장부: 점수 — Humanoid Ledger

Ledger: scores

Distributions and counts of the three retained measurements — sentiment index (0–100), recommendation index (0–100) and author verdict (buy / wait / skip / mixed, with the author’s own words). These measure the source articles and their authors. This site never rates a product; there is no site score, no ranking and no recommendation of its own anywhere on this ledger. Live data: /api/ledger/scores.

Sentiment index distribution

Index bucket	Articles
0-19	6
20-39	22
40-59	64
60-79	51
80-100	62
Not determined	124

Recommendation index distribution

Index bucket	Articles
0-19	0
20-39	0
40-59	2
60-79	2
80-100	17
Not determined	308

308 of 329 analyzed articles gave no recommendation signal — the recommendation index is almost always not determined, because most coverage is news, not a buying recommendation. Nulls are counted, never filled in.

Author verdicts

Author verdict	Articles
buy	29
wait	76
skip	6
mixed	1
Not determined	217

Per-model scores

Model	Analyzed	Sentiment mean	Recommendation mean	buy	wait	skip	mixed	No verdict
Agility Digit 4	18	68.8 (n=12)	—	1	1	0	0	16
Sanctuary AI Phoenix	12	68.0 (n=5)	—	0	3	0	0	9
Unitree H1	8	77.0 (n=5)	80.0 (n=1)	1	0	0	0	7
Apptronik Apollo	13	68.6 (n=7)	—	0	3	0	0	10
Tesla Optimus	107	51.4 (n=61)	81.7 (n=3)	1	34	0	1	71
Boston Dynamics Atlas	16	65.3 (n=15)	80.0 (n=1)	4	3	0	0	9
Unitree G1	35	58.9 (n=25)	68.8 (n=4)	4	8	3	0	20
Fourier GR-2	10	78.3 (n=3)	80.0 (n=1)	2	1	0	0	7
Figure 03	12	65.0 (n=5)	—	0	3	0	0	9
1X NEO	14	57.5 (n=8)	80.0 (n=1)	2	5	0	0	7
XPENG Iron	20	64.2 (n=12)	—	0	4	2	0	14
Agility Digit 5	64	64.7 (n=47)	79.0 (n=10)	14	11	1	0	38

Means are computed over non-null values only; n is shown next to each mean. A mean of few articles is a thin summary of thin coverage — read it with the distribution above, not instead of it.