NFT Market Sentiment
NFT Market Sentiment
NFT Market Sentiment is a 0-100 gauge of the NFT market's mood right now, from Extreme Fear to Extreme Greed.
Every hour it compares three things with their normal levels: NFT floor prices in US dollars, how many people are buying NFTs, and how much they spend. The three count equally. It describes the last day or two, not the last month, so it moves through the day as the market does.
Use it as a market-temperature signal, not as financial advice or a prediction of future returns.
What you'll see
- The gauge, with the current score and its label, for example 42 · Fear.
- A range line through the pointer. It runs from the lowest to the highest score of the selected period. The
|marks are the low and the high, and the small tag names the period. - A period selector: 24H (the default), 7D or 30D. It changes the range line.
- A hover card on the label, showing:
- the score, and the low and high of the period;
- the scores of the three signals: Floor prices, Buyers and Spending. Signal scores are whole numbers; a signal at its very end shows "<1" or ">99".
- When the score is Extreme, one line says why. For example: "Extreme Fear: floor prices are well below their 3-day average, spending is well below normal and buyers are below normal."
- When the score is Neutral but the signals point strongly in opposite directions, one line says so. For example: "Mixed: floor prices are well below their 3-day average, but buyers are well above normal."
- When Floor prices are at an extreme mostly because of ETH's own move, one line says so. For example: "Mostly ETH: ETH is 4.0% below its 3-day average, while NFT floor prices in ETH are little changed (−0.5%)."
- The ⓘ next to the title, which explains the signals and links to this page.
- An expand icon that opens the history chart. The line takes the colour of the band at each point. Hover a point to see its time, score, label and signals. Periods: 7D, 30D (the default), 90D, 1Y and All.
If the latest data is late, the gauge keeps the last score, and the hover card says when it is from, for example "as of 1 Nov, 14:00 · data delayed". If no score can be calculated, the gauge shows no number rather than a made-up one.
Refresh rate: hourly. Every hourly score is stored and shown on the history chart.
Score bands
- 0-24: Extreme Fear
- 25-44: Fear
- 45-55: Neutral
- 56-74: Greed
- 75-100: Extreme Greed
50 means no change from normal. The bands are fixed, so the labels keep the same meaning over time.
The three signals
| Signal | Weight | What it asks | Compared with |
|---|---|---|---|
| Floor prices | 1/3 | Where do NFT floor prices stand now, in US dollars? | Their average over the past 3 days |
| Buyers | 1/3 | How many buyers bought NFTs in the last 24 hours? | A normal day of the past 4 weeks |
| Spending | 1/3 | How much did they spend, in US dollars, in the last 24 hours? | A normal day of the past 4 weeks, read with the price direction |
Floor prices uses the AF Market Cap (Adjusted Floor) of the NFT market tracked by NFT Valuations, converted to US dollars at the hour's ETH/USD rate.
Buyers counts the unique buyers of the past 24 hours. It shows whether people are joining or leaving the market.
Spending is the US-dollar sale volume of the past 24 hours. Bundles and flagged wash trades are excluded. Heavy spending means different things in different markets: in a rising market it is greed, and in a falling one it is selling under pressure. So:
- above-normal spending counts towards greed only as far as prices are rising, and towards fear when they are falling;
- below-normal spending always counts towards fear.
How the signals combine:
- They count equally. On its own, any one signal moves the gauge at most about 17 points from 50, so an Extreme reading needs at least two signals pointing the same way.
- ETH's moves enter only through Floor prices.
- When floor prices collapse, the reading cannot rise above Neutral. If buyers surge at the same time, it can read Neutral, and the hover card says the signals are mixed.
How the score is calculated
- Floor prices: the median of the last 6 hourly values of ln(AF Market Cap in USD), minus their mean over the last 72 hours. The 6-hour median ignores a one- or two-hour data glitch.
- Buyers and Spending: ln(1 + the value for the past 24 hours), minus the median of the same at the same hour on each of the previous 28 days. Comparing the same hour of the day removes time-of-day effects.
- Each signal is scored from 0 to 100 against the size of a usual move:
S = 100 / (1 + exp(−ln 3 × x / s)).- s is the typical deviation: the median absolute value of x at the same hour over the previous 365 days, divided by 0.6745.
- No change scores 50, a typical rise 75 and a typical fall 25.
- The curve flattens gradually towards 0 and 100 rather than hitting them. A fall three times the typical size scores about 4, not 0, because moves that large are not as rare as a bell curve would assume.
- A score below 1 or above 99 needs a move more than four times the typical size.
- Spending is read with price:
Spending = 50 + min(S_V − 50, 0) + max(S_V − 50, 0) × (S_P − 50) / 50, where S_V is the spending score and S_P the floor-price score. - Hourly reading:
R = (Floor prices + Buyers + Spending) / 3. - The score shown is a weighted average of the last 48 hourly readings with a 6-hour half-life. The latest hour counts most, a reading 6 hours old counts half as much, and one 12 hours old a quarter. So the gauge moves through the day without flickering.
Because of that average, the signals in a hover card describe their own hour, while the score also carries the hours before it. When the market turns quickly, the score can lag the signals by a few hours.
Data quality and availability
-
Method changes are bridged. When NFT Valuations changes how floor valuations are calculated, the step in the AF Market Cap at that hour is removed, so it does not read as a market move. The change of 1 April 2026 is bridged this way. If such a step cannot be measured, the gauge keeps its last score instead of guessing.
-
Unannounced steps are held, not scored. Suppose the AF Market Cap moves by about a quarter or more in a single hour (0.25 or more in its logarithm) and the move is not a documented method change. Then the floor-price signal is treated as unavailable until that hour has left the 3-day window. Meanwhile the gauge keeps its last score and shows "data delayed". For scale: the largest other one-hour move since 2023 was 12%, and the method change of 1 April 2026 was a 51% drop.
-
Missing data never turns into a made-up score. A score needs:
- at least 5 of the last 6 hours of floor values, and 60 of the last 72;
- 21 of the 28 reference days;
- 90 days of history to size a usual move;
- 24 of the last 48 hourly readings.
When an input is missing, the signals are not reweighted and nothing is replaced with 50. The gauge keeps the last score, and the hover card shows when it is from.
-
History starts on 22 March 2023, the first hour with enough data for the calculation. Every point on the chart uses the current method.
How to read it
- Fear: NFT floor prices are below their level of the past few days, often with fewer buyers or less spending than usual.
- Greed: floor prices are above their recent level and buying is active.
- Neutral: little has changed, or the signals offset each other.
- Mixed: prices and activity point opposite ways, so the signals offset each other and the gauge can read Neutral. The hover card says so.
- It describes the last day or two. A two-day rebound inside a long decline can read Greed, and a sharp one-day drop inside a strong month can read Fear.
- ETH counts, but only through floor prices.
- Floor prices are in US dollars, so they move with ETH.
- Since April 2026, NFT floor prices in ETH have changed little from hour to hour, so this signal mostly follows ETH. The "Mostly ETH" line says when that happens.
- On its own, ETH moves the gauge at most about 17 points and cannot make it Extreme.
- Buyers and spending are NFT-only.
What it is not
- Not a forecast and not financial advice.
- Not a crypto fear and greed index. ETH counts only through NFT floor prices, one third of the score.
- Not NFT Valuations' estimate of what any NFT is worth. It uses floor prices, buyers and spending.
FAQ
Why only three signals?
Each answers a different question, and none is counted twice. Earlier versions also used listings, turnover, buyer/seller balance and volatility. Those either repeated the price signal or pointed the wrong way at the wrong time. For example, sellers disappear in a crash, so a buyer/seller balance reads as greed just when the market panics.
Why do the three signals count equally?
Each answers a different question: are prices up, are more people buying, are they spending more? None is clearly more reliable over hours to days.
Giving any one of them half the weight or more would let it make an Extreme reading on its own. Equal weights favour no question, and CNN's Fear & Greed index weighs its signals equally.
Until 6 November 2026, prices counted 60%. In practice, that let ETH's 3-day move decide most readings.
Why does the gauge read Neutral on a day when prices fell?
Because buyers or spending were above normal at the same time.
- Heavy spending in a falling market already counts towards fear.
- A rise in the number of buyers counts towards greed whatever prices do, because more people are joining the market. A surge of new buyers can also come from mints or airdrop campaigns.
The hover card shows "Mixed: …".
Why does it move during the day?
Sentiment changes hour by hour. The score compares the latest hours with the past 3 days for prices, and with the same hour on a normal day for activity. It is smoothed over about 6 hours, then recalculated and stored every hour.
Why US dollars and not ETH?
Most people count what they gain or lose in dollars. When ETH rises, NFT holders are richer even if floors in ETH stay flat, and a dollar-based reading reflects that. Because floor prices are one signal of three, ETH cannot decide the score on its own.
How is it different from crypto fear and greed indexes?
It uses only NFT market data from NFT Valuations: no crypto fear and greed data, social media or search trends.
Crypto indexes look at the past 30 to 90 days, so they move more slowly. This gauge looks at the last day or two. It often agrees with crypto indexes on big ETH moves, and disagrees when NFT buying and spending go their own way.
What does "Mostly ETH" mean?
Floor prices are measured in US dollars, so they move with two things: ETH's own price, and NFT floor prices in ETH. The gauge splits the Floor prices move into those two parts, each against its own 3-day average.
The hover card shows the line when both of these hold:
- Floor prices are at an extreme (15 or below, or 85 or above);
- ETH accounts for at least 70% of the move, in the same direction.
For example: "Mostly ETH: ETH is 4.0% below its 3-day average, while NFT floor prices in ETH are little changed (−0.5%)." The score is not changed by this; the line tells you what is behind it.
Why is the gauge still falling when ETH has bounced?
After a sharp ETH move reverses, the gauge can keep falling for a day or two, because it compares prices with their 3-day average.
What does "data delayed" mean?
An input for the latest hour is missing, late or held (see "Unannounced steps"). The gauge shows the last score it could calculate, the hover card shows the time it is from, and it updates as soon as the data arrives.
Can a single collection swing the score?
The score is calculated from global market-level snapshots across all tracked collections. Large collections weigh more in the AF Market Cap, but one collection's move is diluted by the rest of the market. A one- or two-hour glitch is ignored by the 6-hour median.
How often is this updated?
Hourly.
Did the method change?
Yes:
- 1 November 2026: the index moved from pillars ranked against the past year to three signals, so that no change reads 50 and the score reflects the market now.
- 6 November 2026:
- the scoring curve of step 3 changed from the bell curve (
100 × Φ(x / s)) to the curve above, so that large but not unusual moves no longer read as 0 or 100; - the "Mostly ETH" line was added;
- the three signals became equally weighted (they were 60%, 20% and 20%);
- Valuation was renamed Floor prices;
- unannounced one-hour steps in the AF Market Cap are held instead of scored;
- the hover card gained a line explaining Extreme and mixed readings.
- the scoring curve of step 3 changed from the bell curve (
The history chart was recalculated with the current method.
Will the method change again?
Not because of a single reading. The method changes only when one of these holds:
- A defect: the calculation does not match this page. We fix the code, not the method.
- An input changes meaning, is corrected, or stops being available. For example, floors or buyers are counted differently in a way that cannot be bridged as a one-hour step.
- A sustained failure over a full 365 days, measured with the published test definitions. One of:
- ETH/USD momentum alone explains at least as much of the score as the NFT-native inputs;
- over the whole history, Extreme Fear plus Extreme Greed fall outside 5–30% of hours;
- the average 1-day move exceeds 10 points;
- more than 4 days a year swing by 40 points or more.
- A new kind of data artifact moves the score by more than 10 points, and the bridge and hold rules cannot handle it.
Any change is tested against criteria written down before the test, on the same harness and on data it was not tuned on. It is announced here with its date. There is at most one method change a year, except for (1) and (2).
Not reasons for a change:
- a reading that disagrees with crypto fear and greed indexes, with commentators, or with the price chart on a mixed day;
- a quiet year with few Extreme readings.
Updated 10 days ago

