test(cliffs): the ore lever OUT OF SAMPLE - precision holds, unknown count is 14 (#84) - #126
Merged
Merged
Conversation
…count is 14 (#84) Everything the last three PRs established about the ore rule was measured on [1500,1500], because that is the only region `oracle-vulcanus-cliff-ore-direction` re-runs with the resources off. A rule characterised on one region and never tested on another is fitted until proven otherwise, so the lever was captured for the two regions the entities fixture covers and it never did. | region | resources | ON | OFF | suppressed | | --- | --- | --- | --- | --- | | [0,0] | 945 tungsten-ore | 283 | 283 | 0 | | [-1200,800] | 1047 coal | 387 | 387 | 0 | - PRECISION 1.000 SURVIVES. Our predicate fires on zero cells in both regions, so 1992 resource entities across two fresh regions produce no false positive. This is the arm that would have caught a rule fitted to [1500,1500]. - THE UNEXPLAINED POPULATION IS 14, NOT 11. #123 could only park `106,26.5`, `90,38.5` and `-1050,1022.5` as undetermined because the lever's region did not cover them. It does now, and the ore suppresses nothing there. - Re-confirms #110's attribution - 27 calcite, 4 geyser, 0 tungsten and coal - at far larger scale than the arm that produced it. The non-vacuity check is in the fixture rather than argued: "0 suppressed" is also what a lever that never reached the generator prints, and the OFF arms read back 0 resources against 945 and 1047. What this does NOT establish: neither region contains calcite or a geyser, the only two controls that suppress anything. So this is a PRECISION test, not a recall test - the rule's positive evidence is still one region, and #125's cascade result is still measured on [1500,1500] alone. Breaking the rule would need a region with calcite in it, and the oracle set has none. New fixture is deliberately separate from `oracle-vulcanus-cliff-ore-direction` so regenerating it cannot rewrite ground truth four merged PRs depend on. PROVENANCE entry added. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015jeznBmWiZywUc7cBw9xFU
This was referenced Aug 4, 2026
wormeyman
added a commit
that referenced
this pull request
Aug 4, 2026
… effect remains (#84) (#129) * test(cliffs): every route from the ore to a cliff is closed - and the effect remains (#84) #128 closed the tile half and recorded a three-way contradiction. Two more routes are now closed, and the honest summary is stronger than "unknown": the search space is EXHAUSTED, not merely unexplored. | route | evidence | | --- | --- | | the cliff FIELD reads a resource | `cliff_elevation = cliff_elevation_from_elevation = elevation = vulcanus_elevation`, whose 47-node expression closure holds no resource region; and our port's raw cell set is BIT-IDENTICAL under every lever arm (2277 cells, same codes) | | `Surface::wouldCollide` does more | disassembled: per-orientation box, degenerate early-out, `constCollideWithTile`, `collideWithEntity`, both with the cliff's own mask at `proto+0x2b0`. No third input | | its ENTITY half | #124 - disjoint masks at prototype level, ordering | | its TILE half | #128 - 841 tiles move, none crosses lava/lava-hot | | the lever perturbs STRUCTURALLY | this PR - richness x2 and x0.5 move ZERO cliffs | The last row was the only route #128 left open. `control:calcite:richness` appears in `vulcanus_calcite_richness` ALONE - not in `vulcanus_calcite_probability` (where the ore lands) nor in `vulcanus_calcite_region` (which drives the volcanic_jagged_ground_range tile) - so it hands the generator a different CompiledMapGenSettings describing the same world. Over [1500,1500]: 861 cliffs and 3933 resources in all three arms, identical cell for cell and orientation for orientation. The surface reads the changed richness back, so the override applied. So the lever is not a structural perturbation: the effect tracks ore PRESENCE - and nothing that can see ore presence can reach a cliff. This does NOT say the rule is wrong. Box overlap against real resource positions reproduces the game at precision 1.000 over 31 cells and survives out of sample (#126). The rule is right; the mechanism is missing, and the port's model of map generation is incomplete in a way none of these five rows covers. Recorded so the next person does not re-walk a closed route. Three genuinely untouched ideas are listed in the notes: chunk-generation ORDER beyond the mask argument, whatever happens to the queue between `generateCliffs` and `applyCliffs`, and `CliffCraterPlacer::tryToPlaceCliffAsCrater` - ruled out for the residual by position but never as a mechanism. New fixture + PROVENANCE entry; capture takes ~7s. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015jeznBmWiZywUc7cBw9xFU * test(cliffs): close the cliffiness gap in the field argument (#84) The field-closure row named `cliff_elevation` only. The cliff generator reads TWO properties, and a closure argument that covers one of them leaves the gate unexamined - `planet-map-gen.lua` routes `cliffiness = "cliffiness_basic"` for Vulcanus, and nothing had traced that one. It is clean: `cliffiness_basic`'s closure is a single node referencing only `x`, `y` and `cliff_richness`. So the row stands, but now for both properties rather than for the one that happened to get traced first. Found while checking whether crater-cliff's appearance under the lever shared a cause; `crater_cliff` (47-node closure) and both Vulcanus biomes it reads are clean too. Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com> Claude-Session: https://claude.ai/code/session_015jeznBmWiZywUc7cBw9xFU --------- Co-authored-by: Claude Opus 5 (1M context) <noreply@anthropic.com>
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Everything the last three PRs established about the ore rule was measured on
[1500,1500], because that is the only regionoracle-vulcanus-cliff-ore-directionre-runs with the resources off. A rule characterised on one region and never tested on another is fitted until proven otherwise, so the lever was captured for the two regions the entities fixture covers and it never did.[0,0][-1200,800]Precision 1.000 survives
Our predicate fires on zero cells in both regions. 1992 resource entities across two fresh regions produce no false positive, so the rule is not merely right on the region it was built from. This is the arm that would have caught a rule fitted to
[1500,1500], and it did not fire.It also re-confirms #110's per-control attribution - 27 calcite, 4 geyser, 0 tungsten and coal - at far larger scale than the arm that produced it.
The unexplained population is 14, not 11
#123 could only record that the lever's region did not cover
106,26.5,90,38.5and-1050,1022.5, and parked them as undetermined. The lever covers them now and the ore suppresses nothing there:Anywhere "11 unknown + 3 undetermined" was quoted, the number is 14.
The non-vacuity check is in the fixture, not in an argument
"0 suppressed" is exactly what a lever that never reached the generator prints. The OFF arms read back 0 resources against 945 and 1047, so the
autoplace_controlsoverride provably applied - the standing lesson from #111, and the reason the zero is worth anything.What this does NOT establish
Neither region contains calcite or a geyser, which are the only two controls that suppress anything. So this is a precision test, not a recall test: the rule's positive evidence is still one region, and #125's cascade result is still measured on
[1500,1500]alone. A region with calcite in it would be the test that could actually break the rule - the oracle set has none, so getting one means choosing a new region rather than re-running an existing one.Fixture hygiene
New fixture
oracle-vulcanus-cliff-ore-direction-regions.seed123456.json, deliberately separate from the existing one so regenerating it cannot rewrite ground truth that four merged PRs depend on. Same seed, protos andalsoResources, so the two are directly comparable.PROVENANCE.jsonentry added; the capture takes ~8s.Verification
pnpm run verifygreen - 1469 app (up 4) + 13 worker + 3 container tests, 0 warnings, exit 0. Measurement only; nothing shipping changed.🤖 Generated with Claude Code
https://claude.ai/code/session_015jeznBmWiZywUc7cBw9xFU