SONAR Case Study - us_sc

1. Header

Market us_sc - contest us_sc_gov_r_runoff_2026

Election date: 2026-06-23

Generated: 2026-07-16T16:48:12.763626+00:00

Integrity verdict: VALIDATED

2. Locked prediction

Lock id

95d8554d-1f2c-40ed-ab54-55649e84e833

Content hash

b41b6f82d5b3fc8594d3c39aa6f1355a602a26698ff8528da0a6596679aa9899

Locked at

2026-06-21T23:56:01.700490+00:00

Methodology version

stance_balance_impact_weighted

Data window

2026-06-08T00:00:00+00:00 to 2026-06-22T00:00:00+00:00

Cycle

us_sc_gov_r_runoff_2026

External timestamp proof (locks.log lines)

Predicted shares

Candidate	Predicted raw score
Alan Wilson	81.30
Pamela Evette	18.70

Polling benchmark at lock time

Polling benchmark

not recorded

3. Certified result

Source

South Carolina State Election Commission (unofficial certified)

Source URL

https://www.scvotes.gov/

Independent reviewer

rrodgers

MAE (points)

12.8

Winner called correctly

n/a

Decisiveness called correctly

n/a

Certified shares

Candidate	Certified share
Alan Wilson	68.50
Pamela Evette	31.50

Polling comparison (comparator addendum 2026-07-13)

Both comparisons publish regardless of which is closer to SONAR or to the certified result. SONAR's prediction froze at lock time; the latest poll may field far later.

Locked benchmark

none recorded (this lock carries no polling benchmark).

Latest pre-election poll

none recorded.

5. Daily record (30d)

Per-candidate support series with calibration anchors and political-event overlays. Events are labeled as movement in the 48h following the event (correlation, not causation).

No political events in the last 30 days.

6. Multi-instrument agreement

External anchor 2-way A vs SONAR 2-way A on the same date. Divergence is the signed difference (SONAR minus external).

No external calibration anchors on file.

7. Integrity screen

Methodology version

stance_balance_impact_weighted

Concentration ratio (max over window)

0.685

Flagged account count (latest)

n/a

Pool size (min / max over window)

88 / 112

Share ceiling

5% impact cap enforced

7b. Final-week integrity note

Integrity screens showed no unusual movement over the final-week window.

8. Calibration ledger context

This pair MAE

12.8

Ledger mean (prospective validated pairs)

12.80

N prospective validated pairs

This pair included in mean

yes

Retrospective pairs (pair_origin='retrospective') are excluded from this mean, mirroring the live Calibrator which learns from prospective pairs only.