The short version: on-chain data can show you roughly how long coins have sat still, and that's genuinely useful. It cannot tell you who owns them, why they moved, or what anyone intends — and the gap between those two things is where most on-chain analysis quietly turns into fiction.

When food goes out, the same four cats eat immediately. Two more arrive within a minute. The seventh waits until everyone else has finished and then eats whatever's left, every single time, and has done for two years.

From watching the bowl I can tell you a great deal about behaviour. What I can't tell you is anything about the cats. I don't know which are related, which belong to someone, which are the same cat I counted last winter with a different coat. I have excellent data about an outcome and almost none about the thing producing it.

That's on-chain analysis. It's better than nothing, which is why it caught on, and it's worse than its presentation, which is why it needs this article.

1. What the chain actually records

Every output has an age: the time since the coins last moved. That's not an estimate — it's a fact, publicly verifiable, with no methodology to argue about.

From that single honest measurement, analysts build:

Cohorts. Coins grouped by how long they've sat — under a month, one to three months, a year or more. The common convention treats roughly 155 days as the dividing line between "short-term" and "long-term", because historically coins older than that moved far less often.

Supply-in-profit. Comparing each coin's last-moved price to today's price gives a rough measure of how much of the supply is held above or below water.

Realised value. Valuing every coin at the price when it last moved — a kind of aggregate cost basis, and a much more interesting number than market cap, which I've written about before.

Dormancy and coin-days destroyed. When old coins move, they "destroy" accumulated coin-days. A spike means something long-held just woke up.

All of this rests on one real measurement and some reasonable conventions. The conventions are where it gets slippery.

2. The genuinely useful part

Old coins moving is a real signal. When a meaningful quantity of long-dormant supply becomes active, something happened — a custodian reorganising, an early holder selling, an estate being settled. You don't know which. But "something woke up" is information, and it's not available anywhere else.

Aggregate cost basis beats market cap. Realised value tells you roughly what the market actually paid, rather than multiplying the last trade by the total supply. For judging whether the average holder is up or down, it's far more honest.

It's public and verifiable. Unlike exchange volume, nobody can fabricate the chain. The raw inputs are checkable by anyone with a node — which, after my piece on reported volume, should count for a lot.

Long horizons work better than short ones. These metrics are reasonable at describing multi-month conditions and terrible at next week.

3. The four things it cannot see

Who owns anything. The chain has addresses, not people. One person may hold thousands; one address may hold the coins of a million customers. Every "number of holders" figure you've ever read is a guess dressed as a count.

Custodial internals. When an exchange or ETF custodian moves coins between its own wallets, the chain sees movement. Cohort metrics may see long-held supply becoming active. Nothing happened economically. As institutional custody has grown, this noise has grown with it.

Intent. A coin moving might be a sale, a transfer to cold storage, a custodian rotating keys, a loan, a change output, an inheritance, or someone practising. The chain records the movement and is silent on all of it.

Everything off-chain. Enormous amounts of trading never touch the chain at all — within exchanges, within ETF structures, in derivatives. A metric that only sees settled on-chain movement is watching one door of a building with several.

4. Reading it without being fooled

Prefer the raw to the composite. Coin age is a measurement. An index built from seven metrics with chosen weights is an opinion with a chart. The further from the raw data, the more of the author is in the number.

Ask who makes the heuristic. "Exchange balances" depend on a firm's private guess about which addresses belong to which exchange. Those labels are proprietary, imperfect, and revised quietly. Two providers routinely disagree about the same week.

Distrust anything with a precise threshold. The 155-day line is a convention, not a property of the market. Analysis that hinges on a coin being on one side of an arbitrary boundary is analysis that would say something different with a different arbitrary boundary.

Check whether the pattern survived the market changing. Metrics calibrated on the 2017 market are describing a world without spot ETFs, institutional custody, or regulated venues. Many of them have not been re-validated. They're still published.

5. The uncomfortable half

The incentives here are unusually bad. On-chain analytics is sold by subscription, and a subscription needs the data to appear decisive. "This metric is ambiguous and may not apply any more" is a true sentence that cannot be charted and does not renew. The pressure runs permanently toward confident interpretation of soft inputs — not through dishonesty, mostly, but through which charts get published and which don't.

Any metric with enough variants will appear to have predicted things. There are dozens of on-chain indicators, each with adjustable parameters. Afterwards, some will have lined up with what happened, and those are the ones you'll be shown. This is the same multiple-comparisons problem as seasonality, with better graphics.

And the honest version is unsatisfying. On-chain data supports statements like "a lot of supply has not moved in over a year" — true, verifiable, mildly interesting, and not actionable. The step from there to "therefore price" requires assumptions about people the data cannot see. Everyone makes that step. Almost nobody says they're making it.

I'd apply the same test I used on seasonality: what's the mechanism? "Long-term holders are accumulating" describes a measurement — coins sat still — and then names it as a behaviour by people whose identity and intentions are unknown. Sometimes that's right. It is never established by the chart.

I know what happens at the bowl, and nothing about the cats. Describing behaviour precisely is not the same as understanding who's behaving — and the chain only ever shows you the bowl. 🐾⚡