Validation scorecard

# Validation scorecard

_Scorecard fingerprint: `scorecard-v1:de38d7dec7fc4162` · 0/3 marketable._

Do Attention Density signals predict forward returns? This scorecard is regenerated from resolved event→return outcomes and is **honest when red**: if a signal stops working, the table says so. No metric is marketed before it is measured here.

> **Status: pending data.** No resolved outcomes yet — the archive and Flat-Files forward returns are accumulating. Tables below show the structure; numbers fill in as history matures.

**Sample readiness** (resolved outcomes needed before a signal is marketed):

- `confidence`: 0/30 resolved — 30 more to go
- `attention_density_percentile`: 0/30 resolved — 30 more to go
- `social_pressure_index`: 0/30 resolved — 30 more to go

## Marketability

| signal | verdict |
|---|---|
| `confidence` | … pending |
| `attention_density_percentile` | … pending |
| `social_pressure_index` | … pending |

_A signal is marketable only once it is significant AND leads price AND holds broadly AND pays AND generalizes AND is stable._

### Signal: `confidence`

| horizon | n | rank IC | IC | hit rate | mean return |
|---|---|---|---|---|---|
| 1h | 0 | — | — | — | — |
| 1d | 0 | — | — | — | — |
| 3d | 0 | — | — | — | — |
| 5d | 0 | — | — | — | — |

_Methodology in `attention_density/pipeline/scorecard.py`. No metric is marketed before it is measured here **and significant** (\* = p<0.05, ns = not significant); flat/negative results are shown as-is._

**Verdict: PENDING** (0/0 checks measured-and-passed) — not marketable yet.

**Diagnostics (1d):**

- Calibration: — (not a probability signal, or no resolved outcomes).
- Out-of-sample: — (not enough resolved, time-ordered events yet).
- Long-short: — (not enough resolved outcomes to form quantiles).
- Stability: — (need ≥2 periods with a defined IC).
- Payoff: — (need both winning and losing outcomes to size the edge).
- Lead test: — (needs resolved outcomes with a contemporaneous price_return).
- Cross-name breadth: — (need ≥2 names with enough resolved outcomes).
- Overlap-adjusted: — (not enough resolved outcomes to deflate).
- Placebo (permutation): — (not enough resolved outcomes).

### Signal: `attention_density_percentile`

| horizon | n | rank IC | IC | hit rate | mean return |
|---|---|---|---|---|---|
| 1h | 0 | — | — | — | — |
| 1d | 0 | — | — | — | — |
| 3d | 0 | — | — | — | — |
| 5d | 0 | — | — | — | — |

_Methodology in `attention_density/pipeline/scorecard.py`. No metric is marketed before it is measured here **and significant** (\* = p<0.05, ns = not significant); flat/negative results are shown as-is._

**Verdict: PENDING** (0/0 checks measured-and-passed) — not marketable yet.

**Diagnostics (1d):**

- Calibration: — (not a probability signal, or no resolved outcomes).
- Out-of-sample: — (not enough resolved, time-ordered events yet).
- Long-short: — (not enough resolved outcomes to form quantiles).
- Stability: — (need ≥2 periods with a defined IC).
- Payoff: — (need both winning and losing outcomes to size the edge).
- Lead test: — (needs resolved outcomes with a contemporaneous price_return).
- Cross-name breadth: — (need ≥2 names with enough resolved outcomes).
- Overlap-adjusted: — (not enough resolved outcomes to deflate).
- Placebo (permutation): — (not enough resolved outcomes).

### Signal: `social_pressure_index`

| horizon | n | rank IC | IC | hit rate | mean return |
|---|---|---|---|---|---|
| 1h | 0 | — | — | — | — |
| 1d | 0 | — | — | — | — |
| 3d | 0 | — | — | — | — |
| 5d | 0 | — | — | — | — |

_Methodology in `attention_density/pipeline/scorecard.py`. No metric is marketed before it is measured here **and significant** (\* = p<0.05, ns = not significant); flat/negative results are shown as-is._

**Verdict: PENDING** (0/0 checks measured-and-passed) — not marketable yet.

**Diagnostics (1d):**

- Calibration: — (not a probability signal, or no resolved outcomes).
- Out-of-sample: — (not enough resolved, time-ordered events yet).
- Long-short: — (not enough resolved outcomes to form quantiles).
- Stability: — (need ≥2 periods with a defined IC).
- Payoff: — (need both winning and losing outcomes to size the edge).
- Lead test: — (needs resolved outcomes with a contemporaneous price_return).
- Cross-name breadth: — (need ≥2 names with enough resolved outcomes).
- Overlap-adjusted: — (not enough resolved outcomes to deflate).
- Placebo (permutation): — (not enough resolved outcomes).