five-AI game ran from20:53 to the save at21:24:42 local time, approximately31 minutes of gameplay. The user confirmed audible physical game sound; software loopback alone was not used to establish that claim. The saved game4 then loaded with the same six balances and positions. Actual F11 fullscreen switched to4096x2160 and returned to1920x1080; fresh captures are pass33-six-player-reloaded.png and pass33-six-player-fullscreen.png in build/polish-continuation-20261001/. The in-game Full Help button also opened the inspected themed browser page, as recorded above.
The completed600-second six-player window-title sample contains1182 FPS values: median60.0, p5 59.0, minimum57.9,82.0643% at least60 and93.0626% at least59.5. Normal live turns included cards, property acquisitions, rents and dice, with brief waits for human input; no builds or other GPU work overlapped the sample. The previously observed cold42..46 FPS dips were not reproduced in this window. Different events make this an unmatched session observation, not a measured performance gain or a continuous60 FPS floor. Window-title telemetry does not measure individual frame times. Evidence is build/polish-continuation-20261001/pass33-six-player-private-font-fps.csv and pass33-six-player-private-font-fps.summary.log.
The actual Community Chest School Tax illustration also appeared during live play, captured in build/polish-continuation-20261001/pass33-community-school-tax-real.png. This qualifies that live card occurrence, while the complete32-card coverage remains the separate production-fixture proof recorded above.
The actual Release process14192 launched at22:01:28 local time. The intro's original400x300 film remains intact; teal/brass presentation surrounds only its outside area, with an honest18-pixel “Press any key or click to skip” caption. The real before capture is pass34-intro-movie-real-before.png; the inspected after is pass34-intro-real-after.png, both in build/polish-continuation-20261001/. An actual Space input removed the shell and reached PickGame. OpeningMoviePresentationTests and OpeningMoviesTests pass; focused build evidence includes intro-presentation-fix-build.log.
In the same loaded game4, player s retained464 funds while the current-player name changed from black to readable cream. Compare pass33-six-player-reloaded.png with pass35-current-player-name-after.png in that qualification directory. The RaceCar player's name was also observed white during actual dice flow. Focused CPU checks pass for unchanged native footprint, alpha, clocks and cache behavior; this presentation change does not change balances or turn logic.
Focused CPU checks and read-only review pass for the camera avoidance helper. The matched production GPU proof uses the canonical board/usa_procedural_runtime.glb, the same20 loaded scene items and11 measured building bounds. Both1920x1080 captures use Direct3D12, studio lighting, shadows and4x MSAA. Raw camera position changes from(384,63.58,102) to(384,130.684,102); both retain the ground-plane aim(243,0,243).
The inspected before view is filled by the hotel wall from inside the building. The after view clears its roof and shows the actual ivory city. Captures are build/building-camera-proof-20261001/building-camera-before.png and building-camera-after.png; building-camera-proof.tsv records assets, bounds, cameras and rendering settings. This is a camera-only matched fixture, separate from live gameplay.
Ordinary close token framing remains present in the real game. Following actual camera transitions, the inspected live samples no longer showed the original wall-filled view. These observations do not establish exhaustive coverage of every camera path. A later local variable rename removes a shadowing warning without changing behavior.
Pass37 qualifies52 actual cash-control CPU cases and the production Slot3 pressed state, preserving native alpha, glows and clocks. Eight matched GPU fixtures cover left/right idle and each side's three pressed controls. They use real DAT assets through CashDialogPlayback at fixed clocks, not simulated transactions. Before/after captures and actual-cash-gpu.tsv are retained in build/trade-cash-polish-20261001/. The modern presentation is guarded to the full USA/enUS, city0, system13 context with no custom board.
Pass38 restores the missing visible Trade partner chooser title. Actual captures pass37-trade-empty-panels-real.png and pass38-trade-partner-real-after.png in build/polish-continuation-20261001/ show the before/after presentation. The explicit UI/board viewport-mapping correction passes LogicalViewport CPU checks and actual chooser DAT qualification. The earlier raw point(1065,497) incorrectly mapped to(443,207) at scale2.4. In actual Release process17168 it maps to(458,276) at scale1.8, selecting partner0(Horse) for playerA1; the real names and property panels appeared. The chooser title9 is readable.
Live cash controls also qualify: clicking50 showed an offered$50 and projected balances414/876; Clear restored464/826. Entering10 then Okay retained an offered $10 and closed the cash dialog. Main Cancel returned to the board with actual funds464/826 unchanged. These checks qualify partner selection, cash editing and cancellation, not an accepted trade or full AI negotiation. Evidence in that qualification directory is trade-input-fixed-real.log, pass38-trade-chooser-fixed-real.png, pass38-trade-partner-selected-real.png, pass38-trade-cash-live-open.png, pass38-trade-cash-live-50.png and pass38-trade-cash-live-okay10.png.
An isolated ten-minute six-player session in Release process17168 produced1183 window-title FPS samples: median60, p5 59, minimum57.9,79.7971% at least60 and 90.6171% at least59.5. Real inspected events included GO paying200, percentage Income Tax paying148(664→516), rent6(516→510), AI purchases, rolls and movement. The owned game4 save for s was updated at23:00:32; original blank p/r saves and game5 for t were preserved. Evidence is build/polish-continuation-20261001/pass38-trade-camera-game-fps.csv and pass38-trade-camera-game-fps-summary.txt.
This window covers the pass38 chooser, camera and name presentation. Later HUD, auction, cash readout, thumbnail and anisotropy changes were not in this sample. It does not establish individual frame times, a continuous60 FPS floor or a matched performance improvement for those later changes.
The accepted matched GPU comparison retains the same1920x1080 camera at yaw8, elevation25,20 objects,520 batches,2251932 triangles and4x MSAA. All15 created samplers use8x anisotropy in the enabled view, with no1x fallback;57846 pixels differ from the disabled view. Both actual captures were inspected: raster GO salary, Chest instructions and mascot are visibly sharper, while vector names remain unchanged. Proof is in build/anisotropy-proof-20261001/0x/modern-scene-probe.png, 8x/modern-scene-probe.png and comparison-manifest.json.
MeshGPUResources defaults to8x for linear-filtered, mipmapped modern textures. MONOPOLY_MODERN_TEXTURE_ANISOTROPY=0 opts out and retains the authored fallback sampler behavior. Actual Release process3264 was built and launched at1080p after this default changed. No new five-minute live FPS sample has been captured; the visual comparison does not establish a game performance gain.
Pass41's six matched consumer GPU captures were inspected and accepted: intro, going-once and sold each have before/after views under build/auction-backdrop-polish-20261001/. The upper stage is now teal; the floor, lamps and animated elements remain unchanged. These production consumer fixtures qualify presentation, not a live-game auction or its bidding outcome.
Pass42's actual left/right cash dialogs were inspected at$0 and$50. In fresh Release process9104, right-side50 followed by Cancel remained in Trade after25 seconds without input; opening the left cash dialog then worked. Captures are build/polish-continuation-20261001/pass42-cancel-traced-immediate.png and pass42-cancel-traced-after25s.png. This qualifies the visible readout and that cancel/return path, not proposal acceptance or a cash gift transaction.
Pass43's actual modern Trade thumbnails were compared with the earlier chooser capture. build/polish-continuation-20261001/pass43-trade-modern-thumbnails-real-after.png records the inspected result; clicking Horse selected the correct partner. This is live chooser presentation/input proof, not a completed trade. The separate pass39 HUD is accepted from the fresh GPU rerun documented below.
The fresh matched build/native-action-polish-20261001/native-actions18-before.png and native-actions18-after.png were inspected and accepted. The selected-pose manifest selected-pose-images.json records all18 changed actual GPU footprints, each at least40x8 pixels; this rerun replaces the earlier narrow-sliver proof. Exact route poses and clocks are retained. The430-entry table passes1056-frame CPU checks. All18 selected native GPU poses have coherent teal chrome while retaining dark ink, white and colored artwork; settled captions are unchanged. This is complete selected-fixture coverage, not live observation of every phase. Earlier actual Release transitions were observed separately.
The accepted matched1080p GPU pair is build/polish-continuation-20261001/camera-travel-row40722/camera-travel-before.png and camera-travel-after.png. Camera eye height changes from78.209 to217.9125, with the same ground aim, forward direction, FOV and20 assets;8x anisotropy and 4x MSAA remain fixed. The inspected after view shows a wider overhead plaza. Later live camera samples qualify corrected sample203 and excluded sample212. The exact row40722 fixture has not been replicated in real gameplay; these samples do not establish exhaustive camera-path coverage.
The accepted matched GPU pair is build/card-out-polish-20261001/services25-options-out-before.png and services25-options-out-after.png. Ink, alpha, clock and placement remain the same. Thirty-one actual Out payloads are byte-identical to their idle payloads; the different Chance12 payload remains native. The actual Options capture build/polish-continuation-20261001/pass44-card-on-options-real-after.png contains no visible card because it had already ended. It is not matched live Out proof.
Actual Release process17604 completed an authentic trade: the human offered$300 for Horse's Tennessee Avenue and the offer was accepted. The inspected trade-tennessee300-ready-real.png shows both offers and native bottom balances 535/618; trade-tennessee300-accepted-real.png shows Main balances235/918 and Tennessee added to the human hand. Both are in build/polish-continuation-20261001/. This qualifies that real accepted trade, beyond the earlier edit/cancel checks; it does not establish every negotiation path.
The completed ten-minute session in process17604 includes default8x anisotropy, UI themes39–45 and the camera correction, with no concurrent build or other GPU work. Its1181 window-title FPS samples have median60, p5 59.1, minimum57.1, 82.4725% at least60 and93.3954% at least59.5. Evidence is build/polish-continuation-20261001/passes39-45-game-fps.csv and passes39-45-game-fps-summary.txt. This is session telemetry, not individual frame-time proof, a continuous60 FPS minimum or a matched performance comparison; it contains no evidence for future passes46–47.
Actual events included acknowledging Chest25, the accepted Tennessee trade, two AI rounds, human rent14 received(balance249) then B&O rent100 paid(balance149). The private game5/t save was updated on October2 at00:12:20; original slots1/p, 2/blank and3/r were unchanged, while4/s retained its pending card for diagnosis. Actual F11 fullscreen was captured as pass45-f11-fullscreen-real.png in that qualification directory, then returned to the1920x1080 window. Live camera proof qualifies samples203(corrected) and212(excluded), without an exact real-game replay of row40722.
Actual Release process15896 loaded the private game5/t save with the human at B&O and149 funds. The inspected build/polish-continuation-20261001/pass46-trade-instructions-real-after.png retains the authentic LANG body, now clear regular cream text with no underlined heading and the full body visible. Compare the earlier panel in pass42-cancel-traced-after25s.png. Actual LANG/Arial CPU checks fit eight lines with bottom438/675; pass46-panel-real-tests.log and general IBar/text tests pass.
The central Moneybag thumbnail is now black and matches the actual GLB. Inspected captures are pass47-active-token-modern-real-after.png and pass47-active-token-fullscreen-real-after.png in that qualification directory. The earlier same-B&O/amount fullscreen view pass45-f11-fullscreen-real.png shows the old gray central icon. This deliberately replaces the central icon's visual spin with the existing static thumbnail at255; authored CNK clocks and transforms remain unchanged, and the3D token animation still runs. CPU qualification covers all11 roots and328 UAP frames in pass47-current-token-real-tests.log.
These are accepted actual UI renders. The completed passes39–45 ten-minute FPS sample predates both changes; no new ten-minute FPS result is claimed for46–47.
A separate completed60-second warm sample for passes46–47 has118 window-title FPS values: median60, p5 59.8, minimum57.9,81.3559% at least60 and96.6102% at least59.5. It includes approximately15 seconds in the chooser, followed by Main and AI turns, with no concurrent build or other GPU work. Evidence is build/polish-continuation-20261001/passes46-47-warm-fps.csv and passes46-47-warm-fps-summary.txt. This short session sample does not measure every frame, establish a continuous60 FPS minimum or replace a ten-minute test.
Focused CPU qualification covers32 cards in both entry cases. Actual root1005 and leaf1 provenance remain separate from draw order; left contract hover never hides the right-panel card. The isolated800x600 production-consumer GPU fixture passes for Main→Trade at y0 and Trade-start at y136. All six captures were inspected under build/card-trade-polish-20261002/: each entry has before, after and hover views. The card moves into a200-wide upper-right panel, exposing the full chooser; right deed hover makes the card yield, then it restores. Returning to Main restores the native400x240 extent without inventing a reward or changing the card clock transition.
Evidence is build/pass48-idle-trade-cpu.log, pass48-ibar-regular.log, pass48-card-trade-gpu.log and pass48-world2d-regular.log. These focused CPU/GPU checks pass. They are isolated production fixtures; the separate actual 1080p qualification follows below.
Actual Release process2712 at1920x1080 loaded private game4/s with510 cash and a pending Services25 card. Inspected captures under build/polish-continuation-20261001/ qualify the live path: pass48-main-pending-card-real.png, pass48-trade-card-clear-real-after.png (clear chooser, compared with pass44-trade-card-overlap-real.png), and pass48-horse-selected-pending-card-real.png (Horse selected while staying in Trade). pass48-pending-card-deed-hover-real.png shows readable B&O; pass48-pending-card-restored-real.png shows the card restored on pointer exit. Cancel returns to Main with the native card and510 cash in pass48-main-card-restored-real.png. Actual Space acknowledgement awards exactly 25, reaching535 in pass48-card-acknowledged535-real.png. An automatic Main return was not reproduced; no independent routing-bug fix is claimed.
pass48-real-profile.log records startup CPU phases: idle-token warming20.8806ms, environment136.505ms, total assets189.133ms, upload preparation150.307ms and first presentation32.3654ms. These are CPU elapsed measurements without GPU-completion proof. The approximately44.7 seconds since session start includes intro, menus and user-input delay; it is not loading time. This evidence does not demonstrate a clear avoidable startup bottleneck.
The completed600-second pass48 sample contains1182 window-title FPS values: median60, p5 59.8, minimum56,80.9645% at least60 and96.2775% at least59.5. No concurrent build or other GPU work overlapped it. It includes card waiting and menus as well as three human rolls and AI rounds, not600 seconds of continuous movement. Evidence is pass48-real-game-fps.csv and persisted pass48-real-game-fps-summary.json in build/polish-continuation-20261001/. Title telemetry does not measure every frame or establish a continuous60 FPS minimum.
Inspected actual events were Services25 paying510→535, AI-round rent income 535→545, Chance22 loan150 paying545→695, Water Works28 rent24 paid to Hat 695→671, then AI rent income10→681 and a roll11 to the human-owned Boardwalk39 with681 cash. pass48-chance-loan150-real.png exposes unreadable baked native card artwork, a separate presentation issue rather than a failed rule award; pass48-loan150-awarded695-real.png and pass48-water-rent24-real.png retain transaction proof.
Original slots1/p,2/blank and3/r retained their timestamps, and game4's pending Services card save remained unchanged. Private game5/t was updated on October2 at00:57:06 with Boardwalk/681; its prior files were backed up as game5-before-pass48.msv and .sgd in the proof directory. The owned process2712 was stopped after path verification for the next GPU qualification. No pass49 result is included in that pass48 evidence.
Debug regular tests and actual11-root/328-frame qualification pass with identical RGBA and fixed rectangle. Eight matched1920x1080 CurrentPlayerPlayback GPU poses cover RaceCar, TopHat, Horse and Moneybag against a faithful68bc35b baseline, using the same PNG fitting and actual clocks. All16 original before/after frames were inspected and accepted for stable anchor/size and Horse/Roll clearance. comparison-crops2x.png is an inspection sheet made only from original GPU-frame crops enlarged2x by nearest sampling; the full PNGs remain original captures. The manifest is build/polish-continuation-20261001/pass49-qualification.json.
The fixed modern rectangle intentionally differs from per-phase authored corners. Root node, leaf, source, authored matrix, clock and native fallback remain preserved. The earlier static-thumbnail255 replacement of icon spin remains the pass47 choice. Release builds pass in build/pass49-release-build.log.
Actual process5552 loaded private game5/t at Boardwalk with681 cash. Inspected live captures are pass49-moneybag-stable-real-after.png, pass49-race-stable-real-after-a.png after Done/AI progression, and pass49-hat-stable-real-after.png, in the continuation proof directory. The actual RaceCar before is pass49-central-race-small-real-before.png; differing live backgrounds make the matched GPU poses the precise before/after comparison. The completed60-second warm sample has118 title-telemetry samples, median60, p5/minimum59,71.1864% at least60 and88.1356% at least59.5. Evidence is pass49-warm-fps.csv and pass49-warm-fps-summary.txt in that directory. It includes actual AI turns without concurrent build or other GPU tests; it does not prove individual frame times or a continuous60 FPS minimum. The completed pass48 ten-minute sample predates this rectangle change.
Actual pass49 F11 fullscreen is captured in build/polish-continuation-20261001/pass49-f11-fullscreen-real.png, followed by a return to the1080p window. This check occurred outside the warm FPS sample.
The accepted isolated800x600 actual CardPlayback GPU comparison is build/loan-card-polish-20261002/loan150-before.png and loan150-after.png, with native-fallback.png for the fallback. The before already has cream paper but occludes the native caption. The after centers the full authentic300x180 image above a16-pixel footer containing the actual two-line building-loan/Collect $150 text. Its logical footprint remains400x240.
This single trial strictly qualifies root5002B/leaf50985, opaque actual UAP flags5, origin0 and canonical ModernDrawCardText[3]. Invalid font, fit, source, alpha or context returns the original native presentation. Clocks, provenance and logical corners remain preserved; no other cards or Out/FaceIn states are modified. Pass48 reflow/actual hover and pass49 all328-frame regressions also pass.
Logs are build/pass50-focused-build.log, pass50-cpu-rebuild.log, pass50-loan-cpu.log, pass50-loan-gpu.log, pass50-ibar-regular.log, pass50-pass48-cpu.log, pass50-pass49-cpu.log, pass50-pass48-gpu.log and pass50-world2d-regular.log. Release passes50–51 now build successfully in build/passes50-51-release-build.log; actual process34060 runs1920x1080. The loan150 caption has not yet been encountered in the full application. Its accepted GPU/real-font fixture proof remains separate from pending live-caption proof. A replay of private game4/s uses genuine rolls, with native B&O rent100 observed; no forced state is used to encounter the loan card.
The actual before/after comparison uses the same private game5/t and balances 1252/755/1052/1580/784/1020. Inspected captures in build/polish-continuation-20261001/ are pass51-portfolio-amounts-real-before.png and pass51-portfolio-amounts-real-after.png; pass51-portfolio-cash-sorted-real.png and pass51-portfolio-reentry-real.png confirm values remain intact through sorting and reentry. Players text now uses priority502 above coins501 and below deeds510. Coordinates, content and other Stats modes remain unchanged. Eight focused CPU tests pass, with pass51-stats-text-build.log and pass51-stats-text-cpu.log in that directory. This is an explicit presentation-layer correction, not a balance or rule change.
Actual process 34060 at 1920x1080 encountered the Park Place auction. The inspected build/polish-continuation-20261001/pass52-park-place-auction-offer-real.png shows human funds 445, Ship funds 1682 and Park Place unowned. pass41-auction-park-place-live.png shows the real auction at bid 0 with the teal backdrop, lamps and Pennybags. The AI auction then resolved automatically: pass41-auction-return-live.png shows Ship funds 1202, and pass41-auction-battleship-owner1202-live.png confirms Ship owns Park Place in Portfolio. The observed deduction is 480 (1682→1202), with ownership confirmed.
A human click targeted the wrong column, so this session does not qualify human bidding. It establishes the live backdrop and that AI auction result, not every auction interaction or endgame behavior.
Accepted CPU/GPU qualification covers the five canonical idle cards Loan 3 and Community 20/22/23/25, plus their exact Out states, using measured source headers and payloads. The four new Community before/after pairs and two Out comparisons were inspected: authentic one- or two-line text is now readable, with the complete 300x180 illustration inside the same 400x240 logical footprint. Exact companions, native fallback and clocks remain preserved.
Original captures are under build/readable-card-polish-20261002/: loan150, community-bank-error, community-go-jail, community-jail-free and community-tax-refund each retain before/after, native-fallback and Out captures. The build/pass52-* logs pass for the focused build, readable-card CPU/GPU, pass48 CPU/GPU, pass49's 328 CPU frames and regular IBar/World2D checks.
Release passes 52–53 now build successfully in build/passes52-53-release-build.log. The four new Community cards have not yet been encountered in the full application, so these accepted fixtures do not establish live card-caption proof. Pass53 has separate live qualification below.
Actual process 24436 runs at 1920x1080. Inspected captures in build/polish-continuation-20261001/ are pass53-portfolio-pending-card-real-before.png, pass53-portfolio-pending-card-real-after.png and pass53-portfolio-cash-matched-real-after.png; the matched balances and sort confirm that columns and buttons are freed without changing displayed funds. pass53-calculator-description-clear-real.png shows NetWorth hover hiding the card, while pass53-card-restored-after-help-real.png shows empty hover restoring the card with 510 cash. No CardSeen acknowledgement or fund change occurred in these five captures.
Deed-picker and Future-popup guards have not yet been exercised live. A later Deeds click returned to Main unexpectedly; NotifyTradeFinished is only a candidate, with cause unproven. These captures do not establish a routing fix. The separate warm sample is distinct from the completed ten-minute session below.
Completed pass53 warm telemetry: 120 samples over 60 seconds, median 60 FPS, p5 58.9 and minimum 58; 73.3333% at least 60 and 83.3333% at least 59.5. Evidence: build/polish-continuation-20261001/pass53-warm-fps.csv and pass53-warm-fps-summary.txt. No concurrent build or other GPU test ran. This title telemetry covers Portfolio interactions and return to Main; it does not establish per-frame timings or a continuous 60 FPS minimum.
The completed 600-second live session in process 24436 has 1181 window-title FPS samples: median 60, p5 59, minimum 55.5, 81.7104% at least 60 and 92.2100% at least 59.5. No other GPU work or build overlapped it. Evidence is build/polish-continuation-20261001/pass53-long-real-fps.csv and pass53-long-real-fps-summary.txt. It does not establish individual frame times, a continuous 60 FPS minimum or endgame coverage.
Actual Services25 acknowledgement paid 510→535. After AI rounds and a roll of 6 to unowned Indiana23, a real human auction bid of 425 won: funds became 110 and Indiana ownership was confirmed. Inspected captures are pass53-indiana-auction-offer-real.png, pass53-auction-human-bid425-live.png, pass53-auction-human-won-balance110-live.png and pass53-auction-human-indiana-owner110-live.png in that qualification directory. This qualifies a human bid and its result, beyond the earlier AI-only auction.
Mortgage of St. Charles added 70 to reach 180; unmortgage deducted 77 to reach
pass53-st-charles-mortgaged180-live.png and pass53-st-charles-unmortgaged103-live.png retain that proof. pass53-deeds-category-real.png shows normal Deeds navigation working without a pending card; it does not resolve the earlier pending-card routing question. Private game5/t was saved with 103 funds at 02:11:53 on October 2. Its prior files are backed up as game5-before-passes54-55 in the continuation directory.Pass54's actual AuctionPlayback CPU qualification covers 196 poses and all 11 tokens. Four matched 1920x1080 GPU pairs are retained under build/polish-continuation-20261001/pass54-auction-gpu/: auction-0-tick0, auction-0-tick40, auction-1-tick0 and auction-1-tick40, each with before/after PNGs. Two pairs were visually inspected and accepted. This is production-fixture proof; the subsequent full-game encounter is recorded below.
Pass55 qualifies all eight authentic LANG descriptions with real Arial CPU checks. Actual process 14816 at 1920x1080 produced the inspected build/polish-continuation-20261001/pass55-networth-description-real-after.png. Compared with pass53-calculator-description-clear-real.png, the cream text on teal is substantially more readable, with the same 510 funds and the pending card hidden during hover. This establishes the actual NetWorth description view; the other seven descriptions have CPU proof, not separate live captures. Release passes54–55 build successfully in build/passes54-55-release-build.log.
Two additional pass55 descriptions were inspected live in build/polish-continuation-20261001/pass55-future-description-real-after.png and pass55-general-income-description-real-after.png, covering FutureToYou and GeneralIncome beyond the earlier NetWorth view.
Actual process 28272 at 1920x1080 produced the accepted build/polish-continuation-20261001/pass56-player-cash-real-after.png. Compared with pass55-networth-description-real-after.png, the same six balances 618/510/1334/1710/992/1208 remain intact. Larger regular cream text at size 12 fits the strip without obscuring deeds. CPU qualification covers six cash cases, including int64 extremes and complete native fallback, in pass56-cash-real-cpu.log and pass56-stats-regular.log. The Release build passes in build/pass56-trace-release-build.log; its temporary tracing is diagnostic work, not a committable part of this presentation change.
Actual Cash sort, Deeds and Bank work with Services25 pending, no CardSeen and unchanged 510 cash, as captured in pass56-cash-sort-trace-real.png, pending-card-deeds-trace-real.png and pending-card-bank-trace-real.png in that qualification directory. The earlier unexpected Main return was not reproduced; these observations do not establish that its cause has been resolved.
Pass57's actual PlayerCashPlayback GPU fixture covers six players at 1920x1080. The inspected build/polish-continuation-20261001/pass57-coin-gpu/player-cash-before.png and player-cash-after.png are lossless conversions of the original BMPs; unsupported-context fallback is framebuffer byte-exact. Actual game process 32860 also produced pass57-modern-coins-real-after.png, compared with pass56-player-cash-real-after.png at the same six balances 618/510/1334/1710/992/1208. Both fixture and live coin presentation were accepted.
Pass58's inspected pass58-deed-picker-real-after.png shows all 28 deeds, compared with the black picker in pass53-deed-picker-pending-card-real.png. pass58-deed-picker-hover-boardwalk-real.png shows readable Boardwalk in the upper-right without Services-card occlusion. Selecting Boardwalk closes the picker with result 0 in pass58-boardwalk-future-result-real.png; the human already owns Boardwalk, so this is selection/result presentation proof, not deep economic-calculation qualification. CLEAR was observed restoring the pending card and 510 cash in two unsaved views; no capture file is claimed for those. The before/after category backgrounds differ, so this comparison qualifies the foreground picker rather than a matched whole framebuffer.
Pass58 CPU logs qualify all 28 actually decoded deeds and open/close/hover behavior. Release passes57–58 build successfully in build/passes57-58-trace-release-build.log. Captures above are in build/polish-continuation-20261001/ unless otherwise specified.
The diagnostic trace proves an autonomous Dog5↔Race2 trade at AI level 2, ending with Finish A2/B0 and changing Portfolio→Trade→Main without inputs. Evidence is passes57-58-input-trace-real.log and pass59-before-main-after-autonomous-trade-real.png in that directory. This explains the observed autonomous transition; subsequent pass59 qualification is recorded below with separate CPU and live-observation limits.
Focused CPU tests replay the actual duplicate Started and terminal A=0/1/2 sequence from the trace: aiProposing is false, both participants are actual AI level 2, and the origin is Portfolio. Native flag false, human participants, invalid participants, current-menu cases and counter -1 preserve default behavior. All focused root CPU tests pass, including the TCP repeat recorded in build/polish-continuation-20261001/pass59-game-session-root-repeat.log; a child socket failure was not reproduced in the root run. Release builds pass in build/pass59-release-build.log.
Actual process 8680 at 1920x1080 produced pass59-portfolio-before-autonomous-trade-real.png and later pass59-portfolio-restored-after-autonomous-trade-real.png after calculator-state reset, in the continuation directory. Portfolio retains 510 funds and the pending card. These views are consistent with restoration; the brief intervening Trade view was not captured, and no new runtime event trace was added. The exact duplicate-event behavior is independently replayed by the CPU tests.
Additional pass58 live FutureOther selection uses Baltic→RaceCar: pass58-other-player-picker-real.png and pass58-baltic-other-player-result-real.png show the player picker and result 0. This uses production futureValue logic, not a stub; unowned rent 0 is authentic. The Bank/Players cases still retain 510 funds without CardSeen acknowledgement.
The completed 45-minute process8680 FPS test comprises four 600-second segments and one 300-second segment under pass59-active45min-segment1..5 CSVs. The combined result and live-session limits are recorded below.
Pass54 was subsequently encountered in the actual full application during the process8680 long session. The inspected build/polish-continuation-20261001/pass54-auction-modern-thumbnails-fullgame-real.png shows six stable PNG portraits with balanced names and money. At unowned Oriental, the human had 731 funds and requested Auction at raw point (667,512). The actual current bid was 500 from Battleship, with human bid 0. No winner had yet been observed, so this qualifies the live auction presentation, not its outcome.
That session also confirmed GO funds 735 followed by Baltic rent 4 paid to Horse, leaving human funds 731 and Horse funds 426, captured in pass59-longsession-baltic-rent731-real.png in the continuation directory. The completed 45-minute FPS result follows below.
Process 8680 remained alive after the completed 2700-second sample at 1920x1080. The 5319 window-title samples span October 2, 01:16:33.634–02:01:33.557 UTC. Median FPS is 60, p5 is 59, minimum is 54.4; 79.7894% are at least 60 and 92.1602% at least 59.5. All five segment medians are 60, with minima 57.1/55.6/54.4/57.8/55.9. Evidence is build/polish-continuation-20261001/pass59-active45min-combined-summary.json and the five segment CSVs.
This actual six-player human/AI session includes menus, Portfolio, calculator, Trade, auction and animations, not 2700 seconds of continuous movement. Actual transactions were Chance GO 535→735, Baltic rent 4→731, Electric rent 24→707, receipts 50→757, Tennessee rent 14→743, receipts 50→793, Ventnor rent 22→771 and Park Place rent 35→736. Private game5/t was saved with 736 funds at 04:02:14 local time; all six original files in slots 1–3 retain identical SHA hashes.
This is title telemetry, not individual frame-time proof, a continuous 60 FPS minimum or endgame qualification. The unfinished passes60–62 have no GPU qualification yet and are not counted in this result.
The accepted same-private-game5 comparison is build/polish-continuation-20261001/pass60-portfolio-gutters-real-before.png and passes60-63-portfolio-fixed-save-real-after.png. Black foot gutters become teal and the frame is clear. Qualification requires the actual board to be published; rejection/native fallback and movie exclusion remain guarded.
The after view retains the same six balances and holdings. Eight Horse cells, three human cells, five RaceCar cells and four Dog cells are readable. TopHat's Jail-free card keeps exact native fallback. Actual Cash sorting is captured in pass63-portfolio-cash-sort-real.png. BSSM Mortgage selection of St. Charles in column 4 pays 70 (736→806); Unmortgage deducts 77 (806→729). Inspected captures are pass63-mortgage-grid-real.png, pass63-grid-mortgage-st-charles806-real.png, pass63-unmortgage-grid-real.png and pass63-grid-unmortgage-st-charles729-real.png in that qualification directory.
Focused StatsUI, PlayerPlayback, Trade, Routing and Skin CPU checks and clean Release builds pass in build/passes60-63*.log. Engine routes the existing modernTradeText setting through the existing setter; temporary tracing has been removed. Passes61–62 now have separate actual qualification below. Passes39–63 total 25 qualified milestones in this continuation.
Three actual intro views show the copyright, Artech and real MIntro Horse: opening-trademark-real-before.png, intro-trace-visible-logo-real.png and intro-trace-mintro-visible-real.png in the continuation directory. No intro bug was reproduced, so these observations are not an improvement milestone. The nine-second authored splash is not demonstrated to be a loading bottleneck.
Pass61's GPU test passes in build/polish-continuation-20261001/pass61-trade-coins-gpu.log, covering four actual coins and four text surfaces with exact fallback. The original 1920x1080 trade-cash-before/after GPU PNGs were subsequently inspected and accepted: four clean modern coins retain the same dollar values and clocks, with framebuffer byte-exact native fallback. The accepted full-game same-private-game5 comparison is passes61-62-fixed-save-trade-real-before.png and passes61-62-fixed-save-trade-real-after.png in the continuation directory. Horse's eight holdings now form a 4x2 grid; the human retains three holdings. pass61-three-trade-coins-real-after.png shows only three live coins, not all four fixture coins.
Actual Boardwalk selection at (257,325) moves its offer card to (374,349), then returns it to (257,325). Captures pass62-boardwalk-offer-grid-real.png and pass62-boardwalk-return-grid-real.png retain that proof. Cash remains 736/1040, and Propose was never clicked, so this qualifies draft editing/return rather than an accepted trade. The shared grid hit mapping also supports the actual pass63 St. Charles mortgage 736→806 and unmortgage 806→729 recorded above. Private game5 was subsequently restored to 736 funds.
The previously recorded completed 45-minute telemetry and F11 checks retain their original scope; they do not measure performance of these later grid and coin changes. Intro trademark, logos and movie were visible, with no reproduced intro defect and temporary tracing removed.
A separate final 120-second actual clean-Release sample in process 2888 includes initial Load, Main, Portfolio and BSSM Mortgage/Unmortgage, not only steady warm play. Its 237 window-title samples have median 60, p5 60, minimum 44.8, 96.6245% at least 60 and 98.3122% at least 59.5. Evidence is build/polish-continuation-20261001/passes60-63-final-fps.log and its CSV. This is not GPU frame-time measurement or a strict 60 FPS minimum. Temporary tracing is absent; the four Engine comments and CRLF were restored without behavior changes.
Passes64–65 now have accepted full-application proof, bringing passes39–65 to 27 qualified milestones. Passes66–67 remain pending actual qualification and are not counted. The earlier FPS samples retain their original scope.
Pass64 repairs actual font fitting for an authentic style2 Mortgaged badge. All 28 actual retail pairs and regular CPU checks now pass: the earlier font failure was repaired using measured 96-DPI glyph bounds. In actual process20872, mortgage changes 736→806. The same Cash-sort St. Charles comparison is pass64-mortgage-face-cash-sort-real-before.png and its after PNG in build/polish-continuation-20261001/: the authentic red Mortgaged badge is clear, with name and price retained. Unmortgage changes 806→729 in pass64-unmortgage729-real-after.png; pass64-restored-normal-title-real-after.png shows the restored normal face without the badge.
Pass65 qualifies the shared auxiliary header at actual 43x28/42x26 extents, flags6 and measured origins, plus the maximum six-cell Jail grid. Y394/424 placement preserves native pixels, IDs, priority and clock. Focused StatsUI, Auxiliary, actual retail headers and regular CPU checks pass. Actual same-private- game5 Turn-order and Cash-sort views retain the balances (including 736/1040). The inspected pass65-aux-fan-real-before.png and pass65-aux-grid-real-after.png plus the Cash-sort after view show all six TopHat deeds in a visible 3x2 grid. The earlier five-deed observation missed a sixth occluded deed. The native Jail- Free item is separated at y+9; other players' grids and hit mapping are unchanged. These are accepted full-application views in Release64–67, process20872.
Pass66 fits the strict settled top-down preset (verticalSquare0) to the actual viewport. build/polish-continuation-20261001/pass66-top-down-settled-real-before.png shows the cropped board before. Forward/up vectors and projection planes remain preserved, without additional overlay reserve or controller-clock changes. Two focused CPU tests and review pass; matched GPU proof remains pending.
Pass67 gives selected Options On/Off controls brass with dark ink, guarded to four authored poses and their alpha. Its CPU checks now pass for four roots, 20 poses and eight mappings, plus regular tests. Actual runtime qualification remains pending.
1.9.8