Key Takeaways
- What it is: Coin days destroyed on a day, divided by the units transferred that day. The units cancel to leave days, so the output is an average age.
- What the division buys: Comparability. A day moving five times as much supply produces roughly five times the raw coin days destroyed without any change in the age of what moved.
- Why it is spiky: Age has no upper bound while quantity is capped by supply, so the product is driven by the oldest tail. A tiny spend of ancient coins can outweigh a large spend of recent ones.
- It needs smoothing: A single day's value carries very little signal, and the smoothing window materially changes what the chart appears to say.
- The ceiling rises with the chain: The largest dormancy a day can produce is the age of the oldest coin, which grows by one day every day.
- The denominator is contested: Raw and adjusted transfer volume give different dormancy series, and providers do not always say which they used.
Who This Guide Is For
Read this if you already know what coin days destroyed is and want to know why anyone bothers dividing it by volume. The raw total confounds two different things, how much moved and how old it was, and dormancy separates them at a distributional cost that makes the resulting series awkward to use. The numerator is defined on the coin days destroyed guide.
Educational content about measurement, not individualized financial advice. All figures below are invented.
What Does Coin Dormancy Measure?
Coin dormancy measures the average age of the coins that moved on a given day, weighted by how many units each spend carried. Its unit is days, because coin days divided by coins leaves days behind.
That unit is the reason the metric exists. A raw coin-days total grows with either factor, so a day producing a large total might have moved an enormous quantity of very recent coins or a modest quantity of very old ones. Those two situations describe opposite behaviours and are indistinguishable in the raw number. Dividing by volume collapses the quantity dimension and leaves only the age dimension.
What "average age" means precisely
It is a quantity-weighted mean, so a spend of 10,000 units aged 10 days counts a hundred times more toward the day's average than a spend of 100 units aged 10 days. What it is not is an average over transactions, which would weight a dust spend equally with a large one. Two providers agreeing on the numerator can still disagree on the level if one averages per transaction.
Why the ceiling moves
The maximum value a day can produce is the age of the oldest coin on the chain, reached in the theoretical case where that coin is the only thing that moves. Since the oldest coin ages by one day every day, the metric's ceiling rises for the life of the network. An all-time-high reading is therefore partly a statement about the chain's age, and so is any percentile threshold computed over a long history.
How Is Coin Dormancy Constructed?
The numerator is settled. Almost all of the construction risk sits in the denominator.
Which volume goes on the bottom
Raw on-chain transfer volume counts everything the ledger recorded: change returning to the sender, coins shuffled between an exchange's own wallets, and consolidations that merge many small outputs into one. Adjusted volume attempts to strip that out. Because the numerator counts coin days destroyed by every spend regardless, pairing an all-spends numerator with an adjusted-volume denominator produces a systematically inflated dormancy, and pairing an adjusted numerator with a raw denominator deflates it. The transfer volume guide covers how far apart the two series can sit.
Change outputs cancel, and that is unusual
On a UTXO chain, spending a small amount out of a large output forces the remainder back to the sender as change. Suppose a 100-unit output aged 500 days is spent to send 1 unit onward. Coin days destroyed is 100 × 500 = 50,000, and raw volume records 100 units moved. Dormancy is 50,000 divided by 100, which is 500 days, exactly the age of the coins involved. The change output inflated both halves of the fraction by the same factor and cancelled out, which makes dormancy far more robust to change-output inflation than the raw coin-days total is.
The cancellation is not universal. It holds when the spent inputs share one age. A consolidation merging outputs of many different ages destroys a weighted mix of coin days while contributing the full sum to the denominator, and a chain of self-transfers resets ages repeatedly, so later hops add fully to the denominator and almost nothing to the numerator.
Day boundaries and currency
The metric is computed on a calendar day, normally in UTC. For a series whose interesting values come from rare large events, a spend landing either side of midnight is the difference between two moderate days and one record day. Separately, some providers publish a value-denominated variant dividing a currency-valued numerator by currency-valued volume. That variant has a price series embedded in it and should never share an axis with the native-unit version.
Formula and Measurement Logic
Dormancy = coin days destroyed today / units transferred today
The formula is a mean, and the properties that matter follow from the distribution it averages over rather than from the arithmetic.
Age is unbounded above, quantity is not. The largest quantity any spend can carry is bounded by supply; the largest age it can carry grows every day. The product is driven by the upper tail, and the within-day population is typically one large mass of young coins plus a thin tail of ancient ones.
The mean is the wrong summary for that shape. A day whose median spend age is under a week can report a dormancy of several weeks because a few very old units moved, and nothing in the published number reveals which case you are looking at.
Smoothing changes the message. A simple moving average holds a spike at full weight for the length of the window and then drops it in one step, manufacturing a sharp apparent decline exactly one window after the event.
| Choice | Options | Effect on the series |
|---|---|---|
| Denominator volume | Raw transfers or entity-adjusted transfers | Adjusted denominators raise the level, sometimes substantially |
| Weighting | Per unit, or accidentally per transaction | Per-transaction weighting lets dust spends move the average |
| Units | Native coins, or currency-valued | The currency version embeds a price series and is not comparable |
| Smoothing | None, simple moving average, median, log transform | Simple averages create a phantom drop one window after a spike |
| Day boundary | UTC calendar day or rolling window | Splits or merges the rare events that define the series |
| Age source | Native output age, or reconstructed from a balance | Account chains require a convention that changes the numerator |
How Should Coin Dormancy Be Interpreted?
Read a smoothed dormancy series as a statement about the age mix of what is being spent, and a single day as almost nothing. Rising smoothed dormancy means the spending mix shifted toward older supply; falling means recent supply is churning.
Set the smoothing rule before looking
Because the raw series is dominated by rare large observations, the smoothing choice determines what you see, and choosing it after seeing the chart is a way of choosing the conclusion. A rolling median or a log-transformed mean is more honest about this distribution than a simple moving average, at the cost of being harder to explain.
Run the leave-one-out test
Recompute the day's dormancy with its single largest coin-day contributor removed. If the value collapses, the day was one transaction and should be described as one transaction. The days people write about are precisely the days with one dominant spend.
What it does not say
- Nothing about price. Old coins moving to a new custodian, into a multisig upgrade, or onto an exchange all read identically.
- Nothing about who. A single entity reorganizing its holdings and a broad set of long-term holders acting independently produce the same average.
- Nothing about the median holder. The value can sit far above the age of the typical coin moved.
Step-by-Step Workflow
- Establish which volume series sits in the denominator, and confirm the numerator was filtered the same way.
- Confirm the weighting is per unit rather than per transaction.
- Choose and record a smoothing window and summary statistic before looking at the chart.
- For any day of interest, recompute dormancy with the largest single coin-day contributor removed.
- Check whether an apparent decline sits exactly one smoothing window after a spike.
- Compare against the raw coin days destroyed series to see whether a move came from age or volume.
- Note the chain's age when using long-history percentiles.
- Describe the result as the age mix of spending, not as a decision by long-term holders.
Worked Hypothetical Scenario
All figures here are invented to demonstrate the arithmetic and describe no real chain.
Take a hypothetical day on which exactly three spends occur.
| Spend | Units | Days since last move | Coin days destroyed | Share of volume | Share of coin days |
|---|---|---|---|---|---|
| 1 | 12,000 | 6 | 72,000 | 77.2% | 22.1% |
| 2 | 3,500 | 45 | 157,500 | 22.5% | 48.4% |
| 3 | 40 | 2,400 | 96,000 | 0.3% | 29.5% |
| Day A total | 15,540 | 325,500 | 100% | 100% |
Dormancy for Day A is 325,500 divided by 15,540, or 20.95 days. Spend 3 carried 40 units, one quarter of one percent of the day's volume, and contributed 29.5 percent of its coin days. That is the heavy tail in a single line: a rounding error in quantity terms deciding almost a third of the numerator.
How much one small spend moves the average
Remove spend 3 and recompute. Volume becomes 15,500, coin days become 229,500, and dormancy becomes 14.81 days. The 40-unit transaction moved the day's average age from 14.81 to 20.95 days, a rise of about 41 percent. Any commentary about that day's dormancy is commentary about one transaction that a block explorer would show as trivially small.
Why the normalization is worth having
Now compare Day A against a second hypothetical day with far more activity but much younger coins.
| Day A | Day B | |
|---|---|---|
| Units transferred | 15,540 | 96,000 |
| Coin days destroyed | 325,500 | 480,000 |
| Dormancy (days) | 20.95 | 5.00 |
| Reading from the raw total | The smaller day | The larger day |
| Reading from dormancy | Older supply moved | Recent supply churned |
Day B destroyed roughly 47 percent more coin days than Day A, so a chart of the raw total would show Day B as the more significant event and an old-supply-is-moving narrative would attach to it. Dormancy says the opposite: the average coin that moved on Day B was five days old, against nearly twenty-one days on Day A. Day B was a high-turnover day in recent supply. The raw total conflated size with age, and the division separated them.
What Can Make the Interpretation Wrong?
- Reading a single day. One day's value is a mean over a handful of dominant observations. It is a description of those transactions, not of the market.
- Missing the one-transaction case. As the worked example shows, a spend worth a quarter of a percent of volume can move the day's average by 41 percent.
- The moving-average artifact. A simple average drops a spike abruptly at the end of its window, producing an apparent decline that reflects the window length rather than any change in spending.
- Mismatched numerator and denominator filters. An all-spends numerator over an adjusted-volume denominator inflates the level systematically.
- Long-history percentiles. The metric's ceiling grows by one day per day, so a percentile computed over many years compares eras with different achievable maxima.
- Consolidation and self-transfer chains. Merging many outputs of mixed ages, or repeatedly moving coins between wallets, distorts the ratio in ways the change-output cancellation does not cover.
- Treating it as intent. An old coin moving says an old coin moved. Custody migration, wallet software upgrades, inheritance, and selling all look identical here.
Cross-Network and Provider Comparison
Two providers publishing dormancy for the same chain can disagree for a reason that has nothing to do with the numerator, so check the denominator first. Raw and adjusted transfer volume can differ by a large factor on chains with heavy exchange activity, and the ratio inherits that difference directly.
Across networks, the first obstacle is where age comes from. A UTXO chain records the creating block of every output, so age is read rather than inferred. An account chain stores a balance with no lot structure, so the provider has to impose first-in-first-out, last-in-first-out, or average-age accounting before coin days destroyed exists at all, and the three conventions produce different numerators from identical data.
The second obstacle is contract activity. Every deposit into a liquidity pool, lending market, or bridge escrow moves coins and resets their age, and every withdrawal moves young coins back out. A network with an active decentralized finance ecosystem reports structurally lower dormancy for that reason alone.
Fee level compounds both. Where transfers are almost free, holders shuffle coins for operational reasons and keep the average age of moved supply low. Comparing dormancy between a high-fee and a low-fee chain compares fee markets alongside behaviour.
Advanced Analytical Methods
Publish the spend-age quantiles, not just the mean
The same data that produces dormancy can produce the quantity-weighted median, upper quartile, and ninetieth percentile of the day's spend ages. Those four numbers together make it obvious when a mean of twenty-one days sits on a population whose median is under a week.
Leave-one-out sensitivity as a published series
Computing each day's dormancy twice, once normally and once with the largest coin-day contributor removed, and plotting the gap, produces a concentration series. High values flag days that were one transaction.
Log transformation and robust smoothing
Taking logs before averaging, or using a rolling median, keeps a spike from dominating a window and then vanishing from it in one step. Both should be disclosed, since they change the visual shape considerably.
Comparing the two normalizations
Dormancy divides coin days destroyed by same-day volume, which makes it fast, spiky, and about the mix of what moved. Liveliness divides the cumulative version of the same numerator by cumulative coin days created, which makes it slow, bounded, and about the chain's whole history. Reading them together is informative precisely because they are not confirmatory.
Practical Checklist
- I know whether the denominator is raw or adjusted transfer volume.
- I confirmed the numerator uses the same filter as the denominator.
- I confirmed the weighting is per unit, not per transaction.
- I fixed the smoothing window and statistic before reading the chart.
- I recomputed the day with its largest coin-day contributor removed.
- I checked whether an apparent decline sits one window after a spike.
- I noted that the metric's ceiling grows with the chain's age.
- I described the result as the age mix of spending, not as holder intent.
Frequently Asked Questions
Why divide coin days destroyed by volume at all?
Because the raw total confounds how much moved with how old it was. A day that moves five times as much supply produces roughly five times the coin days destroyed even if the age mix is identical. Dividing by volume cancels the quantity dimension and leaves an average age in days, which is the quantity the metric was designed to expose.
How can a tiny transaction dominate a day's dormancy?
Because age has no upper bound while quantity is capped by supply. In the hypothetical day worked above, a 40-unit spend of coins aged 2,400 days contributed 29.5 percent of the day's coin days while representing a quarter of one percent of its volume. Removing it dropped the day's dormancy from 20.95 days to 14.81 days.
Is a record-high dormancy reading meaningful?
Treat it cautiously. The highest value any day can produce is the age of the oldest coin on the chain, and that ceiling rises by one day every day, so records are partly a function of how long the network has existed. Percentile thresholds computed over many years compare periods with different achievable maxima.
Does rising dormancy mean long-term holders are selling?
It means older coins made up a larger share of what moved. The metric contains no price input and no counterparty information, so custody migration, a wallet software upgrade, an exchange reorganizing reserves, and an actual sale all read identically. Pairing dormancy with a price-aware series is the only way to begin addressing intent, and even then the answer stays probabilistic.
What are the units of dormancy, and what does a given value represent?
Dormancy is coin days destroyed divided by the volume of coins spent, so its unit is days. A value of one hundred means the coins that moved that day had been sitting unspent for an average of one hundred days, weighted by size. This makes it a measure of the average age of what moved, independent of how much moved, which is exactly the property the raw coin days destroyed figure lacks and the reason the ratio is calculated at all.
How does dormancy differ from the average age of the whole supply?
Dormancy describes only the coins that moved on a given day. Average coin age describes everything in existence, whether it moved or not. They can diverge completely: on a day when a small amount of very old supply transacts while the rest of the network sits still, dormancy spikes while average supply age is barely affected. Dormancy is a flow measure over the spending population, and average coin age is a stock measure over the whole supply.
What is entity-adjusted dormancy?
It is dormancy computed after removing transfers judged to be between addresses belonging to the same entity, so that internal wallet management does not register as spending. Because those internal moves often involve old supply, they inflate the raw metric substantially. The adjustment depends entirely on the quality of the entity clustering behind it, which is a heuristic process, so an entity-adjusted series inherits the labelling provider's decisions and cannot be independently reproduced from chain data alone.
Why does dormancy tend to drift upward over a chain's life?
Because the pool of supply available to age keeps getting older. In a network's early years no coin can be more than a few years old, which caps how high the average age of spent coins can go. As the chain matures, an increasing share of supply has been sitting for a long time, so any spending drawn from that pool destroys more coin days per coin. Comparing a dormancy reading against levels from many years earlier therefore compares against a period where the high values were not attainable.
Should dormancy always be read with spent volume alongside it?
Yes, because dormancy is a ratio and a ratio conceals the size of its denominator. A very high dormancy reading produced by a handful of ancient coins moving describes a different event from the same reading produced by a large volume of moderately old supply, and the ratio alone cannot distinguish them. Plotting spent volume on the same chart makes the distinction immediate, and it is the single most useful accompaniment to any of the coin-age ratios.
References
These sources should be reviewed during editorial verification. They support data structures and methods, not the hypothetical conclusion. Provider formulas, chain rules, and APIs can change. Confirm current documentation before publication.
- Bitcoin Developer Guide: Transactions: change outputs and output age, which drive both halves of the ratio.
- Coin Metrics: Transaction Metrics: raw and adjusted transfer-value definitions, the contested denominator.
- Glassnode: Metric Catalog: where dormancy and its variants are catalogued.
- Ethereum.org: Technical Introduction: the account model that forces an age-accounting convention on the numerator.
- Coin Metrics: Network Data Glossary: ledger and transfer definitions used across both inputs.