Evidence Chain & Coordination Signals

How to read the evidence chain and interpret coordination metrics — FUD.ai's core differentiator.

Evidence chain

The evidence_chain field is an array of human-readable strings, each prefixed with a category tag that indicates the data source:

Prefix	Source	What it detects
`[SECURITY]`	GoPlus, RugCheck	Honeypot, mint authority, hidden owner, tax manipulation
`[SYBIL]`	Coordination module	Bot networks, duplicate text clusters, burst windows
`[ON-CHAIN]`	DexScreener, Bybit	Liquidity changes, order book walls, exchange flows
`[SOCIAL]`	Twitter, Telegram	Sentiment shifts, mention volume, influencer activity
`[MARKET]`	CoinGecko, DefiLlama	Price momentum, market cap tier, TVL changes

Reading evidence

Each evidence string is self-contained and human-readable. Your agent can log them for audit trails or display them to end users.

[SECURITY] No honeypot detected, contract source is verified
[SYBIL] unique_author_ratio: 0.15 — highly coordinated bot campaign detected
[ON-CHAIN] 45M $PEPE transferred to Binance in last 24h
[SOCIAL] Mention volume up 320% in 2-hour window
[MARKET] Price down 15% with high volatility, market cap: low-cap meme

In the internal pipeline, each evidence item carries a weight (0.0-1.0) that contributes to the final verdict. However, the CROO deliverable schema strips weights for on-chain compatibility — only the string statements are delivered.

If you need weighted evidence for custom reasoning, contact us about an extended API tier.

Coordination signals — the core differentiator

Most sentiment tools aggregate scores. FUD.ai goes further by computing explicit coordination metrics that distinguish organic fear from manufactured campaigns.

unique_author_ratio

unique_author_ratio = unique_authors / total_posts

Range	Interpretation
0.0 - 0.3	Highly coordinated — a small group of accounts generating most of the noise. Likely a botnet or paid campaign.
0.3 - 0.6	Moderate coordination — some amplification but with genuine participation.
0.6 - 1.0	Organic — many independent authors discussing independently. Real community sentiment.

duplicate_text_cluster_size

Measures the largest cluster of near-duplicate posts using Jaccard similarity clustering.

Range	Interpretation
0 - 2	Low duplication — posts are mostly original content
3 - 5	Moderate — some copy-paste behavior, could be organic echo
6+	High duplication — likely coordinated copy-paste bot campaign

cross_platform_burst_window_minutes

Measures the time window in which the largest duplicate cluster was posted across platforms (Twitter + Telegram).

Range	Interpretation
0 - 10 min	Synchronized burst — posts appeared almost simultaneously, strong coordination signal
10 - 60 min	Short window — possible coordination but could be organic virality
60+ min	Spread organically — posts appeared over an extended period, likely genuine

Putting it together

A verdict with these coordination signals:

{
  "unique_author_ratio": 0.15,
  "duplicate_text_cluster_size": 47,
  "cross_platform_burst_window_minutes": 4
}

...paints a clear picture: 15% unique authors, 47 near-duplicate posts, posted within a 4-minute window. This is almost certainly a coordinated FUD campaign, not organic market fear. Your agent can confidently ACCUMULATE or IGNORE_FUD.

Contrast with:

{
  "unique_author_ratio": 0.82,
  "duplicate_text_cluster_size": 2,
  "cross_platform_burst_window_minutes": 180
}

...82% unique authors, 2 duplicate posts, spread over 3 hours. This is organic community discussion — the fear is real and your agent should treat the verdict accordingly.