Bitcoin's returns, year by year

The first thing a reader notices is the asymmetry. Bitcoin's best years are not merely better than its worst years are bad; they are better by an order of magnitude. The strongest year in the series is 2013, which opened at $13.50 and closed at $754, a gain of roughly 5,485 per cent. The worst year is 2018, which opened at $14,120 and closed at $3,744, a loss of about 73 per cent. Put those two numbers side by side and the point becomes unavoidable: a single good year can erase several bad ones, and a single bad year cannot erase a good one.

This is not a claim that Bitcoin always recovers, and it is not a prediction. It is a description of the record as published. The asymmetry exists because the asset has spent most of its life in price discovery, with a fixed supply and a demand curve that has repeatedly repriced the whole market in the space of months. The same mechanism that produces a 5,485 per cent year produces a 73 per cent year. They are not opposites; they are the same property observed at different moments.

The second thing the table shows is how little a single year tells you. Consider 2021: the year closed up 59 per cent, which sounds like a good year, and it was. But the price also fell from a high of $68,790 to a low of $29,300 within those same twelve months — a decline of more than half from the peak. A reader who bought at the top and sold at the bottom lost most of their capital in a year that the annual figure records as a gain. The calendar-year number is a summary of two points, and the path between them is where the investor actually lives.

The third observation is about the early years. The 2010 to 2014 figures come from a market that was thinly traded, with no established exchange and no reliable settlement price. The 2010 row shows an open of five cents and a close of thirty cents, a sixfold gain, but the entire year's trading volume would not fill a single modern block. These rows are included because they are the widely cited annual extremes and because omitting them would flatter the later record by making the base smaller. They should be read as historical reference points, not as prices anyone could have transacted at size.

Finally, the table makes clear why the choice of start date matters more than almost any other decision in return analysis. A reader who starts in 2010 sees a 0-year record with 0 down years. A reader who starts in 2018 sees a shorter record that begins with a 73 per cent loss. Both are looking at the same underlying series. The difference is entirely in where the window opens.

The calendar year is a convention of accounting, not a property of markets. Bitcoin does not know when December ends. A holding period that runs from March to the following February is exactly as real as one that runs from January to December, and it can produce a very different number. When the annual figures are read as a sequence of independent trials, the reader is implicitly assuming that each January resets the asset's behaviour, which is not true.

Rolling periods correct for this by measuring every possible window of a given length rather than the windows that happen to align with the calendar. A one-year rolling return starting in November 2021, for instance, captures the collapse from the all-time high and looks nothing like the calendar-year 2021 figure of plus 59 per cent. A three-year rolling return starting in 2019 captures the pandemic crash, the 2021 peak and the 2022 bear market in a single number. The calendar table is the right place to start because it is legible and comparable across assets; rolling periods are the right place to finish because they describe what a holder actually experienced.

The practical consequence is that any statement of the form "Bitcoin returned X per cent per year" is incomplete without the window it was measured over. The dollar-cost averaging page takes this seriously by measuring steady accumulation rather than a single entry point, and the drawdown record supplies the other half of the picture: how far the price fell, and how long the recovery took.

The calendar table answers a narrow question: what did the price do between two fixed dates? Most readers arrive with a different one. They want to know what a purchase would have returned, or what rate of growth would have turned one balance into another over a number of years. Those questions are answered by two measures that are related but not interchangeable, and the distinction between them is worth stating plainly before the reader meets them in detail.

Return on investment, or ROI, is the simplest of the two. It is the total change in value over a holding period, expressed as a percentage of what was put in. If a position is opened at one price and closed at another, ROI is the difference divided by the entry price. It says nothing about how long the money was committed, which makes it useful for comparing outcomes and useless for comparing rates. A 300 per cent ROI earned over two years and a 300 per cent ROI earned over ten are the same number and very different experiences.

Compound annual growth rate, or CAGR, supplies the missing dimension. It is the single steady annual rate that would take the starting value to the ending value over the stated number of years, assuming the growth compounded. It is a smoothed figure, not a description of any actual year — Bitcoin has never grown at a steady rate for any extended period — but it is the only honest way to put a two-year holding and a ten-year holding on the same scale. The arithmetic behind both measures, the assumptions each one smuggles in, and the reasons a volatile asset strains them are the subject of the ROI and CAGR explainer, which works through the definitions with published figures and shows the calculation in prose rather than hiding it behind a form.

One caution belongs here rather than there. Both measures depend entirely on the two endpoints chosen, and Bitcoin's endpoints are unusually consequential. A CAGR measured from the 2015 low and the 2021 high is a spectacular number; the same calculation shifted by eighteen months in either direction is a much more modest one. That sensitivity is not a flaw in the measures. It is an accurate reflection of an asset whose price has moved in long, violent swings, and it is the reason this site reports the window alongside every rate it quotes.