# Pre-tuning gate assessment

Gate version `0.1.0`; provisional design hypotheses. This assessment flags investigations and does not approve card changes.

Limits and the experiment protocol: [pre-tuning gates](../docs/pre-tuning-gates.v0.md).

Rates use exact counts before rounding. Deck gap and first-player wins exclude draws; draws have a separate guardrail. Conditions are assessed separately.

| Condition | Pairs | Deck gap (pp) | First-player wins | Median / p95 turns | Length gate | Flagged cards |
| --- | ---: | --- | --- | ---: | --- | ---: |
| paired | 1000 | 83.90 (investigate) | 50.85% (within_gate) | 8 / 21 | within_gate | 4 |
| all_random | 1000 | 76.10 (investigate) | 50.35% (within_gate) | 11 / 25 | within_gate | 4 |
| comeback_coven_aware | 1000 | 88.10 (investigate) | 49.55% (within_gate) | 8 / 19 | within_gate | 6 |
| rooftop_tempo_aware | 1000 | 56.70 (investigate) | 51.05% (within_gate) | 12 / 32 | investigate | 4 |

## Source: paired-seat-deck-aware-1000-pairs-seed-42.v0.json

SHA-256: `b267127bd2612690baac932adf44ead19fb65b9d0c9e8862d413d17978e291f8`.

### paired

Seed `42`; policy `deck-aware-v0.1.0`; 2000 games.

Deck leader: `COMEBACK_COVEN`. Round-limit endings: 0.20%. Draws: 0.00% (within_gate).

Cards over 50% seen-but-never-cast, with at least 100 sightings per leg:

| Card | Never cast / seen | Rate |
| --- | ---: | ---: |
| HHX-008 Camera Snap Counter | 1378 / 1670 | 82.52% |
| HHX-011 Surprise Concept Photo | 1493 / 1657 | 90.10% |
| HHX-020 Umbrella Shield | 1411 / 1472 | 95.86% |
| HHX-021 Subtitle Reversal | 1294 / 1294 | 100.00% |

## Source: policy-ablation-1000-pairs-seed-42.v0.json

SHA-256: `90fa06e743cc024b8415ac3f797d5f7a34382d0a317041d7a9837dc936c7fc40`.

### all_random

Seed `42`; policy `random-legal-v0.1.0`; 2000 games.

Deck leader: `COMEBACK_COVEN`. Round-limit endings: 0.60%. Draws: 0.00% (within_gate).

Cards over 50% seen-but-never-cast, with at least 100 sightings per leg:

| Card | Never cast / seen | Rate |
| --- | ---: | ---: |
| HHX-008 Camera Snap Counter | 1482 / 1883 | 78.70% |
| HHX-011 Surprise Concept Photo | 1250 / 1890 | 66.14% |
| HHX-020 Umbrella Shield | 1567 / 1677 | 93.44% |
| HHX-022 Secret Sibling Claim | 1047 / 1643 | 63.72% |

### comeback_coven_aware

Seed `42`; policy `mixed-by-deck-v0.1.0[COMEBACK_COVEN=deck-aware-v0.1.0;ROOFTOP_TEMPO=random-legal-v0.1.0]`; 2000 games.

Deck leader: `COMEBACK_COVEN`. Round-limit endings: 0.10%. Draws: 0.00% (within_gate).

Cards over 50% seen-but-never-cast, with at least 100 sightings per leg:

| Card | Never cast / seen | Rate |
| --- | ---: | ---: |
| HHX-008 Camera Snap Counter | 1319 / 1657 | 79.60% |
| HHX-011 Surprise Concept Photo | 1486 / 1646 | 90.28% |
| HHX-020 Umbrella Shield | 1351 / 1448 | 93.30% |
| HHX-021 Subtitle Reversal | 781 / 1387 | 56.31% |
| HHX-022 Secret Sibling Claim | 969 / 1355 | 71.51% |
| HHX-024 Finale Kiss Cut | 852 / 1396 | 61.03% |

### rooftop_tempo_aware

Seed `42`; policy `mixed-by-deck-v0.1.0[COMEBACK_COVEN=random-legal-v0.1.0;ROOFTOP_TEMPO=deck-aware-v0.1.0]`; 2000 games.

Deck leader: `COMEBACK_COVEN`. Round-limit endings: 0.45%. Draws: 0.00% (within_gate).

Cards over 50% seen-but-never-cast, with at least 100 sightings per leg:

| Card | Never cast / seen | Rate |
| --- | ---: | ---: |
| HHX-008 Camera Snap Counter | 1552 / 1888 | 82.20% |
| HHX-011 Surprise Concept Photo | 1242 / 1890 | 65.71% |
| HHX-020 Umbrella Shield | 1624 / 1731 | 93.82% |
| HHX-021 Subtitle Reversal | 1597 / 1597 | 100.00% |

A never-cast card may have been held by policy, drawn late, unaffordable, or waiting for a response window. This metric does not prove mechanical unplayability.

These are descriptive screens on correlated paired games, with no independence-based confidence intervals. Seed-42 evidence was already known when the limits were chosen. This checker does not establish replication or structured table evidence.