Research · Network Economics
Price against what holders paid. Realised cap estimates the aggregate cost basis of Bitcoin by valuing each coin at the price when it last moved. MVRV compares today's market value with that estimated cost basis. The page shows where today's reading sits in history and tests whether that position has historically carried information about future returns.
Reading the daily snapshot (Coin Metrics community data).
Log deviation of price from realised price, on its own distribution over the fit window. Cuts at the 10th, 30th, 70th and 90th percentiles.
| Starting MVRV | 90d | 180d | 365d | Days |
|---|
Descriptive, overlapping windows, in-sample. Not a forecast.
| Horizon | Spearman rank correlation, as-of percentile vs forward return | Median forward return, bottom decile as-of | Median forward return, top decile as-of | Days |
|---|
| Source | Published finding | This page, live |
|---|
Network value: Metcalfe Value Monitor. Long-run time model: Power Law Monitor. This page: what holders paid.
| Series | First | Last | Days | Min | Max | Latest |
|---|
Realised cap sums every unspent coin at the price when it last moved on chain. In this implementation it is not reconstructed from the UTXO set. The Coin Metrics community tier publishes MVRV rather than realised cap itself, so the page recovers realised cap exactly as market cap divided by MVRV (CapMrktCurUSD / CapMVRVCur); the audit panel states which route was used. Realised price divides it by circulating supply. MVRV is market cap over realised cap. The Z-score follows the common definition, market cap minus realised cap divided by the standard deviation of market cap, computed here over the statistics window rather than all history so that the early years do not set the scale. The deviation used for the ribbon, buckets and forward returns is ln(price / realised price).
The out-of-sample test is the honest version of the usual "MVRV called every top and bottom" claim. Each day's percentile is computed only from the data that existed on that day, and the rank correlation between that as-of percentile and the subsequent return is reported at three horizons. A negative correlation means high MVRV preceded low returns without hindsight; a correlation near zero means the historical thresholds were fitted after the fact.
Supply activity bands (share of coins not moved for one, two, five or ten years) are shown when the feed provides them; the community tier does not, so the page shows the count of addresses with a balance in their place and hides the band chart rather than draw it empty. The ten-year band, where available, is the best public proxy for coins presumed lost.
Descriptive: the price-to-cost-basis relationship and its distribution are measured from the chain and reported. Forecasting: only as far as the out-of-sample table supports it, and the correlation is printed. It does not claim: that any MVRV level is a target or a trigger. Realised cap estimates the cost basis of coins, not the cost basis of individual holders; exchange internal transfers and custodial reshuffles move it without any economic change, and Coin Metrics' community tier is not entity adjusted.
Coin Metrics community API (PriceUSD, CapMrktCurUSD, CapMVRVCur, SplyCur, AdrBalCnt; realised cap derived from the first three) through the daily snapshot; Coinbase spot for the live price. The realised cap concept: Coin Metrics, "Realized capitalization," 2018; MVRV: Mahmudov and Puell, 2018; MVRV Z-score: Awe and Yeung, 2018. Cohort refinements (long-term and short-term holder cost basis) are Glassnode constructions and are not in the public tier used here.
Nothing here is investment advice. MVRV extremes have coincided with cycle turns in a sample of four cycles; four is not many.