---
id: GROUND-RPT-0005
type: report
title: "clay-borg: H2 measured — scoping works; bond SOLVE untested by bots"
domain: consumer
repo: ground-game
status: informational
owner: bernd
topic_slug: whynot
created: "2026-08-08"
extends: GROUND-WP-0007
---

# H2 measurement receipt (CB-EV-0032)

Upstream: clay-borg `evidence/CB-EV-0032-h2-measured.md`, hub 2026-08-08.

| # | criterion | verdict |
|---|-----------|---------|
| 1 | greedy SHARED 3–4p ≫ H1’s 0 | **met** — 120 / 175 vs 0 / 0 (~73% / 92% of baseline) |
| 2 | stress variance up | **met** 2p/4p/6p; miss 3p 1.50 vs baseline 1.57 |
| 3 | DARVO + still sometimes wins | **met** (reactive 13/13/57 at 3/4/6p) |
| 4 | bond SOLVE elevated | **untested** by panel (bots ignore scope incentives); incidental claim rate lower on bond |
| 5 | all-global control → H1 collapse | **met** — 3p/4p → 0 |

**ATTACK:** under H2 occasional attack is **affordable** (rank-75 still wins near greedy) but **not profitable** (never beats greedy; always-attack still 0). F17 still open as design judgement.

**Instrument:** `with_variant()` required for owner assignment; bare `state.variant =` left H2 inert — fixed on their side.

No felt-play. Bond joint-motivation needs a table or multi-agent-aware policy.