On-chain metrics explained: active addresses, exchange flows, MVRV and their limits

On-chain metrics are indicators built from the public record of a blockchain: how many addresses are active, how many coins move into and out of exchanges, and at what price coins last changed hands. The best-known include active addresses, exchange net flows, realised price and the MVRV ratio, which compares the market value of a coin's supply with the value at which those coins last moved. They give a view of holder behaviour that price charts cannot, but the data is noisy, depends on assumptions about who owns which address, and misses most trading that happens off-chain.

This guide explains what the main metrics measure, works through how a few of them are calculated, and spends as much time on the limits as on the metrics, because the limits are what most charts leave out.

Why on-chain data exists at all

Bitcoin, Ethereum and most other public blockchains record every confirmed transaction in a ledger that anyone can download and read. The planned guide to blockchains explains how that ledger works. Because the data is public, analysts can measure things that are invisible in traditional markets. Nobody can see every movement of every euro between bank accounts, but anyone can see every movement of bitcoin between addresses.

What the ledger records is limited, though:

• Addresses, not people. An address is a string derived from a key. One person can control thousands of addresses, and one address (an exchange's, for example) can hold coins for millions of people.
• Transfers, not trades. The ledger shows coins moving from one address to another. It does not say whether that was a sale, a gift, a payment, or someone moving coins between their own wallets.
• Only on-chain activity. When you buy bitcoin on a centralised exchange, the trade is recorded in the exchange's internal database, not on the blockchain. Only deposits and withdrawals appear on-chain. Our guide to crypto wallets and custody explains the difference between holding coins yourself and holding a balance on an exchange.

Every on-chain metric is an attempt to infer behaviour from this partial record. Data providers make different assumptions, so the "same" metric can differ between them.

Activity metrics

Active addresses

The number of unique addresses that sent or received a transaction during a period, usually a day. It is used as a rough gauge of network use: more active addresses might suggest more people using the network.

The limits are significant. Exchanges and payment processors batch many users into a few addresses, which undercounts users. Individuals and automated services often use a new address for every transaction, which overcounts them. Spam and low-value automated transfers can inflate the number cheaply on networks with low fees. A change in active addresses can reflect a change in wallet software or exchange practice rather than a change in the number of people.

Transaction count and transfer volume

The number of transactions and the total value moved on-chain. Raw transfer volume is heavily distorted by "change" outputs (on Bitcoin, the part of a transaction that returns to the sender) and by coins shuffled between addresses owned by the same entity. Providers publish "adjusted" versions that try to strip these out, using heuristics that cannot be perfect.

Fees and block space demand

On networks where users bid for space in blocks, total fees show how much people are paying to use the chain. Rising fees mean demand for block space has risen, but that demand can come from a single popular application or a wave of speculation in a new token type. The planned guide to gas fees explains how fees work on Ethereum.

Hash rate (proof-of-work chains)

The total computing power securing a proof-of-work network such as Bitcoin. It is a measure of security and of miner investment, not of demand. It is estimated from how quickly blocks are found rather than measured directly, so daily figures are noisy. The planned guide to proof of work vs proof of stake explains the difference between the two security models.

Exchange flow metrics

These try to measure coins moving onto and off centralised exchanges.

• Exchange inflow: coins sent to addresses identified as belonging to exchanges.
• Exchange outflow: coins sent from those addresses to others.
• Net flow: inflow minus outflow.
• Exchange reserve: the total balance held in identified exchange addresses.

The usual interpretation is that coins moving onto exchanges may be about to be sold, and coins leaving exchanges are probably being moved into longer-term storage. Large net inflows are therefore often described as a sign of selling pressure, and steady outflows as a sign of holding. That is a plausible story, not a rule. Coins also move onto exchanges to be used as collateral for derivatives, to be lent, or simply because a large holder changed custodian.

Worked example: a misleading net inflow

Illustrative numbers only. On one day, a data provider records:

• Inflows to exchange addresses: 12,000 BTC.
• Outflows from exchange addresses: 10,000 BTC.
• Net flow: 12,000 minus 10,000 = +2,000 BTC, reported as a large net inflow.

Now suppose 3,000 BTC of the inflow came from an address that belongs to the same exchange but had not yet been labelled as such, for example a cold wallet being moved to a new address. Correctly labelled, that was an internal transfer and should be excluded:

• Inflows: 12,000 minus 3,000 = 9,000 BTC.
• Net flow: 9,000 minus 10,000 = minus 1,000 BTC, a net outflow.

The headline flipped sign because of a labelling question. This is not hypothetical in kind: exchange wallet reorganisations and newly discovered exchange addresses regularly cause revisions to historic flow data. When a flow chart shows a sudden spike, check whether the provider has attributed it to a single entity.

Stablecoin flows

The same idea is applied to stablecoins: rising stablecoin balances on exchanges are sometimes read as "dry powder" available to buy crypto. Stablecoins are also used for payments, settlement between firms and as collateral, so the reading is ambiguous. Our guide to stablecoins explains how they work.

Valuation metrics: realised cap, realised price and MVRV

These are the metrics most often quoted in discussions of market cycles.

Market cap

Market capitalisation is the current price multiplied by the number of coins in circulation. It values every coin at today's price. The planned guide to crypto market cap and dominance covers circulating versus fully diluted supply.

Realised cap

Realised capitalisation values each coin at the price when it last moved on-chain, then adds them up. A coin that last moved when bitcoin was 20,000 dollars counts at 20,000 dollars, regardless of today's price. The idea is to approximate the total amount holders paid, their aggregate cost basis. Coins that have not moved for years, including coins that are probably lost, count at the old price at which they last moved.

Realised price

Realised cap divided by the number of coins. It is an estimate of the average price at which the current supply was acquired.

MVRV

MVRV stands for market value to realised value: market cap divided by realised cap. An MVRV of 1 means the market price equals the average cost basis. Above 1, holders are on average sitting on unrealised gains; below 1, on average losses. MVRV minus 1 gives the average unrealised gain or loss as a fraction.

Worked example: a three-coin network

Illustrative numbers only. Imagine a network with just three coins and a current price of 60,000 dollars.

• Coin A last moved at 10,000 dollars.
• Coin B last moved at 40,000 dollars.
• Coin C last moved at 70,000 dollars.

Then:

• Market cap = 3 times 60,000 = 180,000 dollars.
• Realised cap = 10,000 + 40,000 + 70,000 = 120,000 dollars.
• Realised price = 120,000 divided by 3 = 40,000 dollars.
• MVRV = 180,000 divided by 120,000 = 1.5, meaning holders are, on average, 50% in profit.

Note what the average hides. The holder of coin C is at a loss of about 14%, while the holder of coin A is up 500%. Their behaviour in a sell-off may be completely different. Providers address this by splitting the metric by how long coins have been held (short-term versus long-term holders), but the split itself rests on a chosen cut-off.

How MVRV is used, and why to be careful

Analysts have noted that, in bitcoin's history, very high MVRV readings have coincided with cycle peaks and readings below 1 with cycle lows. One major data provider describes readings above about 3.5 as historically associated with tops and readings below 1 with bottoms. The word to notice is "historically": those thresholds were observed across a handful of cycles, and the market's structure has changed since. A threshold that fitted four peaks may not fit the fifth. Our guide to how to backtest a strategy explains why rules fitted to few examples tend to fail.

There is also a structural drift. As more of bitcoin's supply is held by funds and long-term holders, and as lost coins sit in realised cap at very old prices, the "normal" level of MVRV may move over time.

Profit and loss metrics: SOPR and NUPL

SOPR

The spent output profit ratio looks only at coins that moved on a given day. For each one, it compares the value when it moved with its value when it was previously moved. The aggregate SOPR is the total value of coins moved divided by their total value at their previous movement. Above 1, coins moved that day were, on average, moved at a gain; below 1, at a loss.

Illustration: on one day, one coin bought at 50,000 dollars and one bought at 70,000 dollars are both moved at 60,000 dollars. Total value moved: 120,000 dollars. Total previous value: 50,000 plus 70,000 = 120,000 dollars. SOPR = 1.0, even though one holder realised a 20% gain and the other a 14% loss. A transfer between someone's own wallets counts as "spent" too, which is why providers publish adjusted versions.

NUPL

Net unrealised profit and loss is market cap minus realised cap, divided by market cap. In the three-coin example it is (180,000 minus 120,000) divided by 180,000, about 0.33. It carries the same information as MVRV, expressed differently, and has the same limits.

Network value to transactions (NVT)

NVT divides market cap by the daily value transferred on-chain, by analogy with a price-to-earnings ratio: a high reading means the network is valued highly relative to how much it is used for transfers. The weakness is the denominator. On-chain transfer volume is distorted by internal movements, and a growing share of economic activity takes place off-chain, on exchanges, on layer 2 networks and inside funds, so NVT can drift upward without any change in real use. The planned guide to layer 2 scaling explains why activity has moved off the main chains.

Ethereum-specific metrics

Ethereum uses an account model rather than Bitcoin's model of discrete outputs, which changes how some metrics are calculated. It also has its own measures:

• Staked ether. The amount locked in staking to secure the network. The planned guide to ether staking covers it.
• Fees burned. Since an upgrade in August 2021, part of every transaction fee on Ethereum is destroyed. Combined with new issuance to validators, this determines whether total supply rises or falls over a period. Our guide to Ethereum and altcoins explains why ether has no fixed supply cap.
• Smart contract activity. Value locked in decentralised finance applications and the number of contract interactions are often quoted, but they can be inflated by the same funds being deposited in several places at once.

The big limits, in one place

1. Most trading is invisible. Trades between customers of a centralised exchange are recorded in the exchange's books, not on the blockchain. Price is set mostly by that off-chain trading and by derivatives, which on-chain data does not see.
2. Entity labelling is guesswork. Exchange flows, holder cohorts and "whale" metrics all depend on clustering addresses into entities. The heuristics are good but not perfect, and they are periodically revised, sometimes changing the history.
3. Funds and custodians compress many holders into few addresses. Coins held by exchange-traded funds sit with custodians. On-chain, millions of fund investors may look like a handful of very large, very inactive holders. The planned guide to bitcoin ETFs explains how those products hold coins.
4. Few cycles to calibrate on. Thresholds such as "MVRV above 3.5" come from a small number of past peaks. That is the same small-sample problem that affects cycle theories.
5. Metrics move together. Many on-chain metrics are transformations of price and realised cap. Seeing several of them "agree" is often one piece of information counted several times.
6. Different providers, different numbers. Methods differ, so compare like with like and read the methodology.
7. Wrapped and bridged coins. Bitcoin represented as a token on another chain does not appear in Bitcoin's on-chain data in the same way, and the same asset can be counted on two networks.

How traders use on-chain data sensibly

On-chain metrics are best treated as background on holder behaviour and market structure, alongside price, liquidity and macro conditions, rather than as entry signals. Some practical habits:

• Read the provider's definition and methodology before drawing conclusions.
• Check whether a sudden spike has a single-entity explanation (an exchange moving wallets, a fund changing custodian).
• Prefer long-term trends over day-to-day readings, which are noisy.
• Never size a leveraged trade on the strength of an on-chain reading. Size from the stop, as the crypto stop loss placement guide explains, with the crypto position size calculator.

Common mistakes

• Treating addresses as people, and concluding that "new users" have arrived when exchanges have changed their wallet set-up.
• Reading every exchange inflow as selling. Coins move onto exchanges for many reasons.
• Using fixed MVRV or similar thresholds as buy or sell triggers, when they were fitted to a handful of cycles.
• Ignoring revisions. Historic data can change when new exchange addresses are labelled.
• Comparing metrics across providers as if they were the same series.
• Forgetting the off-chain market. Derivatives and exchange order books often move price far more than on-chain flows.

Frequently asked questions

What is the MVRV ratio?

MVRV is market value divided by realised value. Market value is price times supply; realised value adds up each coin at the price when it last moved on-chain, approximating holders' cost basis. Above 1, holders are on average in profit; below 1, on average at a loss. Historic thresholds associated with tops and bottoms come from only a few cycles and are not reliable signals.

What does it mean when bitcoin flows into exchanges?

It means coins were sent to addresses identified as belonging to exchanges. Analysts often read large inflows as possible selling pressure, but coins also move to exchanges to serve as collateral, to be lent or because of internal wallet reorganisations. The figure also depends on the provider correctly labelling exchange addresses.

What is realised price?

Realised price is realised cap divided by the number of coins: an estimate of the average price at which the current supply last moved, and therefore roughly the average cost basis of holders. It includes coins that have not moved for years, some of which are probably lost.

Are active addresses a good measure of users?

Only roughly. One person can use many addresses and one exchange address can serve millions of customers. Spam, batching and changes in wallet software all move the number. It is more useful as a long-term trend on a single network than as a count of people.

Can on-chain data predict crypto prices?

No reliable method exists. On-chain data describes how coins have moved and at what prices they were last transferred, which can give context about holder behaviour. Most trading happens off-chain, many metrics are derived from price itself, and thresholds have been fitted to very few cycles.

Where does on-chain data come from?

From the blockchains themselves, read through nodes or block explorers. Specialist analytics firms process the raw data, label addresses and publish the metrics. Because their methods differ, the same metric can show different values from different providers.