Methodology

Data sources

We maintain continuous collectors against seven independent public APIs. Every figure in every published report originates from one of these sources, with a timestamp recording when our infrastructure captured it.

Analytical workflow

Each report follows the same multi-step editorial process. Nothing is published until every step has completed and passed internal review.

1. 1
   Query selection

   Every report starts from a cross-source analytical query — a SQL question that joins at least two independent data sources. We maintain roughly 100 such queries covering divergences, correlations and anomalies. Source coverage is rebalanced so no single dataset dominates recent output.

2. 2
   Narrative planning

   The most significant finding from the query result becomes the centre of the report. A structured outline is drafted: hook, the specific datapoint, historical context, what it may indicate, caveats, and how it relates to other signals the same dataset exposes.

3. 3
   Analysis

   The report body is written against the outline, grounded entirely in the numbers returned by the query. Claims outside the data we hold are either omitted or explicitly marked as analytical interpretation rather than fact.

4. 4
   Factcheck

   Every numeric claim in the draft is re-queried against the raw dataset to confirm the figure, the units and the time window. Figures that cannot be reproduced from raw data are either corrected or removed.

5. 5
   Correction pass

   Any discrepancies surfaced in the factcheck step are corrected in a second editing pass. Chart placeholders are replaced with concrete Chart.js visualizations derived directly from the same query results.

Verification

Every numeric claim is re-queried against the raw dataset before publication. Figures that cannot be reproduced exactly against a named public source and a specific timestamp do not appear in the final report.

The factcheck step is mandatory — no report bypasses it, regardless of topic. Reports that fail verification are corrected and re-verified before they are queued for publication.

Data freshness and timestamps

Every published report includes a "data snapshot" timestamp indicating the moment our infrastructure captured the figures used. For fast-moving indicators (funding rates, open interest, prices) the snapshot is typically minutes to hours old. For slower-moving indicators (TVL, GitHub activity, FRED releases) the snapshot may be hours to days old depending on the underlying publication cadence.

Snapshot library

Beyond raw data sources, we maintain a library of ~80 named "snapshots" — pre-computed analytical slices that combine and contextualize multiple data sources. Each snapshot returns the same structured payload regardless of when it is queried: current readings, aggregates, and historical context. Reports draw from this library rather than re-querying raw data.

• Price & market structure — BTC daily prices over 30/365 days, top-100 movers (24h/7d/30d/90d), market-cap concentration, drawdown depth and recovery times
• Volatility — rolling 30-day realized volatility, multi-year volatility regimes, RSI(14) for top 10 coins
• Sentiment — Fear & Greed daily values over 180 days, regime streaks, Fear/Greed extreme tally
• Derivatives — top-10 perpetual funding rates, BTC and ETH open interest over 30 days, long/short ratios
• On-chain & flows — whale transfer volumes by chain, exchange inflow/outflow concentration, labeled wallet inventory
• DeFi — TVL by protocol, by chain, total DeFi TVL trend, biggest TVL movers over 30 days
• Developer activity — GitHub commit counts and contributor trends across top 20 crypto projects, declining-activity flags
• Cross-asset — BTC dominance over 90 days, ETH/BTC ratio over 180 days, altcoin season index, BTC+ETH combined dominance
• CR24 Score — 5-component score for tracked coins, distribution across the universe, label transitions over 7 days, top strong/risky lists
• News & events — daily article counts, top headlines per day, upcoming token unlocks and event calendar over 30 days
• Multi-year history — BTC, ETH, SOL, BNB, XRP, ADA, DOGE, TRX, LINK, BCH — all peak-to-trough drawdowns ≥20%, calendar-year returns, multi-year volatility regimes
• Macro — BTC correlations vs DXY, gold and S&P 500 over 90 days (where FRED data is available)

Snapshots refresh every 6 hours from the underlying collectors. The exact timestamp of the snapshot used in any given report is shown in the report's "Data captured" footer.

How to read a CryptoRadar24 report

Every research note follows the same internal structure so you can navigate it quickly regardless of topic.

The at-a-glance indicator panel

Each report opens with a 6-row table sitting between the lede and the first heading. It summarizes the report's key indicators in one block. Columns:

• Indicator — short label of the indicator (e.g. "Peak volume", "BTC dominance")
• Reading — the current data point as cited in the report (e.g. "71.8B USD", "59.54%")
• 30/90-day context — how the current reading compares to its own 30-day or 90-day baseline
• Read — one-word interpretation: Elevated / Contained / Stable / Stretched / Mixed / Weak / Firm / etc.

"What would falsify this read"

Each report ends with a list of conditions under which its analytical reading would be wrong. These are operational anchors, not predictions — they tell you exactly what data shifts would invalidate the article's thesis.

"What to watch next"

Three concrete numeric thresholds the reader can monitor in subsequent data updates, each paired with what crossing it would imply for market structure.

Article types

CryptoRadar24 publishes three types of analytical content. Each follows the same data and editorial process but serves a different reader need:

• Pillar (evergreen) — Long-form research anchors. Each pillar takes a structural theme (e.g. Bitcoin's 12-month market structure, DeFi liquidity concentration) and frames it across 7-9 indicator sections. Pillars live at /research/{topic} URLs without a date suffix and are refreshed periodically with the latest data — they accumulate authority and serve as the SEO foundation of the site.
• Cluster (monthly) — Sub-topics that drill into a specific facet of an active pillar (e.g. "BTC volume analysis" inside the BTC market-structure pillar). Cluster URLs include the publication month — they are point-in-time observations of an active month's data.
• Reactive (event-driven) — Short-form reports that fire when hard market triggers occur (BTC ±10% in 24 hours, Fear & Greed reaching extreme territory, top-10 coin anomalies with high severity). Reactive URLs include the full publication date because the analysis is anchored to a specific event window.

Factcheck stages

Each report passes through two automated factchecks before publication. The first is deterministic: we extract every numeric value, every ticker, every date and every named entity from the polished article and verify that each one appears in the source data the analyst pulled. Anything that appears in the polished version but not in the source data triggers an automatic regeneration.

The second factcheck is semantic: it looks for new factual claims, predictions or comparisons that the polished version added but the source draft did not contain. Drift here is logged but does not block publication — the regex check is the hard gate.

After three failed regeneration attempts, the report is saved as a draft for editorial review and is not published to the public site. This guarantees that no figure, ticker or date in any published report originates from a hallucination — every number traces back to one of the named public sources listed above.

Known limitations

Reports are constrained by the data we collect. We disclose limitations explicitly:

• News headlines come from RSS feeds (CoinDesk, CoinTelegraph, TheBlock, Decrypt, Bitcoin Magazine). These do not include sentiment scoring, so reports do not cite "positive vs negative news ratios" — only article counts and the headlines themselves.
• Macroeconomic correlations (BTC vs DXY, gold, S&P 500) require overlapping FRED data ranges. When FRED has not yet released the indicator we need, the correlation is omitted from the report rather than estimated.
• On-chain whale flow data covers BTC, Ethereum, Solana, Tron and a small set of EVM chains. Coverage on long-tail chains is partial.
• Pre-2017 BTC history is included for full-cycle drawdown context, but pre-institutional cycles (2011, 2013) are noted as range references only — they are not used as primary comparisons for current market structure.
• Reports describe public market data and analyst interpretations of it. They do not constitute investment advice and do not predict future prices.

Corrections and transparency

If you spot an error in a published report, or would like us to disclose the exact data source and timestamp behind a specific figure, please email [email protected]. Verified corrections are applied to the report body and the article is republished with a new "last modified" timestamp.