Editorial
Methodology
- Written by
- The Bitcoin Data Guide editorial team
- Reviewed by
- Data review desk
- Dates
- Published 2026-01-12 · Last reviewed 2026-10-01
Fact
Every published figure descends from one canonical dataset, and each derived series is computed from it by a stated rule.
Forecast
Where a page reads a pattern in the record — a cycle, a trend, a comparison — that reading is interpretation, not a fact, and is labelled as such.
One canonical dataset
Every curated figure on this site descends from a single canonical dataset. There is one record of Bitcoin's price history, issuance schedule, halving cadence and cycle statistics, and it lives in one place. Articles, tables, charts and research pages all read from it rather than carrying their own copies. That is a deliberate constraint: when two pages quote the same number, they are quoting the same source, so they cannot drift apart.
The dataset is held as fixed constants rather than recomputed on each request. A figure changes only when the dataset is deliberately republished, which is why each table carries a vintage range rather than a fetch time. The live series — the spot price and the daily candles — are the one exception, and they are kept visibly apart from the curated record because they are readings rather than settled history.
The provenance, vintage and licence of every dataset are recorded in the source register on the data sources and methodology page, which is served from the same metadata the tables themselves carry.
How derived series are computed
Most of the numbers a reader sees are not stored directly. They are computed from the canonical dataset by a fixed rule, so the same input always produces the same output.
Yearly returns
The change from the first to the last attested price in a calendar year, expressed as a percentage. No figure is interpolated: a year with no attested close does not appear.
Drawdowns and recoveries
The fall from a running peak to the subsequent trough, measured peak to trough, and the number of days from that trough back to the prior peak. Both are read from the daily series.
Rolling windows
Returns over a fixed window — four years, for example — computed at every start date the record supports, so the spread of outcomes is visible rather than a single headline number.
Risk-adjusted ratios
A return divided by the volatility taken to produce it, computed over the same window as the return so the two figures describe the same period.
Because the derived series are computed rather than stored, correcting a canonical figure recomputes everything downstream of it. A change to a single yearly close propagates to the returns, the drawdowns and the rolling windows that depend on it, and the affected vintages move forward together.
How sources are attributed
Every figure is traceable to a published source, and every source is named. A table carries the source it was compiled from and the period it covers; a research article names the datasets it rests on. A reader who wants to check a number can follow the chain from the sentence to the table to the upstream record.
Where two sources disagree, the disagreement is recorded rather than averaged away, because the fact that a number is contested is itself useful information. Where a claim rests on a protocol fact — a block height, a subsidy amount, a halving date — it is treated as verifiable against the block chain and stated without hedging. Where a claim rests on interpretation, it is written as interpretation with the reasoning shown.
The sourcing standard itself is set out on the editorial standards page, and the register of every dataset is on the data sources page.
Review and update cadence
Nothing is published without passing a review gate, and nothing stays published without being revisited.
1. Compile against a source
A figure is entered only with the source it came from and the period it covers. A figure with no traceable source is not published.
2. Cross-check where possible
Where two independent sources cover the same period, they are compared. A disagreement is recorded rather than resolved by averaging.
3. Review before publication
Every page is read by a second person on the team before it goes live, against the sourcing standard and the fact-versus-forecast rule.
4. Revisit on a schedule
Curated material is reviewed each quarter and whenever a source is revised. A material update moves the vintage forward and is dated.
The review status of every published page, with the date it was last reviewed and the reason for the status it received, is recorded on the content quality page.
Fact, forecast and opinion
The most important editorial decision this site makes is the line between what can be stated and what cannot. A measured fact is something a source recorded or the arithmetic of the record shows: a yearly return, a drawdown measured peak to trough, a halving date fixed by the protocol. These are stated plainly.
A forecast is a claim about what will happen next, and the site does not make one. Where a page reads a pattern in the past — that volatility has fallen as the market deepened, that the four-year rhythm has been descriptive rather than mechanical — that reading is interpretation, and it is written as interpretation with the evidence and the counter-argument alongside it. A reader should be able to disagree with the reading while still finding the figures useful.
The full statement of what the site will and will not claim is on the editorial standards page, and how an error is handled is on the corrections policy page.
Continue reading
- Data Sources & MethodologyWhere every figure comes from, its vintage, and the honest limits of the compiled record.
- Editorial StandardsHow the writing is sourced, what the site will not claim, and how it handles being wrong.
- About Bitcoin Data GuideWhat the site is, who publishes it, and the editorial orientation behind it.
- Content QualityThe review status of every published page, with the date it was last reviewed.
Sources
- Canonical curated dataset
- The single source of truth for every curated figure on the site, held as module constants in the canister.
- Derived-series module
- Computes returns, drawdowns, recoveries and rolling windows from the canonical dataset by a fixed rule.
- Coinbase Exchange public API
- The live BTC/USD spot price and daily candles, fetched by the canister and cached for a few minutes.
Related reading
- Research BriefingsDated research notes on the curated record, each naming the datasets it draws on.
- Network StoriesDated, sourced accounts of what happened on the Bitcoin network and how it was measured.
- Institutional NotesHow institutions hold, account for and report bitcoin, drawn from public filings.
- Content QualityThe review status of every published page, with the date and the reason for it.