---
title: "The thinnest ceiling on record · Bitsaga Insights"
description: "One bitcoin in five just flipped from a loss to a profit. Measured on our own node, the supply sitting underwater above today's price is 35.6 percent, the thinnest ceiling at any bear market turn we can measure, against 43 to 46 percent in 2015, 2018 and 2022."
url: "https://bitsaga.be/insights/the-thinnest-ceiling-on-record"
language: "en"
---

> One bitcoin in five just flipped from a loss to a profit. Measured on our own node, the supply sitting underwater above today's price is 35.6 percent, the thinnest ceiling at any bear market turn we can measure, against 43 to 46 percent in 2015, 2018 and 2022.
[Insights](https://bitsaga.be/insights) · Technical

# The thinnest ceiling on record

Less bitcoin is stuck at a loss than at **any bear market turn we can measure**.

By [Rob Segers](https://bitsaga.be/about) · August 21, 2026

![A histogram of bitcoin supply grouped by the euro price its current holder paid. A tall orange block of about four million bitcoin sits just below today's price and is in profit. The blue-grey blocks above today's price are underwater and are visibly smaller. An inset shows the share of supply underwater above the price at four bear market turns: 45.0 percent in 2015, 45.7 in 2018, 43.1 in 2022 and 35.6 today.](https://bitsaga.be/images/insights/costbasis-wall-eur.svg)

> ### The short version
>
> **Only 35.6 percent of all bitcoin is worth less than its owner paid for it. At the last three bear market bottoms that number was 43 to 46 percent.**
>
> Bitcoin has risen about 30 percent off its June low, and 4.34 million coins, one in five in existence, flipped from a loss to a profit on the way up. What is left above the price is the smallest wall of trapped supply we have ever measured at a turn. On 19 August price also reclaimed two levels on the same day, the price recent buyers paid and its own 200-day average, 50 days after the low, which is the rhythm of the last three turns.
>
> The counterweight: no new money has arrived yet, and in 2015 this identical signal failed and price fell another 28 percent first.

## What a rally actually has to eat

Every unspent bitcoin carries a price: the one its current holder paid, set at the block that created the coin. That is its **cost basis**. Group all 18.5 million priced coins by that number and you get a map of what the market actually paid, which is the only honest way to see what a rally has to chew through.

Coins below the current price are in profit, and their holders are deciding whether to take it. Coins above are underwater, and their holders are waiting to get out at break-even. That waiting is what a ceiling is made of. It is not a line drawn on a chart. It is a queue of people with a specific price in mind.

We compute this from our own full node, in euro, at every block since genesis. As far as we can find, nobody publishes the euro version. Everyone works in dollars, which for a European holder is a chart about somebody else's currency.

## One coin in five just turned green

Bitcoin bottomed at about 58,500 dollars on 30 June and trades near 75,100 today, a rise of roughly 30 percent in seven weeks. Here is what that move did to the map.

4.34M

Bitcoin that flipped from a loss to a profit in this rally, 21.6% of all supply

That is not a sentiment reading. Those coins have a recorded cost basis between 51,269 and 75,148 dollars, and the price passed through every one of those levels in 52 days.

The single largest block sits directly beneath today's price: **3.96 million bitcoin** with a cost basis between 54,247 and 63,780 euro, which is about 63,400 to 74,600 dollars. That shelf did not exist a year ago. It held 1.46 million coins at the October 2025 high and has grown every week since. It is, quite literally, the people who bought this drawdown.

## And they are not selling

Here is the part we did not expect. When price broke above that shelf, the coins in it barely moved.

On the first day above the band, the 54,247 to 63,780 euro shelf gave back **30,600 bitcoin, 0.77 percent of itself**. The bin immediately above it gained 37,300. People who bought the bottom and are finally green are, so far, sitting still.

One caveat we have to state: the last day in our data covers only 40 blocks, about 28 percent of a day, because we measured at the tip rather than waiting for a clean close. That single reading needs another week before anyone should lean on it, and we will publish it again either way.

## The ceiling is the thinnest we have measured

What remains above the price is 6.59 million bitcoin, **35.6 percent of priced supply**. Run the identical measurement at the equivalent moment in the last three bear markets and the number is consistently larger.

| Bear market turn | Flipped to profit | Left underwater above price |
|---|---|---|
| 2015 | 21.11% | 44.99% |
| 2018 | 22.49% | 45.71% |
| 2022 | 10.92% | 43.14% |
| **2026** | **23.41%** | **35.60%** |

Measured at each cycle's short-term-holder cost basis reclaim, on our own node, block 963,401

This turn has both the largest flip into profit and the thinnest remaining overhead of any of them. The near wall is lighter than the floor beneath it: 1.27 million coins in the first 10 percent above the price, against 2.25 million in the first 10 percent below.

The real ceiling is further up. 2.93 million bitcoin sit with a cost basis between 88,000 and 104,000 euro, roughly 103,000 to 121,000 dollars. Those are the buyers of the late 2025 top, and they are the wall that a genuine new cycle high has to get through. But that block has already shed 2.84 million coins since October, which is to say most of the people who bought that top have already given up and sold.

## The 19th of August

The cost basis map explains the terrain. One date explains the timing.

On 19 August, bitcoin closed above the **short-term-holder cost basis**, the average price paid by coins moved in the last 155 days, at 67,607 dollars. It had been below that line for 95 days. On the same day it closed above its 200-day moving average for the first time in 289 days.

That twin reclaim landed 50 days after the cycle low. Here is when the same reclaim landed in the last three bear markets, and what happened next.

| Cycle | Reclaim | Days after low | Next 90 days | Next 365 days |
|---|---|---|---|---|
| 2015 | 10 Mar 2015 | 55 | −22% | +41% |
| 2018 | 20 Mar 2019 | 94 | +121% | +53% |
| 2022 | 12 Jan 2023 | 64 | +58% | +126% |
| **2026** | **19 Aug 2026** | **50** | ? | ? |

Two out of three worked, and worked enormously. The third is the reason for the next section.

## Why this might be wrong

We would rather give you the counter-argument than have you find it yourself.

**2015 is what a false reclaim looks like.** Price took back the same level on day 55, then spent only 3 percent of the following 90 days above it and fell 28 percent into a lower retest. The signal we are describing has failed before, in exactly this configuration, at exactly this point in the cycle.

**No new money has arrived.** This is the strongest bear argument and it is ours, not a critic's. The **realized capitalisation**, the sum of what every coin was last worth when it moved and therefore the closest thing to money actually committed, stands at 1.058 trillion dollars and is *down* 2.5 percent over 90 days. Valuation has improved because price rose, not because capital came in. Some comfort: realized cap was also falling at all three prior reclaims and turned positive 111, 14 and 18 days later. It has never been the signal that leads.

**The second gate is unbroken.** Above the short-term-holder line sits the **true market mean**, the average cost of the whole active market, which our node puts at **75,947 dollars**. Price is about 1 percent underneath it. Every prior bear exit needed both gates, and this one has passed one. It is worth knowing that Glassnode independently publishes this level at about 75,800 dollars from their own data. Two separate pipelines landing within a fifth of a percent is a good sign the number is real, and it is being tested as you read this.

**This bear never actually broke anyone.** It never printed an MVRV below 1, never went net-unrealized-profit negative, and never spent a single day with price below the average holder's cost. All three of those happened in 2015, 2018 and 2022. You can read that as a more mature market that no longer needs to wash out, or as a bear that has not finished its job. We genuinely do not know which, and anyone who tells you they do is selling something.

Two smaller ones, stated for completeness. Our entity-level figures showing large holders shrinking are 17 days out of date and rest on address clustering, which is guesswork we would not put in a headline. And the fee market has not recovered at all, so there is no demand-side confirmation from actual block space usage.

## What we think

The terrain is genuinely the best it has been at any bear market turn we can measure. Less supply is trapped above the price than in 2015, 2018 or 2022, the coins that bought the bottom are in profit and holding, and the calendar rhythm matches the last three turns almost exactly.

What is missing is money. Until the realized cap turns up, this is a repricing of coins that already existed and not new capital arriving, and that is precisely the shape a bear market rally has.

So: the ceiling is thinner than it has ever been at this point, and the floor under it is thicker. That is a good setup, not a guarantee. The line to watch is 75,947 dollars. We will publish this same chart again in two weeks whether it confirms or fails, because a chart you only republish when it agrees with you is not a measurement, it is an advertisement.

---

Every figure here comes from our own Bitcoin full node at block 963,401, 21 August 2026, and from a complete replay of every block since genesis. The cost basis map is a 128-bin logarithmic histogram, so the bin edges are exact and any price inside a bin is an approximation. The charts are drawn by scripts in our repository rather than by hand, so they can be rerun against fresher data, and they will be. If you find an error, [tell me](https://bitsaga.be/contact).