Crypto's Favourite Adage Half-Fails as Bitcoin Hits Extreme Fear

Tested against eight years of index history and Bitcoin (BTC) prices, though, only half that wisdom survives. 'Sell the Greed' holds up, and the more entrenched the euphoria, the more reliably it does. 'Buy the Fear' is the shakier half: at the horizons where a frightened investor most wants it to work, buying the bottom has more often underperformed a purchase made at random.

Dark clouds are gathering over Bitcoin. With the 30-day moving average of the index at 16, the 60-day at 25 and the 120-day at 22, all three are below the official Extreme Fear threshold. BTC/USD trades at around $61,500, more than 50% below its all-time high. Every sentiment indicator points to deeply entrenched pessimism.

This is exactly the kind of configuration that the adage 'Buy the Fear, Sell the Greed' invites investors to exploit. If investor sentiment is genuinely mean-reverting – that is, if extremes tend to snap back towards the average – extreme fear should mark an attractive entry point and euphoria should mark an exit. The FNG has become the standard quantitative expression of this contrarian intuition, with a score ranging from 0 (Extreme Fear) to 100 (Extreme Greed) and official thresholds at 25 and 75.

Data, methodology

The FNG aggregates six components (Volatility 25%, Momentum/Volume 25%, Social Media 15%, Dominance 10%, Google Trends 10%, Surveys 15%, currently paused).

Over the study period, the FNG averaged 45.7 with a standard deviation of 22.1. The index spent 23.5% of trading days in Extreme Fear (≤ 25) and 10.8% in Extreme Greed (≥ 75), reflecting the asymmetry of Bitcoin's volatile history, marked by prolonged bear markets and short-lived euphoric episodes.

(Source: Trading View & Alternative.me)

Six signals are built from the moving averages of the FNG at three smoothing horizons (30, 60 and 120 days), using Alternative.me's official thresholds of ≤ 25 for Extreme Fear and ≥ 75 for Extreme Greed. Using moving averages rather than spot readings isolates episodes of durably established sentiment. A 120-day moving average reflects fear entrenched over six months, a far more discriminating condition than a single-day reading. Forward returns are computed at five holding horizons, namely 30, 90, 180, 365 and 730 days.

To avoid treating consecutive days within the same episode as independent observations, a minimum gap of seven days is imposed between successive triggers. Each conditional distribution is compared with an unconditional reference (BTC forward returns sampled every 30 days) via a Welch t-test – a test for whether two averages genuinely differ when their samples vary in size and spread – with a Benjamini-Hochberg correction, a standard adjustment that stops a battery of simultaneous tests from throwing up false positives, to control the false discovery rate at 5% across the 30 tests run at once. Results that survive the correction are marked with a star. Absence of a star does not mean a result is meaningless, only that the sample cannot distinguish it from noise.

(Source: Trading View & Alternative.me)

Results

The chart crosses six signals in rows (the top three for Fear, the bottom three for Greed) with five holding horizons in columns (30d to 730d). Each cell shows the average return an investor would have obtained by buying BTC when the row signal was triggered, then holding for the column duration. Stars mark statistically significant cells (Benjamini-Hochberg correction at 5%). Unstarred cells fall short of that bar.

The unconditional reference is the implicit comparison point for every cell. It represents BTC's average return with no signal, corresponding to a purchase at a random date held over the chosen horizon. Over our period, it stands at +5% at 30 days (30d), +17% at 90d, +40% at 180d, +90% at 365d and +212% at 730d.

(Source: Trading View & Alternative.me)

Short term (30 and 90 days)

In Fear (top), the 30 and 90-day returns shown in the first two columns range from −10% to +6%, all below their respective references of +5% at 30d and +17% at 90d. An investor who bought BTC in Extreme Fear to sell one or three months later would have underperformed a random purchase. One cell delivers a significant result. The 120-day moving average ≤ 25 at 90d delivers −10%, 27 points below the reference, so six-month entrenched fear was followed by clear underperformance at three months over the study period.

In Greed (bottom), the cells display nominally higher returns, very dispersed across signals, with none statistically distinguishable from the reference. An investor who sought to sell on these signals at short horizons would, in general, have missed better performance than by selling on a random date over the period. The one exception is the 120-day moving average, which would have avoided a 7% loss.

At short horizons, there is no actionable rule. Sentiment does not deliver exploitable information, the only robust result runs against the 'Buy the Fear' thesis, and only the 120-day moving average at the 90-day horizon supports the 'Sell the Greed' thesis, without statistical significance.

Medium term (180 days and one year)

Reading the 365d column top-to-bottom on the Fear side, the shorter the moving average, the better the return. The 30-day moving average delivers +125% (light green cell), the 60-day delivers +115% and the 120-day only +33%. An investor buying on recent fear (30-day) would have outperformed the reference by 35 points on average over one year. Conversely, an investor buying at the bottom of an entrenched bear market (120-day) would have underperformed the reference by 57 points, a statistically confirmed gap. Entrenched fear signals a bearish regime that has not yet finished purging. The same hierarchy shows at 180d (+21%, +25%, +8%), with the 120-day return also significant.

On the Greed side over the same columns, the logic reverses exactly. The longer the moving average, the deeper the underperformance. At 365d, we read +53% for the 30-day, +18% for the 60-day and −13% for the 120-day, with the last two statistically significant. The same degradation appears at 180d (+35%, +8%, −4%), with the 60-day and 120-day confirmed significant. Every red cell in the bottom half validates the 'Sell the Greed' thesis. If the signal produces a return below what a random purchase would have delivered, that is precisely because it would have been better to be out of the market at that time.

Two symmetrical trends therefore emerge from these central columns. On the Fear side, the short window captures capitulations that rebound quickly. On the Greed side, the long window captures structurally exhausted euphoria that precedes the deepest corrections.

Long term (two years)

This is the column where the message of the chart becomes unambiguous. On the Fear side (top), the three cells are green (30-day at +229%, 60-day at +259%, the most intense cell on the chart, and 120-day at +196%), all close to or above the +212% reference. Over two years, entrenched fear ultimately corrects. Dispersion across episodes, however, prevents us from concluding that these signals formally beat the reference.

On the Greed side (bottom), the opposite holds. The entire column is red, and all three cells are statistically significant, with the 30-day at −11%, the 60-day at −31% and the 120-day at −48%. Whatever the smoothing, Extreme Greed is a confirmed exit signal at two years. The most striking result stands at the bottom-right of the chart. The 120-day moving average ≥ 75 delivers −48% against +212% for the reference. An investor who had sold BTC during an episode of six-month entrenched euphoria would have avoided an absolute net loss over two years, in an otherwise broadly bullish market.

Synthesis, the rule that emerges from the chart

(Source: Trading View & Alternative.me)

Over the study period and given the chosen indicators, one finding survives every horizon and every smoothing: 'Sell the Greed' works. From 180d through 730d, every 60-day and 120-day Greed cell sits below the reference in a statistically confirmed way, and the effect intensifies as the horizon lengthens – the longer euphoria has been entrenched, the more punishing it is to have stayed in.

'Buy the Fear', the more famous half of the adage, is the half the data quietly undermine. It works directionally at one and two-year horizons with the 30-day and 60-day moving averages, but turns counterproductive with the 120-day at intermediate horizons and loses its robustness at 730d. At short horizons, it does not work at all: the only robust short-term result, the 120-day Fear signal at 90d delivering −10%, runs directly against the contrarian intuition. Buying the bottom of an entrenched bear market – the very moment the adage feels most compelling – has historically been the worst version of the trade.

This asymmetry is not a statistical artefact. It directly reflects the structure of crypto cycles. Because fear phases are frequent and durable (23.5% of the time) while euphoria phases are rare and brief (10.8%), the discriminating moving average is not the same in the two zones. On the Fear side, a short moving average is needed to distinguish transient mean-reverting capitulations from the noise of an extended bear market that has not yet purged. On the Greed side, conversely, only a long moving average can isolate the rare episodes where euphoria takes hold long enough to signal a cycle top.

Current FNG level, limitations

As of 30 Jun, the three FNG moving averages (30-day at 16, 60-day at 25, 120-day at 22) are all in Extreme Fear territory. This is the most extreme configuration the 'Buy the Fear' thesis can identify. Historical data suggest this configuration produced returns below the reference over horizons from three months to one year. On a two-year horizon, returns following similar episodes converged towards the past reference (+196% versus +212%).

The robust message is therefore qualitative and directional. Over the study period, entrenched euphoria was a more reliable exit signal than entrenched fear was an entry signal, and the holding horizon was the decisive parameter. On the historical record, a reading as deeply fearful as today's has not been followed by the swift rebound the adage promises: episodes like it took time to resolve rather than marking a clean bottom.

This reading must nevertheless be framed by several limitations. This study is not a trading indicator. It documents regularities observed in historical data, without prejudging their future reproduction. Statistical power remains modest, since the most significant signal (120-day moving average ≥ 75) rests on only 17 episodes, leaving wide confidence intervals around the reported magnitudes. The available history (1 Feb 2018 to 30 Jun 2026) covers only about three market cycles, and the BTC ecosystem itself has been profoundly transformed over the period (arrival of spot ETFs, institutionalization and changes in liquidity structure). Nothing guarantees that the historical relationship between sentiment and return persists in a more mature market. The FNG is also a proprietary aggregate whose composition evolves. The Surveys component (15% in theory) is currently suspended, which limits the intertemporal comparability of readings. Finally, the study conditions only on the FNG and does not incorporate macro variables, onchain data or market-structure indicators that interact with sentiment.