Methodology

Last updated: April 5, 2026

EchoIndex turns public crypto news into a signal you can read at a glance. This page explains what goes in, how BEI is built, and where the system can still get it wrong.

What EchoIndex is, and is not

EchoIndex is an informational research product. It does not provide investment advice, trading advice, legal advice, or tax advice.

Think of it as a structured read on market narrative, built from recent Bitcoin-related headlines and sources. The goal is to make fast-moving news easier to read, not to replace source checking or risk management.

Data sources

EchoIndex collects public crypto news from thousands of source sites and links back to original publisher pages.

How news becomes usable data

Recent articles and derived analysis fields are stored so the product can show live news, grouped storylines, and the Bitcoin Echo Index (BEI) over time.

• News shows recent articles, filters, and source links.
• Market pages show BEI history, topic context, and related BTC comparisons.
• Dashboard surfaces a faster top-layer read through Daily Sentiment, Topic Heatmap, and Top Narratives.

AI analysis

AI is used to classify articles, estimate sentiment and Bitcoin relevance, help group repeated coverage, and generate parts of the summary layer.

These AI systems are still being improved over time, so outputs may change as the models, prompts, and quality checks are updated.

For each article, the system assigns structured fields such as:

• Topic
• Content type
• Sentiment
• Relevance to Bitcoin
• Impact magnitude
• Confidence

These fields are used both in the web UI and in the BEI pipeline.

Deduplication and event clustering

Crypto media often repeats the same story across many publishers. To avoid counting one event again and again, EchoIndex groups same-event coverage into event clusters before scoring.

In production, the dedup flow combines headline fingerprints, extracted entities, keyword overlap, event anchors, and AI-assisted checks for uncertain cases.

That grouping logic matters even when users are not looking at a dedicated cluster page, because it helps EchoIndex separate one repeated narrative from several genuinely different stories.

Echo Points and BEI

EchoIndex uses an article-level scoring unit called Echo Points (EP) to represent the direction and magnitude of a story's impact.

Each eligible article's EP depends on signed impact, Bitcoin relevance, model confidence, and configurable topic and content-type weights.

Only certain content types are score-bearing in production BEI: official_filing, official_announcement, official_ruling, reported_event, developing_event, and unconfirmed. Recaps, opinion, market analysis, promotional, low-quality, and unknown types can still appear in the product, but they are treated as non-scoring context.

In the current configuration, those score-bearing types do not all share the same base type weight. Rulings and filings are slightly stronger, developing stories are discounted, and unconfirmed stories stay eligible but with a lower type weight. Topic weights are configurable and currently tilt only moderately across themes rather than using extreme multipliers.

BEI is not a plain sum of the last 24 hours of articles. The production index is updated in 5-minute buckets from event-level state. When a cluster is first seen, or when a genuinely new source confirms it, the cluster can add fresh contribution. Repeated coverage from the same source has much less effect unless the newer article is materially stronger.

Existing event contribution then fades with time. In the current production config, many events stay fully weighted for an initial period and then decay over a freshness window, with some topics fading more slowly than others.

The live index is scaled onto a 0-100 range with a neutral midpoint near 50, but the underlying state is driven by weighted event contribution rather than a simple article counter.

What you see on the site

• Dashboard gives a fast snapshot through Daily Sentiment, Topic Heatmap, Top Narratives, and Latest News.
• News lets you filter by time window, topic, and sentiment, and hide lower-signal items.
• Market focuses on BEI history and related BTC context.

The same underlying EP logic is also used to label stories into Pos.1 to Pos.5, Neg.1 to Neg.5, or Noise.

How we evaluate quality

EchoIndex behaves like a retrieval and ranking system. It selects, groups, and ranks news for usefulness, so the practical question is simple: does the product surface the right stories, in the right shape, at the right time?

• Classification quality: are topic, type, and sentiment labels usually correct?
• Deduplication quality: are different events incorrectly merged, or the same event split apart?
• Ranking usefulness: do the most visible stories feel meaningfully relevant to users?
• Timeliness: how quickly does the system reflect new articles, late arrivals, or corrections?

Known limitations

• A strong score can still be driven by a rumor rather than an official confirmation.
• Publisher quality varies, so multi-source agreement helps but does not guarantee truth.
• Headlines and summaries can be ambiguous, truncated, delayed, or occasionally noisy.
• Automated outputs can be incomplete, delayed, or wrong.
• Recent values can shift when late articles arrive or when clustering and analysis are rerun.
• BEI is context, not a trading system.

Privacy and cookies

EchoIndex may provide account and saved-news features, and authentication may be handled through Supabase Auth and related browser session state.

Cookies and similar storage are used mainly for essential authentication, session continuity, and security. For the full policy view, use the footer links.