Over 2014–2024, Bitcoin was the best-performing major asset in eight of eleven years, and the worst-performing in the other three (BlackRock). That single line is the whole thesis. The engine that produced roughly 136% compound annual growth since inception is the same engine that produced four drawdowns of 77% or more. You do not get the ascent without the falls; they are the up-swing and the down-swing of one oscillation.
This page examines the dynamics of both the bull and bear sides of Bitcoin’s value growth over the years. It puts them side by side and weighs the evidence. We can’t identify the bottom, but we can describe the dynamics at play.
—
—
Four completed cycles, four brutal drawdowns
Bitcoin has had five bear markets: four are complete, and the current one is the fifth, still unresolved. Every completed cycle has ended in a fall of at least three-quarters from its peak, and every one of those falls was, in hindsight, followed by a new all-time high. Both halves are in the table.
| Cycle | Peak | Trough | Drawdown | Peak→trough | Then recovered |
|---|
* The 2011 −93% is shown for the record but excluded from every trend line on this page: it partly reflects non-market artifacts (the June 2011 Mt. Gox breach printed a $0.01 tick), on an exchange that was then >70% of all volume, with no real market infrastructure. Recovery magnitudes shrink each cycle, 621× → 130× → 22× → 8×, even as the trough-to-next-peak time held remarkably steady near 1,050 days. The shrinking recovery is as real as the shrinking drawdown; see the counterweight.
Defining the terms, in Bitcoin’s context
The words carry over from equities, but the thresholds do not. Bitcoin routinely falls 30% inside a healthy advance, so the conventional rules mislabel it. Here is how this page draws the line.
The conventional “down 20% from the high” rule is useless for Bitcoin: 20–30% pullbacks happen routinely inside bull runs. This page uses a stricter, two-part definition. A structural bear takes 100+ days to bottom, which is measurable and assumption-free, and has historically run 77% deep or more. That depth describes past bears; it is not a floor the next one must reach. And unlike a conventional equity bear, a Bitcoin bear tends to grind for months, clearing out leverage before it ends.
The sustained advance between two structural bears. Historically the trough-to-next-peak span has held remarkably steady near 1,050 days (about 35 months). A Bitcoin bull routinely contains equity-sized “bear” corrections, drops of 20–30% that resolve upward rather than ending the advance, which is exactly why the conventional 20% rule misreads them.
Bull and bear name the regime you are in, not what happens next.
The drawdowns have been getting shallower, and here is exactly how much to trust that
Two different measures point the same way. Discount 2011, and the completed drawdowns (peak-to-trough depth) run −85%, −84%, −77%. Separately, volatility (the day-to-day swing) has roughly halved over five years. They are different measures, not the same number series. There is a real, mechanical story for why a larger, more liquid asset should swing less. But “shallower each time” is the kind of pattern that can seduce us into timing the bottom, so the evidence for each mechanism is graded below.
- Capital scaling / law of large numbers. A ~$2 trillion asset needs vastly more capital to move a given percentage than a $10 million one did.
- Raised cost-basis floors. Aggregate holder cost basis sat near 43.7% of the prior peak this cycle, versus a third or less historically (Galaxy), a higher floor for mean-reversion to fall toward.
- Volatility decay and supply absorption: large long-term-holder distribution met by comparable ETF and treasury inflows through 2026.
- “Institutions dampen the swings.” The clean version of this theory broke down in the February 2026 sell-off: ETF redemptions transmitted selling straight to spot, and institutions treated Bitcoin as a high-beta tech proxy and sold it in macro stress. What was a supposed stabiliser can become an accelerant. This corroborates the rising-correlation caveat below.
- Regulatory clarity reducing panic. Bitcoin already sits in the cleanest regulatory category: in March 2026 the SEC and CFTC jointly classified it a digital commodity (administrative guidance). The CLARITY Act (the Digital Asset Market Clarity Act) would codify that; it passed the U.S. House in July 2025 and cleared the Senate Banking Committee in May 2026, but as of mid-2026 it is not yet law, still needing a Senate floor vote, reconciliation, and the President’s signature. Clearer rules could plausibly reduce forced-seller panic, but that is a reasonable expectation, not an established effect.
Three completed cycles cannot prove a trend. A skeptic would call “shallower each time” an extreme case of small-sample bias, indistinguishable from three correlated random walks. The right verb is “has behaved,” not “behaves”: this is historical evidence, not a projection.
Half of what it was, still multiples of everything else
Volatility here means how much price swings around its path over a window, measured as annualised realised volatility of returns. It is not the same as a drawdown (a peak-to-trough fall), and not the same as risk (the next section). Measured that way and computed live from this site’s price series, it traces an arc: from ~150% in the early years toward ~40% now, roughly half of five years ago, and today below Tesla and around Nvidia. It has fallen, but it is still high.
Rolling annualised realised volatility of log returns, computed from the shared ~12-day series (annualised at ~30.4 periods/yr). Because the series is sampled, treat this as a trend line, not a point estimate; a daily series would read somewhat higher.
Two caveats on the decline
It was not a smooth 15-year glide. An econometric read finds the 2014–2020 downtrend was statistically weak; the real compression is post-2020: ETFs, a trillion-dollar cap. It would be more accurate to say Bitcoin’s volatility “fell sharply, chiefly since 2020,” than “steadily for fifteen years.”
Daily-close volatility fell; intraday did not. Range-based intraday volatility stays high, near the 79th percentile of the S&P 1500, driven by leveraged perpetual-futures liquidation cascades. That churn is largely a derivatives artifact, not a change in Bitcoin’s underlying value, which is why the next section matters.
Dispersion is not the same thing as losing your money
Volatility measures dispersion, how far price ranges around its path. It is symmetric: a +40% month is penalised by the maths exactly as much as a −40% month. Risk, for a long-term holder, is something else entirely: the permanent loss of capital. They are different quantities, and Bitcoin’s makes the gap unusually wide.
Bitcoin’s returns are positively skewed and fat-tailed: a large share of its “volatility” is upside, especially rebounding off the Power-Law lows. A single volatility number therefore overstates the downside danger to someone with a long horizon. Over 2020–2024 Bitcoin’s Sortino ratio of 1.86 was roughly double its Sharpe ratio, and Sortino only penalises downside movement, so most of the swing it ignores was the good kind. It would be more accurate to say Bitcoin’s volatility is largely to the upside, but it is still high.
The guardrail. “Risk isn’t volatility” must never slide into “Bitcoin isn’t risky.” There is no valuation floor; wrong sizing, a broken thesis, or being forced to sell at a low are all real, permanent-loss risks. The upside skew is a historical characteristic, not a guarantee that the next dip resolves upward.
The drawdown does its damage through you, not to you
A 50% fall only becomes a 50% loss if you sell into it. Whether you do is mostly psychology, and the psychology is well-measured. Kahneman and Tversky (1979) established that losses are felt roughly twice as intensely as equivalent gains, a result replicated across some 90% of studied populations. That asymmetry is the machinery a bear market runs on.
It strikes twice. It keeps newcomers from ever entering (an opportunity cost), and, more destructively, it shakes existing holders out at the lows, converting a paper swing into a realised loss. And it does its worst work not in the crash but in the grind: the first sharp drop clears out leverage, a relief rally sucks people back, and then months of slow bleed, every bounce sold, break the spirit. People capitulate in the grind, not the crash.
Conviction has to be earned, not sold. As Saylor has said, “Everyone gets Bitcoin at the price they deserve.” A real-time case study arrived in 2026: a “Saylor is selling” headline (his firm sold 32 BTC to meet a dividend) triggered retail to panic-sell at the lows, right before the same firm rebought 1,550 BTC lower.
The empirical answer to the fear: the worst annualised return over any five-year hold has been about +3.6% (NYDIG), no five-year holder has lost money, and every four-year hold since 2013 has averaged 60–120% annualised (CoinMarketCap). Cross-check the recovery maths on The Bitcoin Horizon.
The guardrail. “Don’t panic-sell” is not “diamond-hands, never sell, ever.” Even Saylor’s firm sold when it had to. Put precisely: earned conviction, a long horizon, and correct sizing so you are never forced to sell let you avoid panic selling, contingent always on the thesis holding. This argues against emotional selling, not against rational portfolio decisions.
Everything above, argued against
Everything above has a serious opposing case. Here is the strongest version of the other side.
Maturation compresses the upside too
The very forces that shrink the drawdowns (scale, liquidity, institutional depth) mechanically cap future returns. Even Bitcoin-favourable analysts model it: VanEck’s base case is ~15% CAGR for 2026–2050, revised down from 25% a year earlier; Fidelity warns high returns “may not persist as the asset matures.” You cannot sell the shallower drawdowns without also selling the mid-teens returns. They are the same coin.
The “uncorrelated diversifier” claim no longer holds
Post-ETF, Bitcoin’s correlation with equities rose to 0.5–0.88 (near 0.9 with the Nasdaq at points), versus roughly zero in 2018–2020. And it is not a safe haven: across S&P drawdowns worse than 12%, Bitcoin lost ~35% on average while gold gained ~4.7% (State Street). It has behaved risk-off, then uncorrelated, then correlated, as adoption shifted, so pick no single label.
The reflexive floor can be clawed back
The raised cost-basis floor is real but not fixed. In a genuine panic, Galaxy notes, a 10–30% cost-basis decline can pull the implied floor from ~$40K back toward ~$28K. The floor moves under stress, which is the moment you were counting on it.
The model is a tool, not an oracle
The Power Law fits history well but is not a deep structural law: its exponent varies threefold by start date, and the multi-sigmoid variant that fit history best was among the worst out-of-sample (Baquero & Menezes, 2026). Usefully, the simple Power Law still forecasts well at 12–24 months precisely because it doesn’t overfit. Use it descriptively, never as a price oracle.
Twenty-plus analysts have published a “next bottom” spanning roughly $25K to $68K: a deep camp near $25–37K, a consensus near $40–55K, a shallow camp near $54–68K. The width of that spread is the finding. When the professionals disagree by nearly 3×, the conclusion is that the bottom is unknowable, not that a twenty-first number would help. Today’s price sits near the shallow end, so further downside is entirely live if the consensus or deep camps are right, and nobody knows which.
Read cycles as a risk-management framework, not a timing tool. They tell you drawdowns of this magnitude happen and must be survivable in your plan. They do not tell you when.
Position sizing is the survival mechanism
Everything above resolves into one practical lever. The reason to size a Bitcoin position carefully is not timidity; it is that risk contribution is wildly non-linear. On institutional risk-budgeting maths, a 2% allocation contributes roughly 5% of a portfolio’s risk, but a 4% allocation contributes about 14% (BlackRock); JPMorgan puts a 3.5% Bitcoin sleeve at roughly the risk of a 40% bond allocation. Read that as a warning about the curve, not a recommendation to hold 4%.
Which is why the recommended caps cluster low: BlackRock 1–2%, VanEck 1–3% as the “sweet spot,” and Galaxy’s finding that the biggest marginal Sharpe/Sortino gain comes just from 0% → 1%. The figure and its cap belong together.
This fuses the two middle sections: oversizing guarantees the loss-aversion shakeout. Hold more than you can watch fall 77%, and the psychology does the rest; you sell at the low, converting volatility into permanent loss. Correct sizing is precisely what lets conviction survive the drawdown, which is the only way you are still holding to capture the returns. Sizing is not a tax on the thesis; it is what makes the thesis survivable.
The how much belongs to its own page: the Kelly criterion, why its raw answer of ~65% is absurd, and why fractional Kelly lands nearer ~10%. How Much Bitcoin? owns that calculation; this page owns only the reason it matters.
Sources & methodologyhow the numbers are made, and where the claims come from
How we measure a drawdown
- Headline drawdown %s are documented daily-close cycle extremes: the researched, cleansed figures (peak-to-trough on daily closes), not numbers computed from the sampled series. Daily-close is the rigorous standard, deliberately shallower than intraday. For the widest cycle, 2013–15, the definitional range runs −81.6% to −87.7% depending on method; the shallower daily-close read is the one used here.
- The charts use the shared ~12-day Power-Law price series (
PL_DATA), the same series every other channel page on this site reads, so the live “where are we now” number ties exactly to what those pages show. A peak or trough between samples reads slightly shallow on the line; each cycle line is annotated with its true documented depth to reconcile the two. - Volatility is rolling annualised realised volatility of log returns, computed from that sampled series (~30.4 periods/yr). Disclosed as a sampled-data trend, not a point estimate; a daily series reads somewhat higher.
- The live status computes drawdown-from-peak against the 2025 peak of $126,198 (Oct 6 2025) and days-since-peak, both recomputed every page load from the live spot price. The 2025 trough is unresolved; its final depth is left blank on purpose.
- 2011 is asterisked and excluded from every trend line (Mt. Gox artifacts, >70%-of-volume single point of failure).
Computed here vs. cited
Computed live on this page: the drawdown-from-peak and day-count, the cycle-table drawdown geometry, the day-zero overlay, and the rolling-volatility series. Cited (not computed): the worst-five-year-hold figure (NYDIG), the cross-asset volatility multiples, CAGR and Sortino figures, allocation caps, recovery magnitudes, and every named institutional claim below.
Sources
Commercial-interest sources (asset managers with a Bitcoin product) are tagged as such and paired with skeptic-leaning research.
- Kahneman & Tversky, “Prospect Theory,” Econometrica (1979): loss aversion, losses ~2× gains.
- Baquero & Menezes, “Bitcoin’s Power Law: Weak Structure, Strong Forecasts,” arXiv:2605.21316 (2026): the model-is-not-a-law counterweight.
- BlackRock / iShares (commercial interest): best asset 8 of 11 years 2014–24; ~54% vol vs 15.1% gold / 10.5% equities (Jan 31 2025), easing to 43% by Dec 31 2025; 1–2% allocation and the non-linear risk-contribution figures. Its “uncorrelated diversifier” framing is not inherited here.
- VanEck (commercial interest): 15% base-case CAGR 2026–2050, revised down from 25%; 1–3% allocation; on-record on pro-cyclical reflexivity.
- Galaxy Digital (skeptic-leaning research): 4 of 13 bottoming signals fired; cost basis 43.7% of ATH; the reflexive-floor claw-back toward ~$28K; the analyst bottom-estimate clusters.
- Also drawn on: NYDIG (drawdown methodology; worst 5-yr hold +3.6%), Fidelity Digital Assets (maturation, return-persistence caution), State Street (safe-haven comparison), CoinMarketCap (4-year hold returns), Paybis (capitulation grind), and standard composite-index methodology (NYDIG / Coin Metrics / Glassnode) for the cleansed daily-close extremes.