Verdict Anatomy

A Seismograph verdict is not limited to a flat counter of "good and bad" comments. The final result is broken down into three global metrics and a deep-dive categorization.

Global Indexes

Each analyzed video receives three aggregated scores:

1. Mood. Reflects the balance between support for the author's arguments and direct criticism. A high Mood score indicates audience solidarity with the content. A low score indicates prevailing disagreement.

2. Controversy (Polarization). Indicates the heat of the discussion. A polarizing video (e.g., sharp political commentary) might have high controversy when the audience splits into two irreconcilable camps actively arguing in the replies.

3. Parasitic Load. Represents the percentage of informational garbage in the comments. This aggregates spam, phishing, advertising, meaningless emojis, and calls to subscribe. The higher the Parasitic index, the lower the actual value of the comment section.

Stance Distribution

The foundation of these indexes is the classification of each comment into one of 7 categories. The algorithm determines not the sentiment of the words (words can be polite while the meaning is destructive), but the stance of the commenter.

Categories are strictly mutually exclusive: one comment = one category.

A detailed description of each of the 7 categories, along with examples, can be found in the Rubric section.

The Mood-Net Main Index

The core metric of PJQ is Mood-Net, calculated exclusively from the substantive opinion pool (SUP, AGA, NEU).

* Formula: Mood-Net = (SUP - AGA) / (SUP + AGA + NEU)

* The index ranges from -1.0 (absolute rejection) to +1.0 (absolute agreement).

* THIN, OFF, SUS, and AGN are completely excluded from this calculation. They are tracked as raw counters only. Including meaningless positivity (THIN) would artificially inflate the support score, corrupting the true signal.

Reading the numbers in practice

A few rules of thumb make verdicts much easier to interpret correctly:

1. The baseline is negative — calibrate accordingly. Comment sections lean critical by nature: people are far more likely to leave a comment to argue than to agree. Across a broad sweep of videos, the typical Mood-Net sits *below zero*. So a Mood-Net near 0.0 is not "lukewarm" — it already means the content was received unusually well. Don't read mild negativity as failure; read it against this baseline.

2. Read Controversy together with Mood — not instead of it. They answer different questions. Mood is *direction* (agree vs. disagree); Controversy is the *size of the split*. High Controversy with a near-zero Mood is the signature of a genuinely divided audience — two camps of similar size — not a calm one. A low Mood with low Controversy means broad, quiet disagreement.

3. High Parasitic Load means "read with caution". When a large share of the comments is noise (common on entertainment, shorts, and viral clips), the opinion pool that feeds Mood and Controversy is thin. The percentages are still computed honestly from the substantive comments, but they rest on a smaller base — treat them as indicative, not precise.

4. A single verdict has sampling variance — and we show it. A verdict measures a *sample* of the comments, not all of them, so the Mood-Net carries a margin of error. Every verdict displays a ± confidence band next to the Mood-Net (a 95% interval derived from the sampling error of the proportions). A larger sample narrows the band, which is exactly why Deep mode exists: for divisive or high-stakes topics, run Deep for a tighter, more stable reading. The measurement modes (Scout / Standard / Deep) let you trade speed and cost for stability; see the Analyzer section.