Authority-Weighted Velocity Anomaly (AWVA)

What it measures

AWVA is a "change detector."

Instead of asking whether sentiment is high or low, it asks whether behavior is unusually different from recent history.

It tracks both:

• Discussion activity changes.
• Sentiment changes.

And it gives more weight to changes happening in higher-trust communities.

Inputs

AWVA uses:

• Current-day community behavior.
• A rolling historical window to define "normal."
• Community trust scores for weighting.

How it is computed

1. For each community, Rebuzz builds a recent baseline for volume and sentiment using a moving average.
2. It measures how far today's value is from that baseline in standardized units (z-scores).
3. Only large deviations are counted (small noise is ignored).
4. Each large deviation is weighted by community trust.
5. Weighted deviations are added into one anomaly score.

Alert tiers:

• Notable: 3+
• Significant: 5+
• Major: 8+

Output + interpretation

Higher AWVA means "today looks less normal" for that ticker, especially in stronger communities.

In the current implementation, AWVA is non-negative and mainly reflects anomaly intensity.

Edge cases/default behavior

AWVA has explicit safeguards:

• If history is too short, AWVA is withheld instead of guessed.
• Extremely flat history is stabilized to avoid false spikes from tiny variance.
• If no meaningful deviation is found, AWVA stays at 0.

How Rebuzz uses it

AWVA is recomputed daily and included in the composite Rebuzz score when enough history exists.

It helps surface "something just changed" events that may not appear in slower-moving consensus metrics.