Dated source reference. Current release status: handoff state. Full original relative paths are in the developer archive.

API verification — private broker55 / native candidate63

October20 team delivery71 publishes the unchanged accepted API70e/native70d with new bundle-specific checks:634 API tests/126 schemas,25 TypeScript tests,13 example/scenario cases and ten installer groups. Full external ZIP hash and native Paseo desktop/mobile pages verify. Exact-DLL original/college live acceptance remains inherited; packaging starts no game. Historical verification entries below retain their original source scope.

Current work70 retains API0.2.0.dev55/HTTPv1 and existing transports. API70e passes634 Windows tests/126 schemas; effective college overtime accepts matching frozen authority while invalid/missing/mismatched authority still refuses. Unchanged native70d/DLL5cf76196 passes original70liveo/API70c natural9–6 and college70liveq/API70e natural6–3, with strict native C metadata, selected IFF reads, forced-two-point classification and independent immutable results/repeated ACK. Both close/restore98 baseline files/DLL50/INI; final preservation passes and custody is released. Changed free-throw counts retain instruction/headless proof, with only baseline2→2 queried in original and no college FT query. Historical result compatibility and inherited TypeScript compilation retain their scope. Closeout.

Developer tools64 are accepted headlessly on exact runtime source devkit64f, archive95adee1517752258ee23f73e9034f3260bd264b14419cfd5649ffe0e4228fd84. 628 Windows tests pass with zero skips. Source/schema verification validates125 schemas/50 operations,88 unchanged native inputs and eight unchanged native/request protocol inputs against native63i. Strict TypeScript/25 runtime cases retain their compiled64c identity; six matching compilation inputs prove the current client is identical. Independent Node HTTP acceptance passes eight scenarios/63 checks/18 actual response-schema validations, including real SSE cancellation, uploads, SQLite persistence and lost responses. The game inputs/outcomes are explicitly simulated; no native game was started or installed. Work63's exact-DLL live acceptance remains pending. Developer usage.

Extension-platform work63 is the private API54 candidate: frozen timing/reset/college-bonus configuration, FT/classification transactions, pre-reveal copied native metadata and immutable match assets.595 Windows API tests/122 schemas pass; the full native DLL builds with warnings-as-errors. Headless/native-instruction evidence is separate from final game acceptance, which is waiting for the user’s explicit Dell-safe notification. The stable game, public packages and source62 live baseline remain preserved.

Current rule-state and SDK53 checkpoint

Work62 adds per-field legal ball/control/restart/FT interpretation, ten named record kinds with scoped actor roles, a separate rule transport, trusted native C modules, preplay frozen constants and per-step permission decisions. It does not certify the entire NBA/NCAA rulebooks or every SDK enum. SDK guide documents ownership, callback context, required/optional faults and unsupported decisions. HTTPv1/mainABI26/eventABI4/profile4/court3 remain unchanged.

Exact API62zzr source passes561 Windows API tests and all110 schemas/historical-result compatibility. Final SDK62final passes1,018 compiled checks with an independently loaded C DLL; final helpers62final2 pass313 production-adapter/original-instruction checks plus six independent exception/invalid/slow × optional/required processes. The latter use declared stubs and no game/desktop. They cover30 permissions,13 control cases, three actual FT award branches before phase change, composed college early returns/reset/AP publication, startup refusal serialization and abandoned-step cache recovery. Boundary62z separately supplies606 compiled preservation/preflight/return tests. Kicked-ball/double-dribble handlers contribute128 captured-instruction guard/attribution cases; upstream detection is not thereby certified.

Native full-result62zzb verifies a separate C module with18/9 clocks, phase4 binding, natural3–0 college finish, reconciled totals, two independent consumers and repeat ACK; readback62zzb reopens the persisted real-game database and delivers to a third consumer without a game or generated snapshot. Optional-fault62zzo finishes2–4 with frozen values retained and explicitly degraded immutable results; required-fault62zzt suspends/stops the owned child, marks review_required and emits no successful result. Refusal62zzu returns a typed initialization failure in0.459s without menu input or an unattended modal. These runs retain exact DLL scopes in the ledger; later FT/restart/AP/gap corrections are independently checked on the final inputs.

Final non-probe adapter62final2 builds MSVC /W4 /WX: source archive48eb706d3c9835e9ac370bb5c3639a0674a8f1fc5883d86f2c08142ea1af312f, DLLcd992875d5f55b6d3a759e512386445043e9fa5486b6791af035d87367a16f2c. Exact-final62finala passes original control/live/alive/NBA-possession with native24-second clock;62finalb naturally finishes college0–2 with the independent module, frozen18/9, zero module/adapter faults, reconciled totals and identical independently ACKed results. Aggregate acceptance, source comparison and preservation record final identities/restoration. Public/stable installations and historical packages are separate.

Previous metadata connection checkpoint

Roster metadata52 connects optional roster metadata to any external client. Private API0.2.0.dev52/HTTPv1 keeps ABI26/event4/profile4/court3 and all67 native inputs unchanged. Its539-test/98-schema source now has real native61a acceptance: the last two teams of the hosted905-team export bind24 players at phase4, retain six metadata-only entries, expose standard/custom fields and overrides to an independent live scorebug consumer, and finish naturally4–3 with identical frozen context in reconciled independent deliveries. Reopening the persisted real-game database retains context/ACKs and delivers to a third consumer. The broker reads companions; the DLL does not parse or render them. No new DLL or public Windows package.

Source59d archive8e89717d9b24a1588a19308341c026722fcf4623babf2f01e6aab48acc6de4bd/265 inputs;539 exact-source tests/zero skips,98 schemas/16 checks/resolved references and actual editor ZIP SDK/HTTP readback pass. Source-only catalog packaging retains slots. Frozen context provenance, immutable companions, explicit source rows and independent result ACK/restart are covered. Simulated59 Load/Start remains service evidence. Native61 separately verifies actual game Load/Start, phase4 metadata binding, live scorebug reads, natural final results and persisted result reopening with the unchanged52zd DLL. See the ledger and current preservation.

Documentation checkpoint: October 19, 2026 UTC. This is the current evidence ledger for the HTTP contract. Older work-log entries, fixtures and immutable packages retain the versions and limits they had when recorded.

Previous match-control checkpoint

Development51: native/live source52zd/195 inputs archiveab93214f3321ce54dce0de822de4f3330aca7363525b40eebc8b283d65966b50, non-probe DLL4a4f0affa8e251b536dfcec395eb6f77673548c923083a5cec052099208fd88d, Dell MSVC19.51 /W4 /WX. Final API52zh retains all67 native inputs;510 Windows API tests/84 schemas and44 real-response/closed-state checks pass. SnapshotABI26 is74648bytes; older prefixes are unchanged, eventABI4/profile4/court3 remain.52 native control cases execute original timeout eligibility/bookkeeping/queue instructions and compiled adapter guards; picker/commit side effects and special modes are mocked. Retained1104 coaching/defense,111 substitution and27 pause cases pass on the same production DLL.

Live52za confirms corrected half-boundary timing/four native entries/two timeout deductions and natural returns.52w confirms actual keyboard selection/CPU clear/reselection/away original commit before reveal and missing-gamepad refusal.52x confirms original-profile plan replacement/cancellation, pause/resume and full timeout6→5/return. All use zero generated input/adapter errors. Physical gamepad actuation/multiple-human play remains unverified; no controller is available. Exact-clock physical entry, persistent hardware GUIDs, universal legal ball state and broader rule/callback/presentation semantics are not claimed. Prototype52p observer error,52q early-retirement failure and52v intermission-timing audit are preserved.52v separately passed natural4–3 result/inbox but is not accepted scheduler evidence.

Preservation52zm closes all six owned runs/restores98 files/removes six exact-body idle tasks/preserves95 originals, original EXE/INI/stable35/fixtures/immutable public packages/four hosting identities. Isolated DLL restores50; source51 stays frozen. No additional public release, GUI, agent, dependencies or rare-rule/full-match campaign is required.

Previous external game-flow checkpoint

Development50 records source51u archivea7cbf6e344690e55fa01de748d0e41ca2ba6f087c534b76842e37dfec9ab8680/187 inputs and non-probe DLL4f897924002508d689ebb18915f7d2eb0cd49517ad086aba1fb8703eca216f47, Dell MSVC19.51 /W4 /WX.493 exact-source Windows API tests,76 schemas/48 positive-negative checks/unchanged historical results,1,104 prior native cases,111 substitution/ownership/startup cases and27 pause/dispatch cases pass. ABI25 appends128 substitution observations and a56-byte mailbox after ABI24, total68680. HTTPv1/eventABI4/profile4/court3 remain unchanged.

Live51w verifies138 defensive values plus18 coaching values before reveal, phase4 identities, both-team queue/cancel and actual native/on-court entries at a natural foul. Full flow and preservation pass; see the external-flow ledger. Failed51k accepted queues but the automatic lineup builder replaced them at the first foul; corrected availability2/PE version2 preserves exact API-owned pending requests while leaving native entrance handling intact.51k separately passed natural4–2 result reconciliation, independent consumers, repeated ACK, store-before-ACK and matching-owned finished closure. Original closed DB/WAL/SHM hashes are preserved by disposable-copy diagnostics. Exact-minute schedules, physical guarding, worker/SEH failure paths, universal legal ball state and deferred presentation are not certified.

Final51w completes the natural5–4 college match,23-player reconciled result, independent consumers/repeated ACK/store-before-ACK and matching-owned finished closure. Both51 runs close/restore98 files; preservation51z verifies95 originals, protected runtime/packages/hosts and removes two exact-body idle tasks. 33 actual-live schema checks pass. Final documentation source51za changes prose only: all179 runtime inputs, including63 native inputs, match tested51u. No game/build/test is repeated for that prose-only freeze.

Previous live coaching/defense49 checkpoint

Development49 records source49z archive820f60c78c581df826727ca07180e7c81a5075f4c2d1f10718ceebf721ab8049 /179 inputs and non-probe DLL0a99df4af3d5864a1e2bbc0c5cfc08b94921bca59c0264a72ac43b6106eaa961, MSVC19.51 /W4 /WX.471 exact-source Windows API tests,1,104 defensive/coaching instruction cases,27 pause/dispatch cases and71 schemas pass.37 positive-negative fixtures plus unchanged historical results pass;190 actual-live schema checks cover89 commands/receipts and snapshot/settings/lifecycle responses. ABI23 appends652 bytes after ABI22; ABI24 adds16 owned-pause bytes, total68496. HTTPv1/eventABI4/profile4/court3 remain unchanged.

Live49zf verifies18 startup coaching values and42 distinct in-game changes: all nine settings for each side plus all six defensive fields for first/last opposing slots on both sides. Values persist after resume and restore through the public API. Owned pause freezes all three clocks and reported player positions; resume and one-second expiry return natural play. Running writes/stale revisions refuse, duplicate IDs retain immutable receipts, cancellation abandons without a result. Zero adapter errors/generated input. Defense is stored configuration, not physical guarding or universal AI certification. Nested native menu preservation is compiled counter evidence; exhaustive live menus/modes remain unverified. Direct substitutions/startup defense remain unfinished.

Counter-only prototype49v failed live pause acceptance while launch worked; no coaching writes followed its failed frozen-clock assertion. Current availability2/PE simulation-control version2 additionally suspends complete9F1D20 dispatch and excludes counter-only availability1. The earlier9F2640 gate controls rendering, not complete simulation. Failed evidence and closed database originals remain unchanged. Both49 runs close and restore98 baseline files. Public packages, stable35,95 originals, INI, fixtures and hosts stay separate. No public release or deferred niche/full-match certification is added.

Final documentation source49zm updates Markdown metadata only relative to tested49z. All non-Markdown inputs, including native code/current schema/tests, match49z; no new game or repeated regression is required for that documentation-only stage.

Accepted startup coaching47 checkpoint

Development47 coaching records frozen48za sourceac020a51ad81914c72a0c8679adc8ccd8102471759acd686e4986aa4d23a20c1,452 exact-source Windows API tests and64 schemas. ABI22 appends76 coaching bytes after the ABI21 prefix; total67828. HTTPv1/eventABI4/profile4/court3 remain unchanged. All57 non-Markdown native inputs match compiled48u/DLL0ce3673f4ab3a847fa498c8d8cf728093d0606ae91b60387f561487b8a32f520 (/W4 /WX).244 compiled-wrapper/original-handler cases cover numeric transitions, edge wrapping, both sides, invalid inputs and neighboring-state preservation. They are synthetic-memory instruction evidence, not full-engine/AI certification.

Live48w verifies college10/all282/2k/CPU;48x verifies original2/18 reversed coaching values/broadcast/primary-away. Both bind at phase4 and naturally reach phase10 with zero errors/no generated input; coaching values persist in play and owned cancellation passes. Final48za differs from live48u only in API startup-catalog capability gating for older adapters; native/coaching application code is identical.64 schemas cover14 live scenarios and historical results, preserving42 original DB/WAL/SHM byte hashes. All95 originals, the98-file pre-research isolated baseline, stable35, source fixtures, public ZIPs and hosts remain protected. No new release, visual presentation, exact substitutions, rare-rule or hardware certification is implied.

Accepted development40–46 launch expansion

Active ledger records API46/ABI21,441 exact-source Windows API passes on47zzp and compiled non-probe47zze DLL 2721b166874e71e036d09c0980176bf53df5c794121ed055fee5a5a54823d046.55 native code/resource inputs match; changed Markdown is excluded. Source archive 7c6d87fbf946a536499f2dcc22a4f824e91f745eb6f2a44dee94f5266e461c60 identifies167 inputs, and final integrity verifies146 runtime-code files. Later documentation/current-schema edits are separately identified. SnapshotABI21 appends36-byte startup participation after ABI20's67716-byte prefix; total67752. HTTPv1/eventABI4/profile4/court3 remain unchanged.

Combined47zzq passes source46 request execution with shared roster edits/order/minutes, prepared905-team ROS, all13 choices, college10/media4/all264 presets/2k six controls/CPU participation, phase4 IDs, natural phase10/clock598.8780,16MiB export/no-overwrite and owned cancellation. Zero errors/no generated product input. Home47zzf and away47zzi pass the same native DLL; away has guarded rendered marker evidence and stock export. The capture activates only the owned window after independent SDK gameplay observation. A home capture missed cancellation and was refused; it is not visual proof. Earlier47b/h/k/m/p/v/y/za passes retain their original configuration/identity/preset/authoring/request/camera scopes.

61-schema acceptance covers12 live scenarios/188 actual commands and receipts, seven request documents plus no-op shape checks, authoring and immutable historical results. Original36 DB/WAL/SHM hashes remain unchanged because SQLite reads disposable copies. Canonical schema SHAe85cd8d4 differs from frozen47zzp only by accepting empty optional roster_edits, already allowed by runtime validation. Tests441, manifest, combined, away/export, home.

Final preservation47zzs archives final isolated saves/restores95 originals, verifies original EXE/stable35/INI/fixtures/three public ZIPs and unchanged hosts, closes18 runs/removes20 exact-owned idle tasks, and leaves no game/debugger/API listener. The candidate remains isolated. Existing synthetic-fixture pink crowd/floor artifacts, original watermark/quarter labels, physical-device/multiple-human/fullscreen/device-reset/other-PC and full-mode/asset limits remain. No new package, full-match/rare-rule recertification or installed baseline upgrade is implied.

Earlier development39 native startup records355 tests/41 schemas, normal Quick/neutral stock/branded905 flows on DLLc14d0579, and its inactive-Home load restriction. That milestone's phase10-only identity limitation was superseded by the guarded phase4 implementation. Older source archives and verification flags remain unchanged.

Versions and runtime

The frozen production-rule DLL36ck is 61683231d2b425f2531e6cc8d21c29ad2a0a2799e84c28bd453b74f649cf175b. The latest natural OT test used the same rule source with private observer36cs, DLL 94bd34edd755a4a4bb95446811e405fd237573fdbedc6de7a23f5adbdc3bda7d. No private actions executed during that managed match. Observer builds are research tools, not the production DLL.

The preserved baseline is development35 (snapshot ABI11); the Dell was restored to that baseline after testing. Development36 retains the college implementation evidence; development37 appended preparation support; development38 appended native roster loading/Start; development39 added native startup/branding; current46 expansion is explicitly staged in an isolated game, with each run identified in the active ledger. Stable35/public previews are not upgraded. The user deferred another uninstrumented full-game smoke. They subsequently requested a hosted Windows preview, now built from explicitly staged current source; historical candidate ZIPs remain unchanged. Fresh shipped-SDK/runtime tests pass272 cases, and the optional packaged broker reports live ABI14/profile4 data in a short expanded-roster/college match-start check. Acceptance42g records the scope; preservation42h confirms stable35 and95 original saves remain unchanged. The startup log still contains an older ABI12/overtime-unverified sentence; use the actual ABI14/profile4 declaration and this ledger, not that stale startup text. No new full-match certification is implied.

Launch38 exact-source acceptance

Launch ledger records source45r SHA099433229d7ec4d9f8b8524e396599bbe2bcd63262a86e6e50e223fc17637a46 and production/non-probe DLL6280d5df577082adb233dc3a7c652c413c1f15a9b3f692fc239a29980e035d7e.345 Windows API tests and39 schemas pass. Original native load/Start pass from Home with stock105-team and generated905-team rosters. Fresh45r unattended process launch/public SDK passes all13 settings, exact loaded SHA/counts, ten starters, native roster-order/minute preferences, college startup clocks and zero errors. Duplicate command identities preserve earlier receipts. Acknowledgement uses guarded Windows input only during recognized intro/title states; load/Start invoke original native handlers. Early gameplay is not another full-match/OT/rare-rule certification. Generated crowd materials remain pink as in the earlier manual fixture; custom assets are not certified. Public preview2, stable35 and original95 saves remain separate.

Preparation37 exact-source acceptance

Private Dell44c source ZIP SHA256 710d8e8462a3162b80cca500042fdba8c363058b7870aacf7e552f9daa236592, non-probe DLL 07d6e527c53dc4fd0adefe758088fc7c6091b7c3dee30f949f9ce0d682ea0f3e. 307 API tests, 30 compiled native cases, 116 writer tests and 35 schemas pass. Explicit staged-package identity is asserted; embedded Python ignores PYTHONPATH. The native harness documents UTF/default-uniform/preset-bank mocks and does not alone prove rendering/dispatch/SEH.

Live stock client applied all13 controls; original options/arena/books and alternate-uniform gameplay with a running four-minute first quarter match. Original Simulation preset matched all126 native bank values. Generated905-roster startup matches final teams903/904, every requested roster slot, four changed32/10-minute preferences and both starting fives, with zero adapter errors. College clock180/OT60/shot30 remains distinct from quarter setting5. No full-match/export, exact substitution schedule or custom-art certification is added. Uniform menu previews can lag; actual44c alternates passed.

Full scope/usage. These are isolated source/binary candidates, absent from public preview2/editor dev7. 44f preservation verifies95 saves/stable35/originalEXE, no game/API and both hosts alive. Six completed owned tasks were removed after guards. Existing rule evidence, historical results and public ZIPs remain unchanged.

Implemented interfaces

• Sampled player/team box scores, positions, clocks, free-throw schedule, qualifying fouls, bonus, timeout pools and observed presentation phases; physical ball attachment and separate pause/replay state.
• Explicit external match/team/player identity, ordered captured counter/record observations, scoped rebound credit/retraction lineage, replay cursors and loss reporting.
• Reconciled immutable final results, per-period scoring ledgers where continuous evidence permits, independent consumers and idempotent acknowledgments.
• 59 sliders on each user/CPU side; 20 gameplay/presentation options plus 19 original rule switches; 13 preferences across 10 stored controller slots.
• All26 camera descriptors/readbacks; 17 exhibition selections and 63 configurable camera/control combinations checked live. Nine internal camera types remain read-only.
• App-owned version1 request documents; original/college per-child rules,264 startup preset values, exhibition camera and primary-side/CPU control commit before reveal; explicit failure/recovery receipts.
• Optional shared writer: named roster query/edit/lineup/minute transactions, immutable content IDs and authenticated binary export; source files are preserved.
• Guarded zero-stat phase4 external identities/finite match metadata; broker-owned cancellation or finalized-result completion, separate from attached process control.
• Thirteen pregame controls: all nine Game Options entries, both teams/uniforms and paged loaded catalogs. Source-only working-copy roster authoring sets starting order/native minute preferences. Development38 adds public roster loading/Start. Catalog rows are run/roster scoped, not permanent external IDs.
• Opt-in college halves, scaled OT, clock/reset/bonus/arrow/timeout policies and measured court observations, described below and in the college evidence.

These are experimental development interfaces around the original simulation. experimental does not mean absent; unavailable does. Development53 supplies a scoped accepted-transition legal-state interpreter and experimental native SDK. Universal event semantics, uncovered legal transitions, unsupported rule mutation points and safe asset hot-swap remain unavailable.

College implementation and evidence

Exact candidate native suites: 1,066 core/42 guards;144 FT;1,928 dead-ball reset;106 media scheduling;1,192 timeouts;2,882 recovery/control;9,408 original inbound/spot;1,536 AP/reset;2,688 kick;10,800 injury ownership/refund cases. Mocks and boundaries remain in the release-source summary; these counts are not independent natural gameplay events.

End-to-end results

• Latest36ct: two natural60-second halves tied0–0; normal settings restored at OT; Miami4–Spurs0 after one natural60-second OT. Current Result schema, all additive player/team totals, two logical halves plus OT, issue-free reconciled ledger, two consumers, repeat ACK and immutable delivery pass.92 public settings commands comprise46 temporary changes plus46 restorations. No score/clock/pose/outcome writes or private actions during the managed match. Delivery receipt, zero-action proof.
• Earlier36bl: natural10-minute halves, Miami40–Spurs24, four automatic TV cycles and six rim→offensive-rebound→20-second control chains. Reconciled regulation export and independent consumers pass. This used earlier36bc diagnostic hooks; it is not a full-match certification of every later hook. Receipt.
• Both bindings began after some opening activity. Whole-match event completeness remains false. Result finality and per-period coverage must be read from each result, not inferred from a successful reference test.
• Latest36ct closed with all95 saves and original working files restored, no game/API listener. Restoration. Artwork hosting is separate.

The frozen36ck API suite passed271 tests on Windows before this documentation/metadata update. That development36 metadata refresh passed272/272 tests on Windows11/Python3.13.16, including profile-verification and result-certification semantics. All34 canonical schemas, four current/legacy profile declarations,35 retained real36ct snapshots and the immutable historical OT result validate. Full receipts and document checks are in current verification. Earlier Mac broker/client passes retain their recorded versions; no current Mac/Linux game support is claimed.

What verification fields mean

User-excluded rules remain excluded: five-second closely guarded; NCAA-specific charging, continuation, technical/challenge redesign; flawless CPU behavior and compatibility with every gameplay mod. Some native screens still use NBA quarter labels. Midcourt/recovery uses engine actor-center conventions; exhaustive anatomical/airborne boundaries are not certified. These limits do not create a new requirement to hunt rare events before other development.