Moriarno Ratings — Predictive Risk Ratings for Japanese IP

Current Forecasts

v6.4 — Awaiting Release

Title	Grade	Issued	Release	Result
Viral Hit	Aaa	2026-03-19	2026-05-28	Pending
Sins of Kujo	Aa	2026-03-19	2026-04-02	Pending
Soul Mate	A	2026-03-19	2026-05-14	Pending
Straight to Hell	Baa	2026-03-19	2026-04-27	Pending

Results verified against Netflix Global Top 10 data after release.

Rating Scale

6-Tier Risk Grades

Aaa

≤20%

Highest grade — G10 entry very likely

Aa

20–40%

High success probability

A

40–55%

Adequate — mixed success and failure potential

Baa

55–70%

Moderate — failure slightly more likely

Speculative Grade

Ba

70–90%

Speculative — low success probability

B

>90%

High failure risk

Investment Grade (Aaa–Baa): P(fail) ≤ 70%. Speculative Grade (Ba–B): P(fail) > 70%.

Calibration

Grade-by-Grade Track Record

Grade	Titles	G10 Hits	Hit Rate
Aaa	15	14	93%
Aa	12	8	67%
A	16	8	50%
Baa	12	6	50%
Ba	19	3	16%
B	22	0	0%

LOOCV calibration across 96 Netflix Japan Global titles (2020–2025). Hit rate = actual Global Top 10 appearance rate per grade bucket.

Forecast History

v4.0 — Archived

Title	Grade	Model	Issued
Viral Hit	Aaa	v4.0	2026-03-18
Soul Mate	Aaa	v4.0	2026-03-18
Sins of Kujo	Aaa	v4.0	2026-03-18
Straight to Hell	Aa	v4.0	2026-03-18

Historical forecasts are preserved as issued. Numbers reflect the model version at time of publication. Original grades shown in current Moody's scale.

Methodology

How Ratings Are Produced

Ratings are generated by a proprietary multi-factor model trained on the complete census of Netflix Japan Global titles (2020–2025). The model evaluates content characteristics, talent history, production track record, and source material to estimate the probability of reaching the Netflix Global Top 10. Each title receives a grade from Aaa (highest) to B (lowest) based on calibrated probability thresholds. All ratings are validated through leave-one-out cross-validation against historical outcomes.

Forecast Immutability

Published forecasts are never retroactively changed. Each forecast is tied to a model version and issue date. New model versions generate new forecasts; previous forecasts remain on record. When actual results come in, we verify against the original published forecast — not a recalculated number.