Skip to content

Commit 11701e9

Browse files
luisleo526claude
andcommitted
docs: state the merge rule's zero-score outcome as it is now
Since 2026-10-01 a change whose two-band target score is exactly 0 passes when no probe lost a grade or became an engine error and no hard-surface probe got worse on a graded metric; the pass is flagged no-improve-no-regression, or improved when a probe got better where the bands cannot see it. README, CONTRIBUTING and the contributor glossary still called that zero a formal FAIL under a documented native-correctness exception, and CONTRIBUTING and the glossary gave the target rule as "net movement >= 0" with no individual regression, which matches neither lane. All three now state one rule: no regression on the hard surface; a score above 0 passes, a probe that leaves a band falling at most one grade; exactly 0 passes only when nothing got worse; any other 0 fails, and so does a negative score. The README's rule moves under its own heading, "How a change is judged", so it can be linked by anchor. The sentences this replaces were introduced by c45cf5a (README: "strictly positive pooled movement" and "Negative movement is outside this exception"), by 116c461 (Document advisory PR CI and required verification statuses: the README's native-correctness exception), by 4fb9db3 (R5 lane L14-B, item 4: CONTRIBUTING's "net movement >= 0", its exception and "it is not traded against anything") and by 8b7965a (R5 lane L14-B, item 3: the glossary's band 1 / band 2 rule); b2a578c last reworded the README paragraph. Co-Authored-By: Claude Opus 5.5 (1M context) <[email protected]>
1 parent d1d1886 commit 11701e9

3 files changed

Lines changed: 25 additions & 11 deletions

File tree

‎CONTRIBUTING.md‎

Lines changed: 12 additions & 7 deletions
Original file line numberDiff line numberDiff line change
@@ -301,13 +301,18 @@ and the documentation build and skips the rest; `docs/ci.md` names the rule.
301301
### How a parity campaign gates a merge
302302

303303
For changes measured against the closed TradingView test set, a sweep is run
304-
before and after over a **fixed** population, and the merge rule is stated in
305-
terms of movement between grade bands: **no individual regression, and net
306-
movement ≥ 0** across the target excellent and excellent+strong bands. A
307-
documented native-correctness exception permits exactly **net 0** with no
308-
individual regression, after full comparison, independent review and green CI;
309-
its FAIL stays on the record and baseline promotion is deferred. Negative
310-
movement is outside the exception — it is not traded against anything.
304+
before and after over a **fixed** population, and two rules must both hold. On
305+
the hard surface (ETHUSDT.P at 15 minutes) **no probe may regress**: none may
306+
lose a grade, stop reporting a metric or become an engine error. Every other
307+
market and timeframe is pooled and scored over two bands, excellent and
308+
excellent+strong: +1 each time a probe enters a band, −1 each time one leaves
309+
it. A score above 0 passes, as long as no probe that leaves a band falls more
310+
than one grade. A score of exactly 0 passes only if no probe lost a grade or
311+
became an engine error and no hard-surface probe got worse on a graded metric;
312+
it is flagged *no-improve-no-regression*, or *improved* when a probe got better
313+
where the bands cannot see it (a grade rising below them or on the hard
314+
surface, or an engine error fixed). Any other 0 fails, and so does a negative
315+
score.
311316

312317
## Reporting a bug
313318

‎README.md‎

Lines changed: 5 additions & 1 deletion
Original file line numberDiff line numberDiff line change
@@ -267,7 +267,11 @@ Every script is exported from TradingView as-is (its own inputs, its own default
267267
- **strong** — ≥ 95% matched, trade count within 6%, entries within 0.1% and exits within 0.5% at p90;
268268
- **moderate** — ≥ 75% coverage with ≥ 90% of the in-window trades matched; **weak** — at least one match; **minimal** — none.
269269

270-
The closed-test grades are the maintainers' own measurement of a fixed population; only the public board can be re-run from this repository. The formal gate requires **no hard-surface regression** and strictly positive pooled movement across the target excellent and excellent+strong bands. A documented native-correctness exception permits exactly zero target-band movement with no individual regression, after full comparison and independent review; its actual FAIL remains recorded and baseline promotion is deferred. Negative movement is outside this exception. The merge ruleset requires the maintainers' `pineforge/verify` and `pineforge/parity` commit statuses on the exact PR head; GitHub Actions CI is advisory. Baseline promotion also requires a recorded PASS and an exact-head merge.
270+
The closed-test grades are the maintainers' own measurement of a fixed population; only the public board can be re-run from this repository.
271+
272+
### How a change is judged
273+
274+
A change is judged by what it moves: the same population is graded before and after it, and two rules must both hold. On the hard surface (ETHUSDT.P at 15 minutes) **no probe may regress**: none may lose a grade, stop reporting a metric or become an engine error. Every other market and timeframe is pooled and scored over two bands, excellent and excellent+strong: +1 each time a probe enters a band, −1 each time one leaves it. A score above 0 passes, as long as no probe that leaves a band falls more than one grade. A score of exactly 0 passes only if no probe lost a grade or became an engine error and no hard-surface probe got worse on a graded metric; it is flagged *no-improve-no-regression*, or *improved* when a probe got better where the bands cannot see it (a grade rising below them or on the hard surface, or an engine error fixed). Any other 0 fails, and so does a negative score. The merge ruleset requires the maintainers' `pineforge/verify` and `pineforge/parity` commit statuses on the exact PR head; GitHub Actions CI is advisory. Baseline promotion also requires a recorded PASS and an exact-head merge.
271275

272276
### What the closed test taught the engine
273277

‎docs/pages/contributing-llm.md‎

Lines changed: 8 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -286,9 +286,14 @@ to decide a merge. The public corpus is 312 probes; the closed TradingView test
286286
set is larger and is not redistributed.
287287

288288
**band 1 / band 2** — the grade bands a campaign's merge rule is stated in:
289-
*excellent* and *excellent-or-strong*. The rule is no individual regression and
290-
net movement ≥ 0 across both; a documented exception permits exactly net 0 with
291-
no individual regression, and negative movement is outside it.
289+
*excellent* and *excellent-or-strong*. Each entry into a band scores +1 and
290+
each exit −1, summed over both. With no regression on the hard surface
291+
(ETHUSDT.P at 15 minutes), a score above 0 passes if no probe that leaves a band
292+
falls more than one grade;
293+
exactly 0 passes only if no probe lost a grade or became an engine error and no
294+
hard-surface probe got worse on a graded metric (flagged
295+
*no-improve-no-regression*, or *improved* when a probe got better where the
296+
bands cannot see it); any other 0 fails, and so does a negative score.
292297

293298
## Where to read next
294299

0 commit comments

Comments
 (0)