On the ninth of October — a date handed to us without a year, which is the first small tell that something is being smoothed over — a single on-chain metric moved through crypto media like a signal from the future. Short-term holders, the addresses that acquired Bitcoin within the past five months, had pushed roughly 11,000 coins through exchanges over the previous week. That was down 66% from a peak of 32,000 BTC. Within hours the interpretation had hardened into consensus: selling pressure is collapsing, supply is tightening, the squeeze is coming.
I have watched this exact sentence — a percentage decline bolted onto a single flow metric — land in my inbox hundreds of times since 2017. It is almost always right about the direction and almost always wrong about the meaning. A number falling is not the same as a force vanishing. A force weakening is not the same as a price rising. The distance between those three statements is where retail money goes to die, quietly, while a chart in the corner of the screen congratulates everyone for being early.
Silence speaks louder than hype. So let me be quiet for a moment and actually read the number.
Context: Who Counts as a Short-Term Holder, and Why the Label Is Doing More Work Than the Data
Before the phrase short-term holder means anything, you have to accept a definition that the industry invented and never fully standardized. In the common convention used by Glassnode and CryptoQuant, a short-term holder, or STH, is any address whose coins have not moved for fewer than 155 days. A long-term holder, or LTH, is anything older than that. The 155-day line is not physics. It is a heuristic — roughly five months, chosen because it captures coins that changed hands in the most recent market impulse without dragging in the whole prior cycle. It is a reasonable choice. It is also a choice, made by humans, on a spreadsheet, with a business model attached.
Here is why that matters more than the headline. Everything downstream of the label — the selling volume, the percentage decline, the narrative of a squeeze — inherits the weaknesses of the label itself. Two things have to be true for an STH selling figure to be trustworthy. First, the exchange address clusters have to be labeled correctly, so that a coin sent to a trading desk is actually recognized as a coin sent to a trading desk. Second, the time-bucket algorithm has to be consistent, so that the 155-day boundary is the same 155 days on every platform that quotes the number. Neither of those preconditions was disclosed in the flash we are examining. The source was described only as on-chain data analysis, without a named provider.
That is not a small omission. It is the whole game.
I learned to care about this the hard way. In 2017, working as a junior developer in Warsaw, I spent six months manually auditing smart contracts for three mid-tier ICOs, and one of the things that nearly cost me real money was a discrepancy between what a dashboard displayed and what the contract actually did. The dashboard said one thing about the time-crowdsale window. The code said another. The code won. Code does not lie, only humans do — and the humans are usually not lying on purpose. They are simplifying. They are rounding. They are choosing the definition that makes the chart look cleaner.
The same discipline applies here. When a metric arrives without a named source, the responsible first move is not to trade it. It is to ask which platform produced it, what its STH definition actually is, and whether the number has been translated from a primary dataset or copied from a secondary summary. A 66% decline computed on one provider's definition can be a 40% decline on another's. The direction survives the swap. The drama does not.
There is also a quieter problem hiding inside the framing itself. The number we were given is a flow: coins moved to exchanges per week. It is not a stock. It tells us about a rate of change, not a level. And because the STH pool is defined by a rolling time window, the pool itself is constantly changing shape underneath the flow. That single structural fact is enough to erode a large part of the headline, and almost nobody who repeated the 66% figure stopped to account for it.
Core: The Arithmetic Beneath the Decline
Let me start with the baseline that the flash quietly ignored, because without a denominator the numerator is just a mood.
After the fourth Bitcoin halving in April 2024, the block subsidy fell to 3.125 BTC. At roughly 144 blocks per day, that is about 450 BTC of new issuance daily — call it 3,150 BTC per week coming into existence from miners alone. Against that baseline, the 11,000 BTC of STH selling is not a trickle. It is approximately 3.5 times the entire weekly miner supply. The correct reading of 11,000 BTC per week is not that selling has stopped. It is that selling, even in its weakened state, still dwarfs the protocol's own inflation.
This is the single most useful reframe available, and it costs nothing but a division. A narrative that says selling pressure is collapsing invites you to imagine an empty order book. A narrative that says short-term holders are still unloading 3.5 weeks of miner issuance every seven days invites you to imagine the opposite: a persistent, heavy, market-driven supply that has merely stepped back from its peak. Both readings use the same number. Only one of them respects the denominator.
Now the arithmetic of graduation, which is the structural flaw that the flash never mentions. Because the STH bucket is defined as coins younger than 155 days, coins do not stay in it forever. Every day, a cohort of addresses crosses the 155-day line and mechanically reclassifies as long-term holders. This is not a behavioral event. It is a calendar event. An address that did nothing at all — no buy, no sell, no transfer — will leave the STH pool simply because time passed.
What does that do to a selling metric? It shrinks the denominator. If the pool of short-term holders is continuously draining its oldest members into the long-term category, then a falling STH selling volume can be partly explained by the fact that there are fewer STH coins available to sell in the first place. Some portion of the 66% decline is arithmetic migration wearing the costume of behavioral change. I would not claim it explains all of it. I would claim that any analyst who reports the decline without checking the STH total supply is reporting a number they cannot fully interpret.
And that gives us the first real fork in the road, the one the headline cannot resolve. If STH selling fell while the total STH supply also fell, the likely story is coins maturing into long-term hands — supply locking up, which leans constructive. If STH selling fell while the total STH supply held steady, the likely story is holders simply waiting, neither accumulating nor distributing, which is neutral at best. Two opposite meanings, one identical headline. The flash gave us no way to tell them apart.
Next, the distinction that trips up almost every retail reader: potential selling is not realized selling. The metric counts coins sent to exchanges. It does not count coins sold. A transfer to a trading desk can mean a market sell, but it can equally mean a limit order placed far above the market, a collateral deposit, a custody reshuffle, an over-the-counter settlement, or an internal wallet migration. The industry calls this exchange inflow, and the honest translation is potential sell pressure, not actual selling. The coins are loaded into the chamber. Whether the trigger gets pulled is a separate question that this dataset never answers.
So when the flash says short-term holders are selling at a profit, it is telling us something real but narrower than it sounds. It is telling us that coins are moving to exchanges, and that those coins are, on average, in profit — which implies a short-term-holder MVRV above one, meaning the average recent buyer is above water. That is genuinely informative. It tells us the market is in a profit-dominant regime, and that the selling we are seeing is ordinary profit-taking rather than panic capitulation. Ordinary profit-taking is a much calmer thing than a rug pull, and it should not be dressed up as either a bull signal or a bear signal. It is a baseline behavior of a market that has gone up.
Now the tools that would actually let us judge this, and that were conspicuously absent. A serious on-chain read of short-term-holder behavior does not rest on a single flow. It triangulates at least four instruments. STH-MVRV tells you whether recent buyers are in profit. STH-SOPR tells you whether the coins being spent are being spent at a gain or a loss, and how the ratio is trending. Exchange netflow tells you whether the exchange balance is rising or falling overall, not just for one cohort. And long-term-holder behavior tells you whether the strong hands are absorbing the supply the weak hands are releasing. Any one of these alone is a flashlight. Together they are a room.
The flash handed us one flashlight and pointed it at a wall.
Let me be concrete about what a proper SOPR read would have added. If STH-SOPR is above one and declining toward one, it means recent sellers are still taking profit but with decreasing enthusiasm — often a sign that holders are choosing to wait. If STH-SOPR spikes above one and then collapses below it, it means profit-taking turned into loss-taking, which is a very different and much more fragile state. The flash's phrase, still in profit, flattens both scenarios into one word. That word, profit, is doing an enormous amount of unearned work.
The provenance problem deserves its own paragraph because it is the risk that cannot be engineered away with more data. The flash named no provider. In practice, the credibility of an STH figure splits into three tiers. A primary read from a top-tier analytics platform, with a documented methodology, is solid. A secondary summary that faithfully relays that platform is acceptable if the numbers reconcile. A third-hand aggregation, copied across three Telegram channels and reposted by a bot, is a game of telephone where the 155-day definition might have quietly become a 30-day definition somewhere in the chain. Without a named source, you cannot know which tier you are reading, and the reasonable prior is that a flash optimized for virality sits closer to the third tier than the first.
I have run this exact triage on a crisis scale. During the Terra collapse in 2022, I stood up a small team to fact-check rumors rippling through a community of ten thousand people. For three weeks we did almost nothing but trace numbers back to their origin, because a single misattributed figure was capable of triggering a wave of panic selling that no one could undo. The lesson that survived that period is simple and slightly uncomfortable: in a fast market, the reliability of a data point is a more valuable asset than the data point itself. Reliability is what you can act on. A number without provenance is a rumor with a decimal point.
There is a modern twist that makes provenance even harder now, and it is one I have spent the past year building against. In 2026 I helped initiate a joint research project with a Warsaw-based AI startup to build a framework for verifying AI-generated crypto market reports. We cross-referenced machine sentiment reads against on-chain whale movements and published one of the first open datasets on algorithmic manipulation risk. The finding that stuck with me was not that AI gets numbers wrong. It is that AI reproduces the framing of whatever it was trained on, at scale, with total confidence. A single under-sourced STH flash can now be paraphrased into fifty articles that all read as independent. The telephone game no longer needs human callers. This is exactly why I mandate a human-verification layer on every AI-assisted piece that crosses my desk. The machine is not lying to you. It is echoing, and an echo has no source.
Now the blind spot that the crypto-native framing cannot see, and that a reader with any institutional experience sees immediately: the exchange inflow metric may be measuring the wrong plumbing entirely. Spot Bitcoin ETFs, approved in the United States in January 2024, move enormous size through creation and redemption rails that do not always appear as a clean on-chain transfer to a labeled exchange address. Some of the selling pressure that a purist would attribute to short-term holders may in fact be ETF-related flow that the on-chain lens either undercounts or misattributes. This is a directional suspicion rather than a proof — I would not stake a position on it — but it is a suspicion that any honest analyst should voice, because the metric was designed in an era before the largest marginal buyer and seller of Bitcoin was a wrapper product with a ticker.
This connects to something I have argued for years about tokenized and wrapped assets, and it applies with full force here. Wrapped Bitcoin, tokenized treasuries, and the whole category of real-world assets on public chains have been sold as a bridge between traditional finance and crypto. But the honest observation is that traditional institutions rarely need the public chain to do what they want. They need settlement, custody, and compliance, and they can get those from rails they already trust. When a metric like exchange inflow is read as if all meaningful Bitcoin flow happens on the base chain, it assumes a world where institutions live on-chain. In 2026, they mostly still do not. They live in wrappers, and the wrappers do not always show up in your labeled cluster.
The same myopia appears downstream. When people talk about where this data flows, they gesture at Layer 2 and at Lightning and at Stacks as if these were busy, decentralized networks processing the aftermath of base-chain selling. My own read, formed over years of watching sequencer architecture, is that the decentralization of these layers is largely aspirational — most sequencing is a small number of operators, and the promise of decentralized sequencing has been a slide deck for two years running. That is not a reason to dismiss the layers. It is a reason to be precise about what a base-chain flow metric can and cannot tell you about them. It can tell you almost nothing, because the activity it measures is upstream of a set of systems whose real behavior is far more centralized than the branding suggests.
So let me assemble the cross-verification framework that would turn this single flash into an actual signal, because that is the only useful thing to do with a weak data point — build the structure that would strengthen it.
A credible short-term-holder supply-tightening thesis needs four things to line up at once. One: STH total supply falling, confirming that coins are maturing into long-term hands rather than merely waiting. Two: exchange netflow persistently negative across the whole market, confirming that the coins being sent in are being matched by coins being withdrawn. Three: STH-SOPR above one but trending down toward it, confirming that profit-taking is decelerating rather than accelerating. Four: perpetual funding rates that are not already stretched to the upside, confirming that the narrative has not already been fully priced into leverage. If all four align, you have something worth acting on. If one flash alone fires, you have a headline.
Contrarian: The Decline Could Mean the Opposite of What You Were Told
The reflexive read is that falling selling pressure is bullish. I want to sit with the opposite possibility, because it is at least as plausible and far less discussed.
Consider what it means for short-term-holder selling to fall after a peak. One benign explanation is that holders are waiting for higher prices. But there is another explanation that the bull framing cannot accommodate: that the distribution already happened. If the heavy selling occurred earlier, at higher prices, then the subsequent decline in selling volume is not the market coiling for a move up. It is the market running out of sellers who wanted out, having already gotten out. A vacuum of sellers can precede a top as easily as it can precede a rally. Falling selling pressure is directionless on its own. It is the cycle location that gives it a sign, and the flash never told us where in the cycle we were.
There is a third explanation that is even quieter. Falling STH selling could simply mean coins moved off exchanges and into cold storage or long-term idle addresses. That is a transfer of custody, not a statement of conviction. It reduces the flow the metric measures without changing anyone's actual willingness to sell. The metric gets calmer. The holders do not. Nothing about the market's underlying tension has relaxed; the tension has just moved out of the camera's frame.
Truth is often buried under the noise, and the noise here is a very specific rhetorical device. The phrase down 66% from a peak is a peak-framing construction. It anchors your mind to an extreme and then presents the current state as a dramatic retreat from it. That construction is emotionally effective and analytically lazy. The relevant comparison is not against the peak. It is against the baseline of normal behavior and against the miner issuance floor. Measure the 11,000 against 3,150 and the word collapse evaporates. Measure it against a typical mid-cycle week and it may not even be remarkable. A number compared to its worst week will always look like a triumph. A number compared to its normal week tells you what is actually happening.
I will go one step further, because this is where my institutional skepticism earns its keep. A data point this thin is not circulating because it is important. It is circulating because it is useful. It slots neatly into a bullish report, it fits a chart, and it requires no follow-up questions. The most likely home for a flash like this is as one plank in a case someone else is already building. That does not make it false. It makes it borrowed. And borrowed evidence is exactly the kind that gets detached from its context and amplified until it means more than it ever could.
Takeaway: Build the Room, Not the Flashlight
The next time a single on-chain metric arrives wearing a dramatic percentage, the productive response is not to believe it or dismiss it. It is to ask what four instruments it left out, what denominator it avoided, and whose thesis it happens to fit. If STH total supply is falling alongside the selling, watch for supply locking up. If it is flat, watch for stalling. If price fails to respond to a supposed selling collapse, that silence is itself the answer — it means the distribution was already done.
The data will keep arriving faster and cleaner-looking, and most of it will be thinner than it appears. The edge is not in reading more of it. The edge is in reading fewer pieces of it properly, with a denominator and a second source. Code does not lie. Only humans do — and the humans who move your money are counting on you to skip the division.