Bitcoin Rainbow Chart

The Bitcoin rainbow chart, also called the BTC rainbow chart, is a price chart on a logarithmic scale with colored bands drawn around a long-term curve. The curve is a line fitted to Bitcoin’s past prices, and each band marks a range of distance from it. The band the price sits in shows how far Bitcoin is above or below its own long-term trend. It describes where the price has been. It does not say where the price goes next.

This page draws its own version, which we call CDW Rainbow v1, after this site, CDW (Crypto Daily Watch). We fitted the curve to 5,312 daily prices from 23 March 2012 to 7 October 2026, froze it, and cut the space around it into nine bands. The band names state only a position, such as “Well below the curve” or “Near the curve”. Other sites publish their own versions of the rainbow chart: Bitbo uses six bands, while Blockchain Center’s original chart has ten and its current chart nine, like this page. Their curves and names differ, so the band they show for the same day can differ from this one. The Method section lists every number.

The chart has nine bands. Each is 41% above the one below, so two bands up multiply the price by 2.0, a doubling. The colors run from cool to warm: the cool ones (violet to cyan) lie below the curve and the warm ones (lime to red) above it, with green around it. The two outer bands are open: “Well below the curve” covers every price below its upper edge and “Extremely above the curve” every price above its lower edge.

The middle band, “Near the curve”, runs from 16% below to 19% above the curve. The other names count band widths away from it, not percent: “Slightly below the curve” is the first band under the middle one and covers 40% to 16% below the curve, and “Slightly above the curve” covers 19% to 68% above the curve. The band table gives the percent range of every band as its distance from the curve. The names state a position only. The dollar ranges are recalculated every day, because the curve rises over time.

The two right-hand columns continue the fitted curve to 1 January 2027 and 1 January 2028. They show where the band edges would sit on those dates if the curve simply goes on. That is a curve extension, not a forecast: it says nothing about the price on those dates, and a later review of the model would move the edges.

The table rounds each edge to three significant digits, while the percentages on this page are computed from the exact edges. Working a percentage out from a rounded edge can therefore give a slightly different result.

The share of days restates the fit. The curve is a least-squares line through the 5,312 days of the fit window, 23 March 2012 to 7 October 2026, so those days cluster around it. The shares count exactly those days. They describe the past and give no odds for the future. The 21 days since 7 October 2026 are not part of the fit, so the shares leave them out; the year table further down includes them.

The vertical axis is logarithmic: the same height always means the same percentage change, so a doubling takes as much height at $1,000 as at $100,000. The closed bands therefore all have the same height at any date, because each one is the same percentage above the one below.

1. Find the price line. It is Bitcoin’s daily price, and its right end is the latest reading.
2. Read the band the line is in. The color and the name say where the price sits against the curve: violet to cyan lie below it, green is around it, lime to red lie above it. The name is always printed next to the color. The names count band widths, not percent: “Slightly below the curve” is the first band under the middle one and covers 40% to 16% below the curve. The band table gives the percent range of every band as its distance from the curve.
3. Check the distance to the edge. The band table gives each band’s dollar range. The closer the price is to an edge, the less its band says, because a small move changes the label.

Worked example. At the last update (28 October 2026, 02:26 UTC), Bitcoin was at $80,872. That is inside the “Below the curve” band, which runs from $57,802 to $81,648. The curve is at $137,071, so the price is 41.0% below the curve. A rise of 1.0% would take the price into the next band up, and a fall of 28.5% into the band below.

The short answer is that the chart fits the history it was fitted to, and the two tables below show that and little else. The first table gives the share of days Bitcoin spent in each band in every calendar year from 2012 to 2026; the current year counts so far. This table includes the 21 days since 7 October 2026, which are not part of the fit. The second lists, for each cycle, the day with the highest offset and the day with the lowest offset inside a fixed window around it, with the price, the offset from the curve and the band. It uses the 00:00 UTC prices, not intraday highs, so the day with the highest offset need not be the day of the highest price. Offsets are in decades: one decade is a factor of ten, so +0.3 is about double the curve and −0.3 about half of it.

We set the windows by hand, after the fact: 1 October 2013 to 31 March 2014, 1 October 2017 to 28 February 2018, 1 March to 31 December 2021 and 1 December 2024 to 31 December 2025 for the highest offsets; 1 December 2014 to 31 March 2015, 1 October 2018 to 28 February 2019 and 1 October 2022 to 31 January 2023 for the lowest. A different window could pick a different day.

Share of days in each band, per year

Percent of the days of that year. The darker the cell, the more days. The latest year counts only the days so far. Days after 7 October 2026 are not part of the fit; the latest row includes 21 of them.

Highest and lowest offsets in each cycle

Each row is the day with the highest or the lowest offset inside a fixed window around a cycle, found after the fact. The prices are 00:00 UTC prices, not intraday highs, so a row need not be the day of the highest price. Offset is the distance from the curve in decades: +0.3 is about twice the curve, −0.3 about half.

What the tables show. The days cluster around the curve, which is a least-squares fit to the days of the fit window, 23 March 2012 to 7 October 2026: the three bands around it, from “Slightly below the curve” to “Slightly above the curve”, hold about 54.5% of those days, and the two outer bands about 6.5%. The largest gaps are lopsided: the day farthest above the curve sits about 1.02 decades above it, the day farthest below about 0.45 below it. The placement of the bands, three below the band that holds the curve and five above it, was chosen after looking at the data. The 21 days since 7 October 2026 are not part of the fit. The figures above leave them out, and the year table counts them. The shares within the fit window restate the fit; they do not test it. The cycle table tells the same story: its highest and lowest offsets are the largest gaps the curve leaves in each cycle, found after the fact.

Where it misses. The largest distance above the curve has shrunk from cycle to cycle: it was about 10.5 times the curve in 2013, 6.3 times in 2017, 3.6 times in 2021 and 1.4 times in 2024, each lower than the one before. A set of bands with fixed widths cannot show that trend, and the chart cannot say whether it continues. The highest 00:00 UTC price of 2025, $124,728, came at a smaller multiple of the curve, 1.3 times, so the latest row, 2024, marks the highest offset and not the highest price. The lowest offsets of the cycles ranged from 0.46 to 0.57 times the curve. “Extremely above the curve” was last visited on 9 January 2018 and “Well below the curve” was last visited on 23 May 2016. A frozen fit also goes stale. In our test refits, a curve fitted only to data up to the end of 2016 would have stood about 51% below ours on 7 October 2026, and one fitted up to the end of 2018 about 41% above. Ours can age the same way, which is why it is reviewed every January.

How much evidence this is. The cycle table has 7 rows, 4 highest offsets and 3 lowest, from data in 15 calendar years (about 14.6 years). Neighboring days are almost the same observation: the residuals are correlated at about 0.998 from one day to the next, so the 5,312 days hold only a handful of independent cycles. An R² of 0.94 measures how closely a line follows a series that trends upward. It does not show that the line keeps working.

1. Take one price a day. We use Bitstamp’s BTC/USD price at 00:00 UTC of each date, from 23 March 2012 to 7 October 2026 (5,312 days). Later days come from CoinGecko, one 00:00 UTC price a day. They join the history, the chart and the year table, but never the fit.
2. Count the days. The day number is the number of whole days since 9 January 2009, so 10 January 2009 is day 1.
3. Fit the curve. We fitted a straight line by ordinary least squares to log10(price) against ln(day number), every day weighted equally. It gives log10(curve) = −15.9514 + 2.4019 × ln(day number), shown here rounded to four decimals. The model itself uses six.
4. Measure the offset. The offset is how far the price is from the curve in decades: log10(price) − log10(curve). Zero is on the curve, and +0.3 is about double it.
5. Place the bands. The band edges lie at fixed offsets from the curve: −0.375, −0.225, −0.075, +0.075, +0.225, +0.375, +0.525 and +0.675 decades. That gives nine bands, each 0.15 decades wide and 41% above the one below. The two outer bands are open.
6. Find the band. The band is the number of edges at or below the offset, counted from the bottom. A price exactly on an edge belongs to the band above.
7. Turn the edges back into dollars. An edge price is 10^(log10(curve) + edge offset), so every dollar range moves up each day with the curve.

Data

BTC/USD daily prices from Bitstamp: one price per UTC date, taken at 00:00 UTC, which is the last trade of the day before. The window starts on 23 March 2012 because Bitstamp’s first months are thin and partly stale, so the earliest years of Bitcoin are not in the fit. It ends on 7 October 2026, the day the model was frozen. Three days in January 2015, when Bitstamp was suspended, carry the previous price and are kept in the fit. Later days come from CoinGecko. Over the ten days checked (28 September to 7 October 2026) the two sources’ 00:00 UTC prices agreed to within 0.03%. Bitstamp is a single exchange, so against a multi-venue index it can differ by more on volatile days. The headline reading uses the current CoinGecko price, refreshed every few minutes. The history, the chart and the shares of days use the 00:00 UTC prices.

Formula

log10(price) = −15.9514 + 2.4019 × ln(day number). The numbers are rounded for display; the model uses six decimals, −15.951385 and 2.401939, and those are the model. Recomputing a band from the rounded numbers can move a day that sits right on an edge into the next band. Written as a power law, the curve rises with the day number to the power 5.53. The line itself currently climbs about 35% a year, and that rate falls every year. This describes the line, not the price.

Epoch

Days are counted from 9 January 2009, six days after the first block, which is dated 3 January 2009. The exact day hardly matters: counting from 3 January instead and refitting the curve moves it by about 0.2%. A much later or earlier origin would move it: shifting the origin by a year and refitting moves the curve by about 9% to 11%.

Band width

Each band is 0.15 decades wide, which is 41% above the one below; two bands up multiply the price by 2.0. The middle band runs from 16% below to 19% above the curve. Three bands lie below the middle band and five above it. We chose that placement after looking at how far prices have strayed from the curve, so that every band has held days since 23 March 2012.

Fitted to the same data

The curve and the bands are drawn over the prices they were fitted to, so they follow the past by construction. The band widths are chosen, not estimated, and carry no probability. Some earlier rainbow charts, such as Blockchain Center’s, were redrawn after the price left their bands. A chart that adapts after each miss describes; it is not tested.

Frozen and versioned

The constants never change silently. A refit would be a new method version with the whole history recomputed and the change noted here. We review the model once a year, in January, and say whether it is unchanged. The price leaving the bands is not a reason to refit. This is method version 1.0.

Start date

Where the fit starts moves the curve. On 7 October 2026, a fit starting later would have put the curve as follows: one starting in 2013 about 11% lower, one starting in 2014 about 9% lower, one starting in 2015 about 9% higher, one starting in 2016 almost the same and one starting in 2017 about 17% lower. A fit that also used 2010 and 2011, from another data set, would have put it about 5% higher. Each shift is well under one band, but together they show that the curve is a choice.

What R² does not show

R² is 0.94 over the fit window: the line follows log price closely. A series that trends upward will usually give a high R², so it does not show that the line keeps working. The residuals are correlated at about 0.998 from one day to the next: the price stays above or below the curve for months or years, so the 5,312 days hold only a handful of independent cycles. The typical gap between price and curve, the residual standard deviation, is 0.2825 decades (a factor of 1.9, or 1.9 band widths of 0.15 decades).

Other sites

Blockchain Center, Bitbo and others publish their own versions of the rainbow chart: Bitbo uses six bands, while Blockchain Center’s original chart has ten and its current chart nine, like this page. Their curves and names differ, and some have been redrawn over time. On the same day their “current band” can therefore differ from the one on this page. A difference is not an error on either side: the bands are choices, not measurements. The constants and the texts here are our own.

Worked example

On 28 October 2026 the day number is 6,501, so ln(day number) = 8.779711 and log10(curve) = −15.951385 + 2.401939 × 8.779711 ≈ 5.1369. The curve is therefore at $137,071. The price of $80,872 has an offset of −0.2291 decades, which is 41.0% below the curve. That offset lies between the edges at −0.375 and −0.225, so the band is “Below the curve”.

The rainbow chart has several authors, and this page’s definition is its own. According to Blockchain Center’s own history, Reddit user azop posted the first rainbow-style log charts in 2014. In the same year BitcoinTalk user Trolololo published a logarithmic regression of Bitcoin’s price, the mathematical ancestor of the curved version, and revised it several times over the next years. Blockchain Center popularized the live chart and its playful labels from 2019, running from “Fire sale” to “Maximum bubble territory”; those labels are community names, not instructions. It added a lower band in 2022 after the price fell out of its original bands. Harold Christopher Burger and Giovanni Santostasi later described Bitcoin’s price as a power law in time, which is the same idea in other words. We use none of their constants or texts, the fit and the band widths are ours, and none of them endorses this page.

• It describes; it does not predict. The curve is a line fitted to past prices. Continuing it into the future is arithmetic, and the band edges shown for later dates are a curve extension, not a forecast.
• It was fitted to the data it describes. The curve and the bands follow the past by construction. Some earlier rainbow charts, such as Blockchain Center’s, were redrawn after the price left their bands, and ours will be reviewed every January: a refit would move every band.
• It rests on very few cycles. Residuals are correlated at about 0.998 from day to day, so the 5,312 days hold only a handful of independent cycles. An R² of 0.94 does not change that.
• The curve is a choice. Starting the fit in 2013, 2014, 2015, 2016 or 2017 moves the curve by between 0% and 17%, and other sites publish versions with other curves and names. Their “current band” can differ from ours on the same day.
• The bands are chosen widths, not odds. Each band is 41% above the one below, a width we picked. The shares of days are history and give no odds of reaching a band. The largest distance above the curve has shrunk from cycle to cycle, and fixed bands cannot show that.
• A price near an edge is a weak statement. A price within a percent or so of an edge can fall on either side depending on which exchange’s price is used and when it is read. That is why this page prints the distance to the neighboring bands beside the band name.