Where the numbers come from.

The canonical USD price oracle. Daily Bitcoin closes from 28 April 2013 onward; merged with the pre-2013 CSV (below) into the price spine every chart references.

Backs. Every chart on the site reads the price spine for spot or model fits. The direct fits are Rainbow, Power-Law, Stock-to-Flow, Pi Cycle, and 200-Week MA; the pure-compute statistical charts (drawdown, rolling CAGR, volatility, profitable days, halving-cycle overlay, time-in-band, monthly returns) derive from the same series.

Coverage and gaps. Daily-close coverage starts 28 April 2013; the pre-2013 CSV (§04) covers earlier history. The upstream occasionally emits two points for the same UTC day on refresh; we deduplicate by ISO day, last-write-wins.

The on-chain, sentiment, derivatives, ETF, dominance, stablecoin and macro feed. Twenty-eight charts are built from CoinGlass-only inputs; another three (MVRV, MVRV-Z, HODL Waves) combine its holder-supply series with locally-derived market cap.

Backs. Fear & Greed, Puell Multiple, NUPL, ETF Flows, Exchange Flows, AHR999, Reserve Risk, RHODL, SOPR, Realized Price, Golden Ratio, Long/Short, Funding Rate, Open Interest, Liquidations, Dominance, Stablecoin Supply, Altcoin Season, Coinbase Premium, BTC vs M2, Active Addresses, New Addresses, Correlations, MVRV, MVRV-Z, HODL Waves.

Substantive divergences from common references.

• Exchange Flows — coverage starts 1 May 2024; pre-2024 context isn’t available. Coinbase appears under two casings upstream (“Coinbase” vs “Coinbase Pro”); both are normalised to one venue label so the per-day breakdown is internally consistent.
• Long/Short Ratio — no aggregated cross-venue feed exists; the highest-volume venue (Binance BTCUSDT) is shipped as the public-market proxy.
• Funding Rate and Coinbase Premium — rate values are in percent, not fraction (0.0584 = 0.0584%), and that convention is preserved end-to-end. The Coinbase Premium regime is keyed off the 7-day rolling mean to suppress raw daily outage spikes.
• BTC vs M2 — year-on-year M2 growth is derived locally from the supply series (closest observation to the same calendar day one year prior, ±14-day tolerance) rather than the upstream-published growth field, so both baskets stay on one auditable definition.
• Correlations — the upstream Pearson series uses aligned trading days, which runs more extreme than research-desk 90-day tables. Shape is reliable; absolute level is source-specific.
• MVRV and MVRV-Z — realized cap is a two-bucket short-term-holder + long-term-holder approximation of per-UTXO ground truth; tracks within a few percent across the full history. Documented on the chart page.
• NUPL — pre-exchange zero-fill rows from 2009 are dropped so the Y-axis doesn’t compress around zero.

Hash rate, on-chain transaction count, transaction-fees-USD, and estimated-transaction-volume-USD — four free, fifteen-year-deep daily series the rest of the data stack doesn’t ship.

Backs. Hash Rate, Hash Ribbon, Transactions, Fees, NVT.

Substantive divergences from common references.

• NVT — regime thresholds recalibrated to 100 / 220 (from Kalichkin’s canonical 45 / 150) because the estimated-transaction-volume-USD denominator runs about twice as high as Kalichkin’s adjusted-volume figure; the 20 / 80 percentile split on post-warmup history sits at 100 / 220. Both cycle tops still peak well above 220; 2019 and 2022 lows sit around 100.
• Hash Ribbon — the published 30/60 hash-rate cross without Edwards’s price-momentum confirmation filter. The popular price-confirmed variant lags the canonical signal-date list by three to ten weeks on this dataset; readers wanting the strict-Edwards signal can layer the 10/20 price-SMA cross on top.
• Hash Rate — pre-mining-rig days where converted hash rate falls below 1 TH/s are dropped to keep the log axis legible.

A small CSV of 1,015 daily UTC closes from 2010-07-18 (Bitcoin’s first week of Mt. Gox trading) through 2013-04-27, sourced from the CoinMetrics community data. The CoinGecko window starts 2013-04-28; without the early history the regression would lose Bitcoin’s first three years. Newer data wins on overlap.

The chart data published on btc oak is released under CC BY 4.0 — reuse is welcome with attribution to btcoak.com. Upstream feeds remain subject to their providers’ terms; consult each provider’s documentation before redistributing raw upstream data.