ground-game/reports/260808-clay-borg-h1-measured.md

id	type	title	domain	repo	status	owner	topic_slug	created	extends
GROUND-RPT-0004	report	clay-borg: H1 measured — reaches unregulated seats; group success collapses	consumer	ground-game	informational	bernd	whynot	2026-08-08	GROUND-WP-0006

H1 measurement receipt (clay-borg CB-EV-0030 / CB-EV-0031)

Upstream: clay-borg evidence/CB-EV-0030-h1-measured.md, evidence/CB-EV-0031-a-seat-that-does-not-regulate.md, hub message 2026-08-08. Variant selection + H1-A/H1-B implemented in their kernel from rules_delta.yaml. Baseline unchanged: list asserted.

Instrument caveat (their words): three rounds of adversarial review; twelve defects were in their harness, not in H1 design. Stable finding: H1 reaches the unregulated seat, and the table stops winning. Supporting numbers moved across corrections — treat policy-dependent tables as directional, not absolute.

No felt-play. SHARED GROUND dominant for regulation comparison; attack-value covers three modes for arms.

---

Criteria §3.2 (design note)

#	criterion	verdict
1	DARVO arm rate non-trivial	Policy-dependent. 0 for greedy. Met for unregulated reactive (~1 arm/seat/game at 3p+). Under rank-75, H1 reduces arms to 0 at 3p+
2	ATTACK rises for some subpopulation	Met only for unregulated seat. rank-75 attacks fall under H1 (often to 0 at 3p+)
3	group success does not collapse	Fails hard. Greedy SHARED: 165/190/200 → 0 at 3/4/6p; 2p 132 → 68
4	Bond/GROUND better than DARVO	Vacuous for winning policies (they never arm)

Mechanism (not what H1 assumed)

H1-A is answered by regulation (GROUND), not by climbing to DARVO:

• Stress plateaus ~3 under greedy (below gate 4 and arm 5).
• Actions spent on stress hygiene → fewer SOLVEs → solve-rate tax.
• H1-B never fires under competent play (Stress never ≥4).

Seed-1 3p illustration: baseline clear board / success; H1 ends Stress 3,3,3 with 2 unclaimed and under threshold.

Tax scales with open board (harder at 3p+); intended DARVO effect does not.

Baseline Stress economy (larger than H1)

Under greedy and reactive on baseline r0: peak Stress held never exceeds start 2 — Stress only falls — when those policies never select ATTACK. ATTACK appears to be the sole inbound pressure for those policies. That is a sharper account of RPT-0003 than “ATTACK never pays”: the pressure ATTACK would answer never occurs unless someone already attacks.

(Always-attack rank-95 does drive Stress under baseline — so the economy is reachable; just not by non-attacking competent play.)

Invariant (unexplained)

Under reactive + H1 at 3p/4p/6p: DARVO arms exactly seats × games (one arm per seat per game). clay-borg does not explain it. H1-B disabled → arm count unchanged (withdrawn “H1-B suppresses attacker DARVO” story).

Withdrawn by clay-borg

Story that H1-B suppresses DARVO in the attacker / concentrates it in the target — from a broken multi-preference policy.

Their offered tuning directions (not findings)

If the aim is Stress sometimes reaching 5: pressure scaling with count of unclaimed Problems; delay pressure until later rounds; lower DARVO arm threshold. Each is a rules choice for ground-game.

---

Ground-game reading (for WP-0006-T04/T05)

1. Do not promote H1 as packaged into baseline — criterion 3 fails.
2. Direction is right: Problem → Stress coupling addresses the real gap (inbound pressure only via ATTACK).
3. Shape/magnitude is wrong: flat +1/end is a board-clear tax that competent players absorb with GROUND before DARVO.
4. Next experiment should change how pressure is applied (scaled/delayed/targeted), not only add status stress — though Idea 2 last-place (semi) remains a candidate H2 per design note §6.