How we calculate
the Gap Score

We measure mention volume and acceleration across topic-relevant subreddits. The base velocity uses an exponential curve over on-topic post count (score = 100 × (1 − e^−n/15)), so going from 0 to 15 posts matters more than going from 50 to 65. An engagement bonus of up to 10 points is added when verified upvote data is available from Reddit's public API.

Platforms in SVI calculation:

Reddit data is sourced from two complementary feeds: Reddit's public JSON API (real upvote counts, comment totals, verified timestamps) and ForumScout keyword search (broader topic coverage). Results are merged and deduplicated. Only posts from topic-relevant subreddits contribute to the velocity score. Engagement numbers are only displayed when they come from a verified API response.

We ingest article counts, headline prominence, and broadcast minutes from 180+ mainstream media sources across left-leaning, right-leaning, and centrist outlets. MCI is a 0–100 score where 100 = saturation coverage.

We deliberately include both sides of the political spectrum to avoid ideological bias in the score itself. A story buried by CNN and Fox simultaneously scores much lower on MCI than one buried only by CNN — because cross-partisan coverage avoidance is a stronger signal.

After subtracting MCI from SVI, a square root compression curve is applied before multipliers. This spreads scores across the full 0–100 range — without it, topics with genuinely near-zero MSM coverage would cluster at 90–100 with no differentiation. The curve maps the raw gap to a base score (rawGap 60 → 50, 80 → 58, 100 → 65), leaving room for sentiment and coverage anomaly multipliers to push higher-signal stories to 90+.

Before computing MCI, articles are passed through a topic relevance filter. Each story has a curated set of required keywords — an article must contain at least one in its headline or description to count toward coverage. This prevents loosely related articles (e.g. a Pentagon budget piece matching a UAP query) from artificially inflating the MCI. If the 10-article sample all fail the filter but NewsData found matching articles overall, a conservative 30% coverage floor is applied — acknowledging the topic is likely covered even when our keyword sample was imperfect.

When social sentiment and MSM framing diverge sharply, the raw coverage gap understates the story's significance. A story the public feels positively about that MSM frames negatively (or vice versa) receives a multiplier of up to 1.35×.

Sentiment is scored using a fine-tuned model trained on media framing and social discourse data, producing a score from -1 (very negative) to +1 (very positive) for both the social and MSM signals.

The most nuanced component. We look for anomalies associated with editorial avoidance patterns — distinct from simply low interest. This is our most interpretive signal and the one most susceptible to false positives. See limitations #09 for the full caveat.

• →Story has high social velocity but zero or negative MSM article growth
• →Story is trending in multiple geo-regions simultaneously with low MSM coverage in all
• →Social platform engagement throttling detected (sudden velocity drop + comment restrictions)
• →Story involves entities with known PR/media relations infrastructure
• →Cross-partisan silence: absent from both left and right outlets simultaneously

Coverage Anomaly Weight ranges from 1.0 (no anomaly signal) to 1.25 (strong coverage anomaly pattern). This is the most proprietary component of the Gap Score.