Every Bitcoin Bear Market, Measured by Depth and Time

Measured peak to trough on daily closes, every bitcoin decline of more than 50% since 2013 has taken between 88 and 406 days to reach its low, and between 189 and 1,177 days from the peak to the day the old peak was closed above again. There are five completed episodes in that period and one still open, so any single "bear markets last N months" figure is a summary of five data points, and the number moves with the rule used to draw the boundaries.

Nothing on this page is a forecast. The figures record what a public price series did under one stated rule, and the same series gives different answers under other reasonable rules. They say nothing about how long the current episode will last or what comes next.

The rule, and the data it runs on

The series is Coin Metrics' bitcoin PriceUSD, daily frequency, pulled from the Coin Metrics community API on October 23, 2026. Coin Metrics describes the metric as "the fixed closing price of the asset as of 00:00 UTC the following day," generated by its own fixing service. In plain terms, the value labelled with a date is the reference rate at midnight UTC at the end of that date. It is a reference rate, not the price on any one exchange, and it is a close, not an intraday high or low. The pull has 5,941 daily rows, from 2010-07-18 to 2026-10-22, with no missing days. Every figure on this page uses closes through October 22, 2026.

The drawdown rule is close-to-close, measured from a running all-time high:

Two consequences of that rule are worth stating before the results. First, an episode is defined by the old high, so a 52% crash that happens inside a longer decline from a higher peak is part of that longer episode and does not get its own row. Second, because closes are fixings at one moment per day, an intraday wick below the daily close is not in the data. The real lows were at least as deep as the ones listed. The script that produced every table here is a short Node file that reads the downloaded JSON and applies exactly the rule above, and the numbers on this page are its output, copied without rounding beyond what each column shows.

Every decline of 50% or more, 2010 to 2026

Source: CryptosEyes calculation on Coin Metrics community API PriceUSD, daily, 2010-07-18 to 2026-10-22, pulled October 23, 2026. Close-to-close, from a running all-time high. Prices are the reference rate, not any single exchange's trade.

Read the five completed episodes from 2013 onward as a group and the ranges are wide. Peak-to-trough time runs 88, 98, 364, 366 and 406 days, with a median of 364. Depth runs from 53.1% to 84.5%, with a median of 76.7%. Trough-to-regain time runs 91, 122, 481, 716 and 771 days, median 481. Peak-to-regain time runs 189, 210, 847, 1,080 and 1,177 days, median 847. Those medians are arithmetic on five numbers. They describe this set and carry no more weight than five numbers can.

A few rows deserve a closer look.

Rows 1 and 2 are a different market. The first row is a fall from about 40 cents to 20 cents in the autumn of 2010, and the series itself starts in July 2010 at 8.6 cents. At those prices a single trade could move a day's reference rate by double digits, and the 50.1% depth only just clears the threshold. Row 2, the 2011 decline from $29.03 to $2.11, is the deepest in the table at 92.7%, and its trough-to-regain time of 459 days came when the price was single dollars. We keep both rows because dropping data to flatter a pattern is its own kind of rule-tuning, but the medians above use only the 2013-onward rows, which is the fairer comparison.

Rows 6 and 7 sit back to back. The April 2021 peak at $63,446 fell 53.1% to $29,767 on 2021-07-20, and its old high was regained on 2021-10-19. The market made a new high on 2021-11-08 at $67,542, and the decline from that peak is row 7: 76.7%, down to $15,758 on 2022-11-09, 366 days after the peak. A reader who treats 2021 as one cycle would count one decline. This rule counts two, because the series closed above the April high in between.

Row 8 is open. The peak was 2025-10-06 at $124,824. The lowest close so far was $58,525 on 2026-06-30, which is 53.1% below the peak, 267 days after it. As of the last close in the data the old high has not been regained. A later section sets out where the series stands today.

Why the way back takes longer than the way down

A fall and the gain needed to undo it are different sizes, which is why regaining a high takes a different kind of time than reaching the low. A 50% fall needs a 100% gain from the trough; an 84.5% fall needs about 546%. The volatility guide on this site works the loss-and-recovery formula as a position-sizing input, in the drawdown and recovery tables of the Bitcoin volatility survival guide. The table here adds what the series did.

Source: CryptosEyes calculation, same series and rule as the main table. The open 2025 episode is excluded because it has no recovery date.

Across the closed episodes the recovery took longer than the fall in all but one: the April 2021 episode regained its peak in 91 days after falling for 98. For the 2011 episode, recovery took 2.8 times as many days as the fall; for 2013-12 and 2017-12, close to twice as many; for 2021-11 and 2013-04, 1.3 to 1.4 times. Those ratios are spread across a factor of three, so the table does not support a rule such as "recovery takes twice as long as the fall". Two episodes with similar depths, 2013-12 at 84.5% and 2017-12 at 83.8%, produced similar ratios of 1.9 and 2.0, and that may be coincidence between a pair.

The practical point is about what a long recovery means in dollars. The 2017 peak was $19,641 and the trough was $3,185. For the next 716 days, the series spent its time somewhere between the trough and the old peak, and any holder at the peak was below water for 1,080 days in total. The time below the old high is the cost that a depth percentage does not show.

How much of the time bitcoin spends below its old high

Because bitcoin has kept making new highs, it is easy to forget how rarely the series is actually sitting at one. Measured on the same running-high basis, for each day from 2012-01-01 through 2026-10-22:

Source: CryptosEyes calculation on the same series. The running high is computed over the full history, so days in 2012 and early 2013 are measured against the 2011 peak of $29.03.

On this series, 96.4% of days since 2012 closed below the highest close recorded up to then. Roughly seven days in ten were 30% or more below the running high, and close to half were 50% or more below it. Starting the window in 2014 instead, to avoid the 2011 high that the 2012 prices were measured against, gives 96.9%, 69.2% and 44.4%. The share of days sitting at an all-time high is therefore around 3.6% of the period. This does not say a given future day will be below a prior high; it describes the spread of past days, and it is the reason an investor's experience of "the market" depends on when they arrived.

Same series, different threshold: 30%, 40%, 50% and beyond

Coverage of bitcoin bear markets often uses the term without a stated depth, and the answer to "how many bear markets has bitcoin had" depends on where the line is drawn. The same rule and the same series, with only the minimum depth changed:

Source: CryptosEyes calculation. Median depth is the median of the listed episodes' depths.

At 50% there are 8 episodes. Lower the line to 40% and one is added: a decline of 41.1% in the first days of the series, July 18 to July 25, 2010, which starts on the very first observation and so is really "the first price on record to the first low". At 30% there are 13, with four more: February 2011 (37.0%), May 2011 (31.0%), June 2017 (35.8%) and September 2017 (34.0%). The last two sit inside the 2017 advance, two of the pullbacks that never make a bear-market list at 50% and do at 30%. Raise the line to 60% and the count falls to 5, as the 2010, April 2021 and open 2025 episodes drop out; at 80% only three remain: 2011, 2013-12 and 2017-12.

So "how many bear markets" has answers from 3 to 13 on a single series, and every one of them is a defensible reading. The honest thing to do with a count is to quote the rule beside it.

Same rule, different series

The second source is the Blockchain.com market-price chart, which its own API describes as the "average USD market price across major bitcoin exchanges." It is also free and daily, and it is built differently from a midnight fixing, so it is a fair test of whether the table depends on one vendor.

There is a housekeeping step. Blockchain.com's daily labels run one day ahead of Coin Metrics': the value labelled with a date matches Coin Metrics' value for the previous date. We measured it rather than assuming it. Comparing Blockchain.com's labelled-day value with Coin Metrics' value for the same day, the mean absolute log difference is 0.0364 in 2012 to 2014, 0.0268 in 2015 to 2018, 0.0254 in 2019 to 2022 and 0.0167 in 2023 to 2026 October. Comparing with the previous day, it falls to 0.0159, 0.0041, 0.0022 and 0.0010. We re-dated every Blockchain.com value one day earlier and ran the identical rule.

Source: CryptosEyes calculation on Blockchain.com market-price, pulled October 23, 2026, re-dated one day earlier, same rule and 50% threshold. Dates are UTC.

Four of the rows match the Coin Metrics table on every date: the 2013-12, 2017-12, April 2021 and 2025-10 episodes, with depths within 0.6 percentage points. Several others do not.

We cannot explain every difference. Blockchain.com does not publish, on the endpoint we used, the venue list or weights behind its average, and Coin Metrics' fixing uses its own methodology, so what we can show is that the two disagree, and by how much. The overall picture is stable: the large declines are in both series, within days of each other. The fine statistics, such as the number of days from peak to trough, are not.

Ending the decline on a 20% bounce instead of a new high

The running-high rule has one more built-in choice: an episode lasts until the old peak is regained. A different and common way to describe market cycles breaks the series at reversals instead. Here is that alternative, as an explicit rule on the same Coin Metrics closes. A decline ends when the close is 20% above the lowest close of the decline; a rally ends when the close is 20% below the highest close of the rally. Each decline is measured from the rally's high to the decline's low. Declines of 50% or more:

Source: CryptosEyes calculation on Coin Metrics PriceUSD. The 20% reversal size is an arbitrary choice.

This rule returns ten declines of 50% or more, against eight episodes under the running-high rule, and its list is different in kind. The long decline from December 2013 to January 2015 is not one row: it becomes a 52.0% decline from June to October 2014 and a 58.4% decline from November 2014 to January 2015, with a rally of at least 20% between them. The 2017-2018 decline becomes two legs, 60.0% in January to February 2018 and 62.1% from July to December 2018. The 2011 crash splits into three. And two episodes the main rule includes disappear altogether. The 2021-2022 decline lists only one leg of 50% or more, the 59.9% fall from 2022-03-29 to 2022-06-18, because the final slide to the November 2022 low started from a rally high on 2022-08-13 and covered 35.5%, short of the line. The 2025-2026 decline lists none, because its first leg bottomed at 49.1% on 2026-02-05, a 20% bounce followed (confirmed 2026-04-17), and the later slide to the 2026-06-30 low was shallower than 30% measured from the rally high.

The same rule also lists something the running-high rule cannot: the 2020 decline, 52.1% from 2020-02-14 to 2020-03-12, 27 days. Under the running-high rule that fall is inside the 2017-2020 episode, which was still open then, because price had not regained $19,641 until late November 2020.

This variant shows the cost of any single definition. If you ask "how long did the 2022 bear market take", the running-high rule answers 366 days from the November 2021 peak to the November 2022 low, 847 days to a new high; the 20% bounce rule answers 81 days, for one leg that was part of that interval. Both answers are correct for their rule, and they cannot both be called "the" length of the bear market.

Where the price stood after each peak, at fixed intervals

A different question avoids picking a trough at all: if the price is some number of days past a peak, where is it relative to that peak? The table gives the Coin Metrics close, as a percent change from the peak close, at 90, 180, 365, 540 and 730 days for each episode of the main table from 2011.

Source: CryptosEyes calculation. Blank cells (n/a) are dates later than October 22, 2026.

The spread is the finding. After 180 days, the series was between 14% and 90% below the peak. After 365 days, one episode was 92% above its peak (April 2013, which had regained the peak by November 2013 and carried on to higher highs) while others were 31% to 84% below. After 730 days, the 2011 episode was 280% above its peak, the April 2013 episode was 6% above, and four later episodes were between 47% and 68% below.

Three of the episodes sit near 47% below at day 180, in the 2013-04, 2021-11 and 2025-10 rows (-47%, -47%, -46%). With three observations, the similarity is a coincidence you could find in nearly any set of three numbers drawn from a range this wide; it is not a pattern, and nothing in the table supports treating day 180 as a marker for anything. The 2021-04 row shows how misleading a single checkpoint can be: it stood only 14% below its peak at day 180, then fell to 68% below by day 540, because a new high in November 2021 started a deeper decline that this peak-anchored view still counts.

The current episode, as of October 22, 2026

The open episode in the series is the one that began with the high on 2025-10-06. In the same terms:

Source: CryptosEyes calculation on Coin Metrics PriceUSD, latest close 2026-10-22.

The fall to the low took 267 days, which sits between the two short completed episodes since 2013 (88 and 98 days) and the three long ones (364 to 406 days). At 53.1% it matches the April 2021 episode's depth to one decimal place and is, with it, the shallowest in the post-2013 set. The latest close sits 34.9% below the high, and regaining the high from there would take a 53.6% gain. For reference, the earlier episode with a similar depth, April 2021, regained its high 91 days after its low; this one is 114 days past its low and has not.

That comparison has to stop at description. Two episodes of the same depth do not imply the same path, and five completed episodes do not define a distribution that a sixth must follow. Whether the 2026-06-30 close remains the low is unknowable from the series, and a later fall would make this episode deeper and longer in the table. For the cycle structure behind the supply schedule, the block-level record is in the Bitcoin halving cycles history; one on-chain valuation ratio computed across these same cycles is explained in the MVRV Z-score guide.

What a sample this small supports

There are six rows from 2013 onward if the open one is counted, and they come from one asset across years in which the venues, the instruments and the kinds of holders all changed. The table cannot separate those changes from the price record. Here is the narrow set of claims the table does support, each of which can be checked against the rows above:

Forced selling by traders who borrowed is one mechanism that can deepen a decline, and it is a separate topic from the price record itself; the liquidations explainer covers how forced closures work, and the table above has no way to attribute any episode's low to a particular event. We deliberately leave out dated news explanations for each peak and trough. A cause assigned to a date in hindsight is a narrative, and none of the rows above needs one.

Two limits on the data itself. The series is in US dollars, so it is silent about purchasing power. And the reference rate is a close; intraday lows were deeper, so a reader who needs the worst intraday print for a stress test should use a source that publishes highs and lows and should state its venue.

Other cycle measures built from on-chain data sit in the on-chain and cycles hub. If the aim is to turn these depths into a loss budget for a portfolio, the next useful step is the sizing arithmetic in the volatility survival guide, which takes a stress drawdown such as 53%, 77% or 84% and works backwards to a maximum position weight.

Frequently Asked Questions

How long does a crypto bear market last?

It depends on where you start and stop counting. On Coin Metrics daily closes, the five completed bitcoin declines of more than 50% since 2013 took 88 to 406 days from peak to trough, with a median of 364. Regaining the old peak took 189 to 1,177 days from the peak, median 847. A rule that ends a decline on a 20% bounce gives much shorter legs, between 2 and 144 days.

How many bitcoin bear markets have there been?

Under the running-high rule used here, 8 declines of 50% or more in the series from July 2010 to October 22, 2026, one of them still open and two of them in 2010 and 2011 when the price was in cents and single dollars. Since 2013 it is six, one open. At a 30% line the count is 13, and under a 20%-bounce rule at 50% it is 10. There is no single correct count.

Is bitcoin in a bear market now?

The series has an open episode that began with the close high on 2025-10-06. As of the 2026-10-22 close it was 34.9% below that high, after a low of 53.1% below on 2026-06-30. Whether that is called a bear market depends on the definition: it meets the 50% line used here because of the June low, and it fails the 50% line under the 20%-bounce rule. This page does not say which way it will go.

How long did bitcoin take to recover after the 2018 and 2022 lows?

Measured from each trough, the old peak was regained after 716 days for the December 2017 peak (trough 2018-12-15, regained 2020-11-30) and 481 days for the November 2021 peak (trough 2022-11-09, regained 2024-03-04). Measured from the peak instead, those are 1,080 and 847 days. The two numbers answer different questions, so check which one a source is quoting.

What was bitcoin's biggest drawdown?

On this series, 92.7% from $29.03 on 2011-06-08 to $2.11 on 2011-11-18, in a market where prices were single dollars. The deepest since 2013 is 84.5%, from $1,135 on 2013-12-04 to $176 on 2015-01-14. Both are close-to-close figures, so intraday lows were deeper.

Why do articles disagree about when bear markets started and ended?

Because the definitions differ and the data differ. A peak can be a daily close or an intraday high, a bottom can be the lowest print or the day a rally is confirmed, and a bear market can end at a bounce or at a new high. Even two free daily series disagree in places: for the 2013 decline the trough is 88 days or 7 days after the peak depending on the series, and for 2022 the low falls on one of two dates twelve days apart.

Sources

CryptosEyes publishes general educational research, not investment, legal, or tax advice. This page describes the past behavior of a price series under stated rules; it makes no forecast and is not a recommendation to buy, hold or sell any asset.