Buying Bitcoin at extreme fear: Fear & Greed Index evidence

Extreme fear was not a better entry than an ordinary day

“Be greedy when others are fearful” is often translated into crypto as buying when the Crypto Fear & Greed Index shows extreme fear. We compared Bitcoin’s returns after every day in each index band from February 2018 to May 2026. Extreme fear (0–24) covered 21% of days.

After extreme-fear days the median 90-day return was −4.5%, with 44% of those windows positive. After any day it was +3.5%, with 54% positive. Over one year the median was +36.5% after extreme fear against +48.6% after any day. The average one-year return after extreme fear was higher (+100% against +86%), because a few readings near the December 2018 and March 2020 lows were followed by very large gains; the typical outcome was not better.

Thirteen separate episodes, not 637 signals

Fearful days arrive in long runs, so 637 days are far from 637 independent tests. Counting only the first extreme-fear day after at least 30 days without one gives 13 episodes. Seven of them were lower 90 days later, and the median 90-day return was −8.6%.

The episodes that worked were memorable: from 9 March 2020 BTC was up 23.4% after 90 days, and from 6 August 2024 up 20.9%. Others continued down: from 27 November 2021 −28.0%, from 15 July 2019 −24.1%, from 12 October 2025 −21.5% and from 29 August 2022 −18.9%. Fear readings often appear early in a decline rather than at its end.

Why a sentiment index is a weak timing tool

The index is built mainly from recent price behaviour, such as volatility, momentum, volume and social-media activity, so “extreme fear” largely restates that prices have already fallen sharply. In a bear market that condition can persist for months: 2022 had 193 extreme-fear days, and 90 days after them BTC averaged −9.4%, the same as after any day that year.

Extreme greed did not mark tops reliably either. Its median 90-day return of +10.1% was above the all-day median, because greed readings cluster in rising markets. In this sample the index described the mood of the market rather than predicting its next turn.

How the numbers were calculated

Index values are Alternative.me’s daily history, dated by UTC day. Prices are Coin Metrics’ PriceUSD, a UTC day-end reference rate. Each index day is matched to that day’s closing price, and forward returns run 30, 90 and 365 calendar days. Windows that would end after 23 May 2026, the last price in the snapshot, are excluded, so later bands have fewer observations. Medians and shares are computed over all days in the band.

The five bands are our grouping of the 0–100 scale and differ slightly from the index provider’s own labels in the middle of the range. No fees, spreads or taxes are included.

Is extreme fear a good time to buy Bitcoin?

Not reliably in this sample. After extreme-fear days the median 90-day return was −4.5%, against +3.5% after any day. A few fear readings near major lows were followed by large gains, but more of the 13 separate episodes were lower 90 days later than higher.

Does extreme greed mean Bitcoin is about to fall?

Not in this data. Median returns after extreme-greed days were above the average day at 30 and 90 days. Greed readings tend to occur during rising markets, so they described momentum rather than signalling a top.

Limitations

One sentiment index, one asset and one bull-and-bear history since 2018. Daily closes miss intraday entries; forward windows overlap; only 13 independent episodes exist. The index methodology may have changed over time. Historical returns are not a forecast or investment advice.

Sources and data access

• Alternative.me: Crypto Fear & Greed Index and methodology
• Coin Metrics: public BTC daily data
• Coin Metrics: PriceUSD definition and timestamp convention

Cited sources document external facts or data formats; they do not endorse our findings. The public snapshot contains the table and methods. Source-file fingerprints are included for recalculated datasets. Full archived raw files are not distributed here; fingerprints identify the files used, rather than independently proving the results.

Aggregate data snapshot (JSON) · Markdown · Methods and limitations

Live prior 30-day hourly baseline — separate from the historical study

Only the current watch list is shown, up to 12 assets. Missing or stale estimates remain unavailable. Venue and quote currency are kept separate; beta needs matching BTC data. Hourly statistics are not minute-move probabilities.

Observation cutoff (UTC): 24/10/2026, 14:46

Related research