What Are MVRV, Realized Price, NUPL and SOPR?
On-chain cycle indicators: MVRV, realized price, NUPL and SOPR. How realized cap, realized price, MVRV, MVRV Z-score, NUPL, SOPR, the Puell Multiple and long-term holder supply are built, what they showed at past turns, and where they mislead.
They are on-chain measures built from the price at which each bitcoin last moved. Realized price is the average last-movement price, MVRV is market value divided by realized value, NUPL is the net unrealized gain as a share of market value, and SOPR measures the profit or loss on coins that moved on a given day. Their past peak values have declined each cycle, so fixed thresholds have not held.
You’ll learn
- Define realized cap, realized price, MVRV, MVRV Z-score, NUPL, SOPR, the Puell Multiple and long-term holder supply, and compute each from given numbers
- Attribute each indicator to its original authors where the record is clear, and say where it is not
- Compare indicator readings at past peaks and troughs, and explain why their peak values have declined
- Explain how lost coins, exchange wallet moves, fund custody and look-ahead choices distort these measures
- Assess the 2025–26 case, in which price stayed above realized price, without treating a past threshold as a rule
The bear market that stayed above realized price
In the 2018 and 2022 bear markets, bitcoin’s price spent months below its realized price, a measure of the average price at which each coin last moved on-chain. In the decline that followed the October 2025 peak, it did not. Glassnode’s weekly report of October 14, 2026 stated that this bear market “never closed below” realized price, and that even its lowest daily close, on June 30, 2026, was about 10% above it. Coin Metrics’ independent data give the same picture: on June 30, 2026 the daily reference price was $58,525 and its realized price about $53,070, a ratio of 1.10. Many analysts had treated a dip below realized price as a typical feature of a cycle low. A rule built on three earlier cycles did not repeat in the fourth. This lesson builds the measures from first principles, so that a reader can judge what such a result means.
Market value and realized value
Bitcoin records coins as unspent transaction outputs (UTXOs). Each output has an amount and a creation time, which is the moment its coins last moved. That structure allows a valuation that ordinary markets cannot produce: each coin can be valued at the price on the day it last moved. Summing those values gives an estimate of what holders, in aggregate, paid or last valued their coins at.
Market capitalization: Current supply times current price: Mt = St × Pt.
Realized capitalization: The sum, over all UTXOs, of each output’s amount times the price on the day it was created, that is, the day its coins last moved. Coin Metrics defines it as the sum of USD value based on the closing price on the day each unit last moved.
Realized price: Realized cap divided by supply: the average price at which coins last moved on-chain. It is often described as an aggregate cost basis, but it is a last-movement basis, which is not the same thing (see the limits section).
qi is the amount in output i, c(i) is the day output i was created, Pc(i) is the closing price that day, and St is total supply.
Realized cap changes only when coins move. If an output created at $20,000 is spent today at $50,000, the new outputs are valued at $50,000, and realized cap rises by the difference times the amount. If coins created at $60,000 move at $50,000, realized cap falls. Coins that never move keep their old valuation forever, so coins lost in 2010 count at a price near zero.
Attribution. Glassnode’s documentation credits Antoine Le Calvez with coining realized capitalization in September 2018 and links to Coin Metrics’ article “Introducing Realized Capitalization.” Other accounts describe the measure as developed by Nic Carter together with Le Calvez, and Carter presented it publicly in 2018. The Coin Metrics article could not be opened for this lesson, so the exact division of credit is left open here.
MVRV and the MVRV Z-score
MVRV ratio: Market value to realized value: MVRVt = Mt / RCt, which equals Pt / RPt. Glassnode credits Murad Mahmudov and David Puell, October 2018.
MVRV has a direct reading. MVRV − 1 is the average unrealized gain per dollar of realized value. An MVRV of 2 means coins, in aggregate, are marked at twice the price at which they last moved. An MVRV below 1 means the average coin is marked below its last-movement price, so the network as a whole holds an unrealized loss. Glassnode’s documentation describes values above 3.5 as having generally accompanied late-stage bull markets and values below 1.0 as typical of capitulation. Those are descriptions of past behavior, not rules.
Example: MVRV and realized price from Coin Metrics data, November 20, 2026
- Market cap = $1,183.8 billion. Realized cap (market cap ÷ Coin Metrics’ MVRV) = $1,071.4 billion. Supply = 20,114,895 bitcoin.
- MVRV = 1,183.8 / 1,071.4 = 1.105.
- Realized price = 1,071.4 billion / 20,114,895 = $53,262. Check: price ÷ realized price = 58,851 / 53,262 = 1.105.
- Average unrealized gain = MVRV − 1 = 10.5% of realized value.
- Compare October 6, 2025: market cap $2,487.9 billion, realized cap $1,087.8 billion, MVRV 2.29.
MVRV Z-score: (Market cap − realized cap) ÷ standard deviation of market cap. Glassnode credits Awe & Wonder, October 2018, in a Medium article titled “Introducing … The Bitcoin ‘MVRV Z’ Metric That Predicts Market Tops with 90%+ Accuracy.” Glassnode’s page does not state the window used for the standard deviation.
The Z-score rescales the gap between market and realized value by the history of market cap’s variability. The choice of window matters a great deal. If the standard deviation is computed over the full history, including data after the date being scored, the value for a past date uses information that did not exist then. That is look-ahead bias, and it makes past readings look cleaner than anyone could have seen at the time. A version that uses only data up to each date, called an expanding window, is honest but gives different numbers. The table below uses an expanding window.
NUPL
Net unrealized profit/loss (NUPL): (Market cap − realized cap) ÷ market cap. Glassnode credits Tuur Demeester, Tamás Blummer and Michiel Lescrauwaet of Adamant Capital, February 2019, in “A Primer on Bitcoin Investor Sentiment and Changes in Saving Behavior.”
NUPL is the share of market value that is unrealized profit, net of unrealized loss. It is above zero when the network holds a net gain and below zero when it holds a net loss.
Because NUPL is a function of MVRV, it carries no new information. It changes the scale: MVRV runs from 0 to any positive number, while NUPL is bounded above by 1. An MVRV of 2 gives a NUPL of 0.5; an MVRV of 4 gives 0.75; an MVRV of 0.75 gives −0.33. The bounded scale is why NUPL is often drawn with named bands, and why its upper readings compress: going from MVRV 4 to 8 only moves NUPL from 0.75 to 0.875.
SOPR: profit and loss realized when coins move
MVRV and NUPL describe the stock of unrealized gains. SOPR describes the flow: the profit or loss locked in by coins that actually moved on a given day.
Spent output profit ratio (SOPR): For all outputs spent in a period, the USD value at the time of spending divided by the USD value at the time of creation. Glassnode credits Renato Shirakashi, April 2019, “Introducing SOPR: spent outputs to predict bitcoin lows and tops.”
Above 1, coins moved that day were, on average, valued above their creation price. Below 1, they moved at a loss. Exactly 1 is break-even.
Example: SOPR for one day
- Example: on a day when the price is $50,000, two outputs are spent.
- Output A: 1 coin created at $40,000. Value created $40,000, value spent $50,000.
- Output B: 2 coins created at $60,000. Value created $120,000, value spent $100,000.
- SOPR = (50,000 + 100,000) / (40,000 + 120,000) = 150,000 / 160,000 = 0.9375.
- Now add an exchange moving 10 of its own coins, created at $20,000, between its own wallets. Numerator +500,000, denominator +200,000. SOPR = 650,000 / 360,000 = 1.81.
Glassnode’s adjusted SOPR (aSOPR) removes outputs with a lifespan under one hour, because very short-lived outputs are mostly relays and change, not changes of ownership. That filter handles noise from fast-moving outputs. It does not handle the case in the example, where old coins move inside one custodian. That requires entity labeling, which is an estimate.
The interpretation commonly attached to SOPR is behavioral. In rising markets, a fall to 1 means holders stopped moving coins at break-even. In falling markets, a value below 1 means losses were being taken. These are descriptions of who moved coins. SOPR does not show who received them or why.
The Puell Multiple and long-term holder supply
Puell Multiple: Daily issuance value in USD divided by its 365-day moving average. Glassnode credits David Puell, March 2019.
The Puell Multiple measures miner revenue from new coins against its trailing year. It rises when price rises fast and falls when price falls or when a halving cuts issuance. A halving cuts it roughly in half overnight by construction, as Glassnode’s own documentation notes. So any comparison of Puell readings across a halving compares different regimes.
Example: the Puell Multiple across the 2024 halving
- Coin Metrics issuance value on April 19, 2024: $59.38 million. 365-day average: $35.38 million. Puell = 59.38 / 35.38 = 1.68.
- April 21, 2024: $28.42 million against an average of $35.39 million. Puell = 0.80.
- Price barely changed between these dates. The drop came from the subsidy cut.
- November 20, 2026: $26.85 million against $33.81 million. Puell = 0.79.
Long-term holder (LTH) supply: In Glassnode’s classification, supply held in coins older than about 155 days, a threshold after which its historical analysis found coins increasingly unlikely to be spent. Coins younger than that are short-term holder supply. Glassnode smooths the boundary with a transition 10 days wide around 155 days, and applies it to entities, not raw addresses.
LTH supply tends to rise in long declines and quiet periods, as coins age past the threshold, and to fall in strong rallies, when old coins move. A fall in LTH supply is an observation that old coins moved. It is not, by itself, evidence of a sale: the movement may be an internal transfer, as the limits section shows.
Readings at past peaks and troughs
The table gives readings at each daily-reference price peak and the lowest daily reference price after it, from Coin Metrics data, computed for this lesson. Glassnode and other providers use different price feeds, supply definitions and entity methods, so their values differ somewhat.
| Date | Event | MVRV | NUPL | Z-score (expanding window) | Puell |
|---|---|---|---|---|---|
| Dec. 4, 2013 | Peak | 4.72 | 0.79 | 7.71 | 8.22 |
| Jan. 14, 2015 | Trough | 0.56 | −0.77 | −0.60 | 0.31 |
| Dec. 16, 2017 | Peak | 4.43 | 0.77 | 9.36 | 6.35 |
| Dec. 15, 2018 | Trough | 0.69 | −0.45 | −0.49 | 0.39 |
| Nov. 8, 2021 | Peak | 2.85 | 0.65 | 3.53 | 1.47 |
| Nov. 9, 2022 | Trough | 0.75 | −0.33 | −0.36 | 0.48 |
| Oct. 6, 2025 | Peak | 2.29 | 0.56 | 2.53 | 1.07 |
| June 30, 2026 | Lowest so far | 1.10 | 0.09 | 0.18 | 0.66 |
Source: Coin Metrics community data (CapMVRVCur, CapMrktCurUSD, IssTotUSD), retrieved November 21, 2026. NUPL is computed as 1 − 1/MVRV. The Z-score uses the standard deviation of market cap from July 2010 up to each date. Puell uses a 365-day average including the current day.
Four patterns stand out. First, peak readings declined cycle after cycle: MVRV went from 4.72 to 4.43 to 2.85 to 2.29 at the price peaks. Second, the highest MVRV of each cycle did not always fall on the price peak. It was 5.88 on November 18, 2013, 4.72 on December 7, 2017, 3.96 on February 21, 2021 and 2.78 on March 11, 2024, the last two months or more before their cycle’s price peak. Third, in the 2021 and 2025 cycles MVRV never reached the 3.5 level that Glassnode describes as typical of late bull markets. Fourth, trough readings rose: MVRV was below 1 on 304 days in the 2014–15 decline, 134 days in 2018–19, 179 days in 2022–23, and on no day through November 20, 2026.
Glassnode’s own review of MVRV (Checkmate, 2023) records an early example of the second pattern: between the two 2013 peaks, price rose a further 392% while MVRV fell from 5.74 to 5.43. Its numbers differ slightly from Coin Metrics’, which illustrates the point about methods. The same review reports that, since 2017, MVRV was below 1.0 on about 15% of days and above 3.2 on about 6%.
Computing the measures yourself
The example below uses invented numbers for five outputs. It computes market cap, realized cap, realized price, MVRV and NUPL, and checks the identity NUPL = 1 − 1/MVRV.
Example: what lost coins do to the same ledger
- Suppose the 0.5 coins last moved at $1 are in fact lost. Usable supply = 9.5 coins.
- Market cap on usable supply = 9.5 × 50,000 = $475,000. Realized cap barely changes, because the lost coins counted at $0.50 in total: $349,999.50.
- MVRV = 475,000 / 349,999.5 = 1.357, against 1.429 with the lost coins counted.
Where on-chain indicators mislead
Last movement is not cost basis. Realized value records the last time coins moved on-chain, not what any owner paid. A person who bought on an exchange at $60,000 and never withdrew appears nowhere in the UTXO set as an individual; the exchange’s coins carry whatever creation price its own wallet moves gave them.
Exchange and custodian wallet moves. A custodian that reorganizes its wallets moves old coins without any sale. In late November 2025, Coinbase moved about 798,636 bitcoin in 286 internal transactions, from older address types to newer ones. Bitcoinist, citing analyst Axel Adler, reported that this showed up as a large apparent long-term holder distribution, including a phantom $68 billion spike in realized profit, along with distortions in coin age bands and exchange netflow. The example in the SOPR section shows the mechanism in miniature. Entity labels can correct for such moves only if the provider knows which wallets belong to the custodian.
Funds move ownership off-chain. When shares of an exchange-traded fund change hands on a stock exchange, no coins move. When a fund creates or redeems shares, coins may move between the custodian and other parties, at times set by fund operations rather than by investors’ decisions. The more supply sits in funds and large custodians, the less of the ownership change shows up in UTXO data. On-chain realized price then describes a shrinking part of the market.
Look-ahead and refitting. Thresholds such as MVRV above 3.5 or NUPL above 0.75 were read off charts that already contained the peaks they mark. A Z-score whose standard deviation uses the full history uses future data for past dates. The honest test is a rule written down before the outcome, applied without changes, on data available at each date.
Few events. The record has three or four peaks and troughs. Any claim that an indicator marked every top is a claim about a handful of events chosen after the fact, with the multiple-testing problem discussed in the previous lesson.
Method differences. Providers differ in price feeds, the treatment of lost and very old coins, entity clustering and time zones. A reading from one provider should be compared with that provider’s own history, not with a threshold quoted from another.
Do fixed thresholds still mean anything?
The case that they do. Across four cycles, MVRV and NUPL were high near every peak and low near every trough, in the sense of being near the top or bottom of their own range for that cycle. The indicators measure something real, the aggregate gap between market value and last-movement value, which plausibly relates to holders’ willingness to realize gains or losses. Relative versions, such as a percentile within a rolling window, can preserve that information when the level shifts. The case that they do not. Every fixed level drawn from the early cycles failed later. The 2021 and 2025 peaks came below 3.5 on MVRV, the 2026 decline never reached 1, and the cycle’s highest MVRV came before the price peak in both 2021 and 2024–25. Structural change, including fund custody and large custodians, means the UTXO set reflects a different market than it did in 2013. An indicator whose thresholds must be reset each cycle is hard to tell apart from one fitted after the fact. What can be said. The measures are useful descriptions of the current state of on-chain profit and loss. Their past threshold crossings are a small, non-repeating sample. Treat them as descriptions to compare across time with stated methods, not as levels that must recur.
Common mistakes
- Calling realized price “the price everyone paid.” It is the average last-movement price of coins on-chain, and coins on exchanges or in funds do not reflect their owners’ purchases.
- Treating NUPL as a separate confirmation of MVRV. It is a transform of the same ratio.
- Reading a high SOPR or a drop in long-term holder supply as selling without checking for custodian wallet moves.
- Comparing Puell readings across a halving as if the regime were unchanged.
- Quoting a Z-score without saying how its standard deviation was computed.
- Applying another provider’s threshold to your own computed series.
- Concluding that a bottom has or has not occurred because a past threshold was or was not reached.
Problem set
- An example network has 100 coins at a price of $30,000. Its realized cap is $2.4 million. Compute market cap, realized price, MVRV and NUPL.
Show solution
Market cap = 100 × 30,000 = $3.0 million. Realized price = 2,400,000 / 100 = $24,000. MVRV = 3.0 / 2.4 = 1.25. NUPL = (3.0 − 2.4) / 3.0 = 0.20, which equals 1 − 1/1.25 = 0.20. - In the same network, 10 coins created at $40,000 are spent at $30,000. No other coins move. Compute the new realized cap, MVRV and that day’s SOPR.
Show solution
Realized cap changes by 10 × (30,000 − 40,000) = −$100,000, to $2.3 million. MVRV = 3.0 / 2.3 = 1.304. SOPR = (10 × 30,000) / (10 × 40,000) = 0.75. Realizing a loss lowered realized cap and so raised MVRV at an unchanged price. - Find the NUPL values that correspond to MVRV = 0.8, 1, 2.29 and 4.72. Why does NUPL change less between 2.29 and 4.72 than MVRV does?
Show solution
NUPL = 1 − 1/MVRV: 0.8 gives −0.25; 1 gives 0; 2.29 gives 0.563; 4.72 gives 0.788. NUPL is bounded above by 1, so high MVRV values are compressed: a doubling of MVRV moves NUPL only part of the remaining distance to 1. - Daily issuance value is $40 million and its 365-day average is $32 million. The next day a halving occurs and price is unchanged. Estimate the Puell Multiple before and after, assuming the average barely moves in one day.
Show solution
Before: 40 / 32 = 1.25. After: issuance value halves to about $20 million, so Puell = 20 / 32 = 0.625. The change is purely mechanical, which is why a low post-halving reading cannot be compared directly with a low reading in a year with no halving. - A custodian moves 50,000 coins, last moved when the price was $10,000, between its own wallets when the price is $60,000. What happens to realized cap and to that day’s SOPR if the move is counted? What correction is needed?
Show solution
Realized cap rises by 50,000 × (60,000 − 10,000) = $2.5 billion, although no one bought or sold. That day’s SOPR gains $3.0 billion in the numerator and $0.5 billion in the denominator, pushing it sharply above 1. Long-term holder supply also falls, because old coins became new outputs. The correction is to identify both wallets as one entity and exclude the transfer, which depends on accurate custodian labeling. - Example: suppose a rule, written in 2018 from 2011–2017 data, says MVRV above 3.7 marks a bitcoin top. What would a fair out-of-sample test look like, and what do the 2021 and 2025 peaks show?
Show solution
Fix the threshold and the data source as of the date the rule was published, then check only later events. In Coin Metrics data, MVRV peaked at 3.96 in February 2021 and was 2.85 at the November 2021 price peak and 2.29 at the October 2025 peak. The rule crossed early in 2021 and did not cross at all in 2024–25, so it did not mark either later price peak.
References
- Glassnode. Realized Capitalization. Metric guide, read November 2026. Source
- Glassnode. MVRV Ratio. Metric guide, read November 2026. Source
- Glassnode. MVRV-Z Score. Metric guide, read November 2026. Source
- Glassnode. NUPL (Net Unrealized Profit/Loss). Metric guide, read November 2026. Source
- Glassnode. SOPR (Spent Output Profit Ratio). Metric guide, read November 2026. Source
- Glassnode. aSOPR (Adjusted SOPR). Metric guide, read November 2026. Source
- Glassnode. Puell Multiple. Metric guide, read November 2026. Source
- Glassnode. Supply Held by Long and Short-Term Holders. Metric guide, read November 2026. Source
- Checkmate and Glassnode (February 22, 2023). Mastering the MVRV Ratio. Glassnode Insights. Source
- Theissen, F., and Glassnode (October 14, 2026). Boxed In. The Week On-chain, Week 41, 2026. Source
- Coin Metrics. MVRV (CapMVRVCur) and Realized Market Cap (CapRealUSD). Metric definitions, Coin Metrics knowledge base. Source
- Coin Metrics. Community API, asset metrics CapMVRVCur, CapMrktCurUSD, SplyCur, PriceUSD and IssTotUSD for BTC, daily, retrieved November 21, 2026. Source
- Villafuerte, S. (updated November 27, 2025). Coinbase Wallet Rebalancing Creates False $68B LTH Distribution Signal. Bitcoinist. Secondary report citing Axel Adler. Source
Exercise
- Open the link and add start_time and end_time parameters for a month of your choice, for example start_time=2022-11-01&end_time=2022-11-30.
- For three days, compute realized cap as CapMrktCurUSD ÷ CapMVRVCur and realized price as realized cap ÷ SplyCur. Check that PriceUSD ÷ realized price equals CapMVRVCur.
- Compute NUPL for the same days and confirm that it equals 1 − 1/MVRV.
- Pull IssTotUSD for the 365 days before a date and compute the Puell Multiple yourself. Repeat for a date a few days after April 20, 2024, and explain the difference.
- Write down, before checking, a rule using MVRV that you would apply to future data. State the data source and threshold, and the outcome that would count against it.
Check yourself
1. Why does realized cap fall when coins last moved at $60,000 are moved again at $50,000?
2. MVRV is 2.0. What is NUPL, and what new information does it add?
3. What is the main risk of computing a historical MVRV Z-score with the standard deviation of the full history?
4. An exchange moves old coins between its own wallets. What is the likely effect on raw SOPR and long-term holder supply?
5. Why does the Puell Multiple roughly halve at a halving even if price is unchanged?
6. Lost coins bias MVRV in which direction, and why?
7. In Coin Metrics data, the highest MVRV of the 2020–21 and 2024–25 cycles came before the price peak. What does this imply for a rule that marks tops when MVRV is at its cycle high?
8. The 2025–26 decline never closed below realized price. Which conclusion is best supported?
Common questions
What is realized price in Bitcoin?
Realized cap, the sum of each coin valued at the price on the day it last moved, divided by supply. It is an average last-movement price, which differs from what owners actually paid when coins sit on exchanges or with fund custodians.
How is MVRV calculated?
Market cap divided by realized cap, which equals price divided by realized price. A value above 1 means the average coin is marked above its last-movement price; below 1 means an aggregate unrealized loss.
Why can on-chain indicators give false readings?
Lost coins inflate market cap, custodians moving coins between their own wallets look like old coins being spent, fund shares trade without moving coins, and thresholds read off past charts carry look-ahead bias.
This lesson teaches methods and describes past events. It is not investment advice. Live panels: tracked whale wallets, updated 2026-11-22 07:40 UTC.