Skip to content

AutoStory: seeded disturbance sim (wild battles, drift, wanderers, missed screens) - #262

Draft
dayfine wants to merge 45 commits into
bdsp/center-run-settlefrom
bdsp/story-sim-monte-carlo
Draft

dayfine wants to merge 45 commits into
bdsp/center-run-settlefrom
bdsp/story-sim-monte-carlo

Conversation

@dayfine

@dayfine dayfine commented Nov 8, 2026 •

Copy link
Copy Markdown
Owner

Depends on #375 (base bdsp/center-run-settle, the consolidated BDSP console stage; the diff shows only this change).

The story sim ran each row once, tile-exact, with no wild battles, so today's 01.03 console failures (a battle re-plan leaving Roark's room a row off; an open leg ending off the Center door's column) passed in the sim. This adds a seeded disturbance run: per pinned seed, wild battles on encounter tiles, sub-tile drift on stick holds, wandering-NPC seed and start time, and sometimes a missed fade or banner; it reports a pass rate per row and the failing steps with a reproducing seed. The runner and disturbance model are game-agnostic (Pokemon/AutoStory/Pokemon_StoryMonteCarlo, to move to #122); the BDSP binding runs over GridSim. The rates are measured on lab-B's BDSP logs of 00.00-01.11 (2026-10-31 to 11-08), with how each was counted in DisturbanceRates: caves 16 battles in 838 tiles, grass about 10 in 1,200, D-pad taps without drift, 2% per fade or banner. With them and #226's re-plans, 01.03 passes 396/500 seeds (57/500 before) and all rows 1044/1150 at 50 seeds (628 before). The old-01.03 test keeps the first, harsher rates and still finds both console failures. Default mode runs 20 seeds per row; PA_BDSP_STORY_MC_SEEDS / PA_BDSP_STORY_MC_ROWS give a heavy run, PA_BDSP_STORY_MC_TRACE=<seed> logs one seed's run. Tested with the headless SerialProgramsTests runner.
Core stage 7: DisturbanceRates::missed_hold drops one fade's hold, drawn from its own random stream so the other draws do not change. In each row's test, every seed that passed with no miss runs again with one hold dropped and must pass again: 220 reruns, none failed. Every seed's outcome matches its outcome before the change (1060 runs).

Core stage 5: DisturbedRouteIO's taps record their battles' screen events, so seeded battles are placed as on the console (53 placements over the 01.0x rows). Every seed's outcome matches its outcome before (1060 runs, 911 pass), and the 200 single-missed-hold reruns all pass.

Core stage 5 (abf6571): DisturbanceRates::dialog_between_steps pops a dialog up before a step. It draws from its own stream, taken last from the seed, so no other draw changes. New case "dialogs pop up between steps, 00.00 to 01.11" runs 5 seeds a row (3 runs each) at 10% a step. Of 115 runs, 110 pass without dialogs, 110 pass with them cleared by #226's executor, and 71 pass with them left up (the engine before this change). The test fails if a seed that passes without dialogs fails with them cleared. Every other row's numbers match the earlier run at the same seeds, and all 200 single-missed-hold reruns pass.

🤖 Generated with Claude Code

https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss

dayfine and others added 5 commits November 8, 2026 12:55
Each story row runs on the story simulation once per pinned seed, disturbed
as the console disturbs it: wild battles on grass and in caves, drift off
the tile grid on taps, stick presses and holds stopped short of a wall, the
maps' wanderers with a seed and start each, and sometimes one missed fade
or banner. Each row's test logs its pass rate and, per failing step, the
failure kinds with a reproducing seed, and fails under its pinned floor.
PA_BDSP_STORY_MC_SEEDS and PA_BDSP_STORY_MC_ROWS run a heavier set.

The seeds, the per-seed plan, the runner and the report are game-agnostic
(Pokemon/AutoStory/Pokemon_StoryMonteCarlo); the BDSP binding is
SimDisturbance over GridSim and SimExecutor, which moves from the story sim
tests into PokemonBDSP_AutoStory_StorySim.h/.cpp.

The old 01.03 (before be57db1 and 4f13c19) must fail off Roark's row
after a wild battle and at the Center's door.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
The disturbance model was harsher than the console: 01.03 seeds drew 8
to 14 wild battles where lab-B met 0 to 2, D-pad taps drifted onto the
mine's warp, and a quarter of the seeds missed a fade. The rates now
come from lab-B's BDSP runs of 00.00 to 01.11, 2026-10-31 to 11-08,
each with how it was counted: caves 16 battles in 838 tiles, grass 10
in about 1,200, taps without drift, a 2% miss per fade or banner
looked for. The wanderers' step pace is a rate too; halving it did not
change how often Oreburgh's hiker stays in the way.

The old-01.03 harness test keeps the first, harsher rates, and finds
the Center-door miss from the mine's warp. PA_BDSP_STORY_MC_TRACE=<seed>
logs one seed's run with its battles and encounter tiles.

01.03 at 500 seeds: 57 to 396 passes (with #226's re-plans on a
battle at a leg's first input). Floors are today's 20-seed rates.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
…' into bdsp/story-sim-monte-carlo

# Conflicts:
#	SerialPrograms/Source/PokemonBDSP/Programs/AutoStory/Simulation/PokemonBDSP_AutoStory_StorySim_Tests.cpp
#	SerialPrograms/cmake/SourceFiles.cmake
…' into bdsp/story-sim-monte-carlo

# Conflicts:
#	SerialPrograms/Source/PokemonBDSP/PokemonBDSP_Tests.cpp
#	SerialPrograms/Source/PokemonBDSP/Programs/AutoStory/Simulation/PokemonBDSP_AutoStory_StorySim.cpp
#	SerialPrograms/Source/PokemonBDSP/Programs/AutoStory/Simulation/PokemonBDSP_AutoStory_StorySim.h
#	SerialPrograms/Source/PokemonBDSP/Programs/AutoStory/Simulation/PokemonBDSP_AutoStory_StorySim_Tests.cpp
#	SerialPrograms/cmake/SourceFiles.cmake
# Conflicts:
#	SerialPrograms/Source/PokemonBDSP/PokemonBDSP_Tests.cpp
dayfine and others added 18 commits November 9, 2026 02:13
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss

# Conflicts:
#	SerialPrograms/Source/PokemonBDSP/Programs/AutoStory/Simulation/PokemonBDSP_AutoStory_StorySim.cpp
…sed still passes

DisturbanceRates::missed_hold drops the hold of one fade a row looks for,
drawn from a stream of its own so the other draws stay put. The BDSP test
runs each row's seeds with and without it (no other miss) and fails on any
seed that passes only with every hold seen.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
…old unseen

The separate serial test ran every row twice (about 19 min); each row's
test now reruns only its seeds that passed with no miss, in parallel with
the other rows.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
…lk does

DisturbedRouteIO's taps now fill WalkResult::events and count the taps
that moved after a battle, so the walker and the story's taps place a
seeded battle by its screen events, as on the console.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
# Conflicts:
#	SerialPrograms/Source/PokemonBDSP/Programs/AutoStory/Simulation/PokemonBDSP_AutoStory_StorySim.cpp
#	SerialPrograms/Source/PokemonBDSP/Programs/AutoStory/Simulation/PokemonBDSP_AutoStory_StorySim.h
#	SerialPrograms/cmake/SourceFiles.cmake
DisturbanceRates::dialog_between_steps pops a dialog up before a step,
drawn from a stream of its own (DisturbancePlan::dialogs, last from the
seed, so no other draw moves). The new Monte Carlo case runs every row
to 01.11 with dialogs at 10% a step, cleared by the executor and, for
comparison, left up as before it looked: a seed that passes without
them must pass with them cleared.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
dayfine and others added 10 commits November 10, 2026 16:20
Brings in the base stack's code-health splits and outputs-last core
signatures.

Conflict resolutions:
- StorySim.h/.cpp: SimExecutor::run and run_step take (step, state)
  as on the base; this branch's place/settled/record_failure kept.
- SourceFiles.cmake: union of both sides, AutoStory block regenerated
  from the files on disk.
- SimDisturbance.cpp: includes WalkRules.h, where standable() now lives.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
…after #226's merge)

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
…te-carlo

Brings #128's seam work down through #167 and #226. No conflicts.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
…te-carlo

Picks up #226's code-health fixes after the #128 merge. No conflicts.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
Retargets #262 onto the consolidated BDSP console stage (#375's branch).
Conflict: StorySim.cpp SimKnobRouteIO, whose base here is
DisturbedRouteIO: walk_taps and #375's new walk_to_wall check the fade-in
first, then call DisturbedRouteIO's, as lab/dev-bdsp has it.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
Picks up the consolidated stage's StorySim split and code-health fixes. No conflicts.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
dayfine added a commit that referenced this pull request Nov 12, 2026
StorySim.cpp stood at 999 of its 1000 lines, so neither #262 nor #400
could add a line. The story variables and the coord-event triggers they
arm (story_vars(), story_triggers()) move unchanged to SimTriggers.cpp;
the declarations stay in StorySim.h.
Review with: git diff --color-moved HEAD~1

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
@dayfine
dayfine changed the base branch from bdsp/autostory-mine-floaroma-wip to bdsp/center-run-settle November 12, 2026 06:12
Picks up the consolidated stage's test-file split. No conflicts.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
dayfine and others added 4 commits November 12, 2026 01:22
…(red)

Every Monte Carlo seed of 00.05 stood still on east_to_the_wall: a hold
east at a walk from (46, 54) moved no tile, where the grid sim's own
hold walks 6 to the wall at (52, 54). This test holds east, west and
north at a walk and east at a run, with no disturbance, both ways. It
fails now on the walks east and west.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
disturbed_hold() holds a tile at a time, each timed at the walker's
pace (route_ms_per_tile(), 160 ms east or west at a walk). The grid
sim walks east and west at 200 ms a tile (fe2da03), so each of those
holds moved no tile and the whole hold stood still. Every Monte Carlo
row with a walk east or west to a wall failed on all its seeds: 00.01,
00.03, 00.04, 00.04b, 00.05, 01.05 and 01.11. It now times each tile
with sim_ms_per_tile(), GridSim::hold()'s own pace.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
The Center's-door half rebuilt the old row by removing steps that are
gone from the table (south_of_the_centers_door, ...), and cut the town
rows at up_to_the_mines_sign, also gone, so they ran no steps and
passed 200 of 200: it could not show the 10:39 failure any more.
Retired; today's 01.03 reaches the door by a counted wall leg up column
46. The Roark half still finds the 11:55 failure after a wild battle.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
dayfine added a commit that referenced this pull request Nov 12, 2026
… sim's pace; old 01.03 keeps Roark's room) into lab/dev-bdsp
…carlo

Picks up the single registration of the BDSP Dialog/Menu detector suites.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_013VUubhHeYSMGSSDXZkKZss
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant