objective() reads GroundState::score (now public) rather than restating
what winning is; a copy in the bot would disagree with the kernel the
first time ground-game rules on F28.
Working out WHERE the modes can differ was most of the task and it
bounds the result: SOLVE always claims for the actor, so own-score and
group-score want the same SOLVE nearly everywhere. That is a fact about
GROUND's action set, not a shortcoming of the bot. Two real divergences,
both readable off the table: SUPPORT regulates someone else (worth less
against a rival, worth MORE under coalitions where a Bond merges them
into my side), and SOLVE's value is the card's value, which greedy
ignores entirely.
THE RESULT — F27 splits in two:
group success UNCHANGED in 34 of 36 cells
who wins MOVES: BONDED COALITIONS at 4p goes 2.04 -> 2.98,
2.12 -> 3.29, 2.05 -> 3.01 winning seats per game
So "the competitive modes are scoring lenses over cooperative play" was
too strong and is withdrawn. The sharper claim: GROUND's scoring modes
change WHO WINS, not WHETHER THE GROUP SUCCEEDS. And the effect is
seat-band dependent -- 2p none, 4p largest, 6p none under coalitions;
two relation slots capping network growth is a candidate explanation and
is untested.
The panel now prints BOTH policies side by side. That was a correction
mid-task: the first version printed only the new one and I compared it
against a figure remembered from CB-WP-0047 -- a comparison against a
board nobody re-ran.
Control that makes the numbers mean anything: under SHARED GROUND the
two policies agree at all but <=2 decision points across 12 boards, so a
moving column is mode-awareness and not simply a different bot.
Also: two T01 tests keyed on `status: proposed`, which ground-game
renamed to `ready-for-implement` mid-session. They now find the module
by asking resolve() -- the structural property is ours and does not move
when another repo edits its vocabulary.
Also: `make vendor` replaces three hand re-vendors with a tool that
regenerates digests by walking editions/, and reports one-sided files
rather than resolving them.
Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
6 KiB
Edition & rules catalog (schema 2 — composable modules by aspect)
Authority: this file + catalog.yaml on main.
Aspect map: ASPECTS.md — full design-space dimensionality for GROUND.
Simulator research: clay-borg research/CB-RES-0010-game-aspects-and-design-space.md
Purpose: pin baseline content and independent rules modules so humans and clay-borg can run any module alone or in combination, measure, keep, or reject — without silent baseline edits.
Model
┌─────────────────────┐
│ baseline content │ editions/ground-darvo-rN/
│ (CSVs, print data) │
└──────────┬──────────┘
│
┌─────────────────────┼─────────────────────┐
▼ ▼ ▼
problem_stress attack_relief end_condition …
(≤1 module) (≤1 module) (≤1 module)
▲ ▲ ▲
└────────── aspects of the game ────────────┘
| Concept | Meaning |
|---|---|
| Aspect | An orthogonal design dimension of the game (how stress is routed, how the game ends, how Problems enter play, …). Full list: ASPECTS.md. |
| Baseline | Edition data package (Problems.csv, modes, …). Frozen columns still require rN→rN+1 (GROUND-WP-0002 T01). |
| Module | One implementable choice on one aspect. Defaults are named modules with no delta. |
| Profile | Named list of modules (convenience). Not a third rules source — expands to baseline + modules. |
| Configuration | Baseline + resolved module list = what is actually played. |
| Composition | At most one module per aspect. Missing aspects use default_module. |
Independence rule: a module may only declare rules on its own aspect. Cross-aspect effects belong in a profile (explicit combo) or a new aspect — never silent coupling inside one module.
Legacy experiments (editions/experiments/h1-…, h2-…) remain on disk as measured packages; they map to profiles. Prefer module_id / profile_id going forward.
Synonym: older docs said axis for aspect. Machine field is aspect:.
Aspects with modules (registered)
| Aspect | Default | Other modules |
|---|---|---|
problem_stress |
none |
flat_any_open (H1-A, reject), scoped (H2, keep) |
attack_relief |
none |
self_soothe_ge4 (H1-B) |
end_condition |
fixed_rounds_5 |
hybrid_clear_collapse (ready-for-implement) |
problem_deal |
fixed_setup |
pressure_deck (ready-for-implement) |
Many other aspects exist but are fixed in r0 until a competing design appears (see ASPECTS.md § full map).
Selection (clay-borg / trials)
Preferred API
baseline: ground-darvo-r0
modules:
- problem_stress.scoped
- end_condition.hybrid_clear_collapse # when implemented
# other aspects → defaults
Or:
profile: h2
# expands to modules: [problem_stress.scoped]
Or multi-aspect:
profile: scoped_plus_attack_soothe
# problem_stress.scoped + attack_relief.self_soothe_ge4
Validation
- Every
module_idexists andselectable: true. - No two modules share the same aspect.
status: proposedmodules may be selected only if the host declaresallow_proposed: true(kernel may no-op or refuse).- State / recordings must store the resolved module list (and baseline id), not only a legacy experiment string — so A/B names the point in design space.
Implementation
- Baseline CSVs: vendor as today.
- Each module: apply
rules_delta+ optionaldata_overlaysfrom itspath. - Apply modules in aspect order listed in
catalog.yamlaspects:(stable, documented). - Conflict: refuse composition rather than last-write-wins.
Profiles (named combos)
| profile_id | modules | note |
|---|---|---|
baseline |
(defaults) | Pure r0 |
h1 |
flat + attack soothe | Legacy H1; reject-as-baseline |
h2 |
scoped | Legacy H2; keep-as-experiment |
scoped_plus_attack_soothe |
scoped + soothe | Unmeasured combo |
scoped_plus_hybrid_end |
scoped + hybrid end | Needs kernel for end |
scoped_plus_pressure_deck |
scoped + pressure deck | Needs kernel for deal |
Adding a module (checklist)
- Confirm the aspect exists in ASPECTS.md; add the aspect first if needed.
- Create
editions/modules/<aspect>/<slug>/withMODULE.mdand, if non-default,rules_delta.yaml(+ CSV overlays if needed). - Register under
modules:withaspect,path,status,decision. - Optionally add a profile that includes it alone and/or with other modules.
- Message clay-borg:
module_id, aspect, delta path, compose examples. - Measure alone first when possible; then measure interesting profiles.
- Update
utility_estimate/decisionper module (and per profile if combo-specific).
What must not happen
- Silent edit of frozen baseline columns inside
ground-darvo-r0. - A new mega-experiment that re-bundles three aspects into one non-decomposable id (use a profile instead).
- Modules that hard-require another aspect’s module without documenting a profile.
Git
main + catalog is discovery; commits pin digests; branches optional for WIP. Modules live as directories on main.
Legacy map
| Old experiment | Prefer |
|---|---|
h1-problem-stress |
profile h1 or modules problem_stress.flat_any_open + attack_relief.self_soothe_ge4 |
h2-scoped-problem-stress |
profile h2 or module problem_stress.scoped |