How to Read the Crypto Fear & Greed Index

Most of the numbers in this series come straight from a market: a funding rate, a price, an order book. This one doesn't. The Fear & Greed Index is a recipe. Someone takes a handful of market numbers, weights them and squeezes them into one score between 0 and 100.

It's probably the most-shared number in crypto, and it comes with a rule attached: "be greedy when others are fearful." Buy when it says Extreme Fear, sell when it says Extreme Greed. So we took every daily reading since 1 February 2018, 3,154 of them, and lined each one up with bitcoin's price to see what the rule actually got you.

What goes into the number

Alternative.me publishes the recipe. It's six inputs, all about bitcoin:

Look at the top two rows. Volatility and momentum make up 50% of the score, and both are just the recent price wearing different clothes. A sharp fall raises volatility and kills momentum, so the score drops. A steady rally does the opposite. The social-media and search inputs tend to follow the price too: people post and search more when bitcoin is moving.

The score is then sorted into five labels. The source's own cut-offs are 0 to 25 for Extreme Fear, 26 to 46 for Fear, 47 to 54 for Neutral, 55 to 75 for Greed and 76 to 100 for Extreme Greed. It has never read below 5 or above 95.

Finding 1: it's a thermometer for the last month

Put the index under the price and the two lines move together. When bitcoin fell hard, the score sank. When it rallied, the score climbed. The correlation between the index and bitcoin's move over the previous 30 days is 0.70. That's high for anything in markets.

Another way to see it is to compare each day's price with bitcoin's average over the year before. On Extreme Fear days, bitcoin sat a median 24% below that average. On Extreme Greed days it sat 85% above it. So "extreme fear" mostly means "the price has fallen a lot lately", and "extreme greed" means it has risen a lot.

That's useful to know. It isn't a secret signal, though. It's a summary of something you can already see on the chart. And a summary of the past month doesn't say much about the next one: the correlation with bitcoin's next 30 days is 0.13, and with the next 90 it's 0.21. Both are weak, and both are positive. If anything, a higher score came before slightly better returns, which is the opposite of what the contrarian rule needs.

Finding 2: fear days were cheaper, every year

Here's the part of the rule that works. Because Extreme Fear shows up after a fall, the price on those days is usually lower than it was the rest of the year. We checked every calendar year from 2018 to 2026. In all 9, the average price on Extreme Fear days was below the year's average.

The gap varies a lot. In 2020, all 49 Extreme Fear days came around the Covid crash. They averaged $6,561, against $9,956 for the year. In 2022, when bitcoin fell almost all year, 207 days read Extreme Fear and they were barely cheaper than the rest. In 2023 there were only three.

Add it up and it matters. Putting $10 into bitcoin every single day from February 2018 got an average price of $16,208. Putting $10 in only on the 739 Extreme Fear days got $13,424, about 17% less. Each dollar spent that way was worth 6.3 times as much by 24 September 2026, against 5.2 times for the every-day buyer. Buying only on Extreme Greed days would have turned each dollar into 3.1.

One catch: the fear-only buyer put in far less money, and some years they barely bought at all. In 2023, when bitcoin rose 156%, they'd have bought on three days.

Finding 3: extreme fear isn't the bottom

The rule's promise is timing: Extreme Fear marks the bottom. It usually doesn't. We grouped the Extreme Fear days into 27 separate sell-offs. After the first Extreme Fear day of each one, bitcoin fell a median 24% further before it hit bottom. In 15 of the 27, it fell another 20% or more. Only three times was that first day within 5% of the low.

The fear also lasts. The longest unbroken run was 74 days, from 6 May 2022 to 18 July 2022. The most recent sell-off first read Extreme Fear on 12 October 2025, with bitcoin at $114,959. It kept coming back to Extreme Fear on and off until 17 April 2026, and printed a 5 along the way, the joint-lowest reading ever. Bitcoin didn't bottom until $60,885 on 6 June 2026, 47% below that first reading.

There are times it nailed it. In March 2020, the first Extreme Fear day came three days before the Covid low. But the 2022 bear market started reading Extreme Fear on 12 April 2022 at $40,075, and bitcoin went on to $18,791.

Count it by day and the picture is the same. Across every day in the sample, bitcoin was higher 90 days later 54% of the time. After an Extreme Fear day, it was 44%, and the median move was −5.6%. That's the lowest of the five bands.

It's not one bad stretch, either. Split the sample in two at May 2022. Before that, Extreme Fear days were followed by a higher price 90 days on 49% of the time. After it, 39%. Both are below the all-days rate.

Over a full year it looks kinder. A year after an Extreme Fear day bitcoin was higher 61% of the time. But after an ordinary Fear day it was 70%, and across all days 66%. Bitcoin mostly went up over these eight years, and Extreme Fear didn't add to that.

Finding 4: extreme greed missed the biggest tops

The other half of the rule says to sell Extreme Greed. We took every record close that was followed by a fall of 20% or more before the next record. There are seven since 2018. The index read Extreme Greed on four of those days.

The four that did read Extreme Greed were followed by the smaller falls. The three biggest crashes came after readings of 74, 75 and 71, in April 2021, November 2021 and October 2025. The November 2021 record came before a 77% fall. The October 2025 record, $124,659, came on a reading of 71, and the index never went above 74 in the month around it. Bitcoin then fell 53%.

Extreme Greed's record after the fact flips too. Before May 2022, bitcoin was higher 90 days after an Extreme Greed day 72% of the time, with a median of +48%. That was the 2020-21 bull run. Since then it's 37%, with a median of −3.9%. A rule that worked in one half and failed in the other isn't much of a rule.

How to read it yourself

• Read it as a summary of the last few weeks. A low score means bitcoin has fallen and swung around a lot lately. A high score means it has risen. Check the chart and you'll usually see why.
• Use Extreme Fear to spread your buying, not to time it. Fear days were cheaper than average in all 9 years, so adding a bit more on them lowered the average price. Expecting the first one to be the bottom is where it goes wrong: the median sell-off had 24% left to fall.
• Expect it to stay low for a while. Extreme Fear ran for 74 days straight in 2022, and the 2025-2026 sell-off kept going back to it for six months.
• Don't wait for Extreme Greed to sell. The October 2025 record came on a 71. If you have a reason to take profits, don't hold on waiting for the index to reach 76.
• Remember it's one company's recipe. Other sites publish their own fear-and-greed scores, the inputs have changed over the years, and a 25 today isn't built from exactly the same things as a 25 in 2018.

Method, and what this cannot tell you

We took every daily reading of Alternative.me's Crypto Fear & Greed Index from 1 February 2018, and bitcoin's daily close on Binance (BTCUSDT) for the same days. Each reading was sorted into the source's own five bands. For every day we measured bitcoin's return from that day's close to the close 30, 90 and 365 days later; the reading is published at midnight UTC, so that close comes after it. Runs of extreme fear no more than 14 days apart count as one episode, and an episode's low is the lowest close from its first day to 60 days after its last. A top is a record close followed by a fall of 20% or more before a new record.

• One index. Alternative.me's is the one most people mean, but other sites publish their own fear-and-greed numbers with different inputs, and they don't always agree.
• The recipe has changed. Its survey input has been paused, and the social-media and search inputs depend on what X and Google make available. A reading from 2018 and one from 2026 aren't built from exactly the same things.
• Days overlap. A 90-day return starting on one day shares 89 days with the next day's. 739 extreme-fear days aren't 739 separate tests. They're about two dozen episodes, which is why the article counts those too.
• Eight and a half years is a few cycles. The sample holds three big bear markets (2018, 2022, 2025-26) and two big bull runs. Any rule that looks good or bad here is mostly about how those went.
• Four days are missing from the source (14-16 April 2018 and 26 October 2024). No fees or taxes are counted anywhere.