Is “Uptober” Real? Bitcoin Rose in the Second Half of October in 10 of the Past 12 Years, a Median +7.8% — but It Is Down 2.5% So Far This Year, and the Years That Started Weak Were a Coin Flip

“Uptober” is the crypto community’s name for the idea that October is bitcoin’s friendliest month. By our calculation it has a basis: bitcoin finished October higher in 9 of the 12 years from 2014 to 2025, with a median gain of 12.9%. But 2026 is not following the script. Through Friday 16 October, bitcoin is down 2.46%, from a 30 September close of $83,556.14 to $81,500.04 on Coinbase, and TheStreet noted on 16 October that the usual October rally “failed to materialize” (TheStreet, 16 October 2026). So is the pattern real, and does a weak first half tell us anything about the second? We ran the numbers.

The data and the method

We took daily closes (00:00 UTC to 24:00 UTC) for bitcoin from Yahoo Finance’s BTC-USD history, which starts on 17 September 2014, and measured three things for each year from 2014 to 2025: the change from the 30 September close to the 16 October close (the “first half”, matching where 2026 stands today), the change from the 16 October close to the 31 October close (the “second half”), and the full month. For 2026 we used Coinbase Exchange candles for the same dates. Yahoo’s early-year prices come from aggregated sources and differ from any single exchange, so the 2014–2016 figures are approximate.

Twelve Octobers in one table

Three headline results stand out. The full month was positive in 9 of 12 years (mean +16.6%, median +12.9%). The first half was positive in 7 of 12 years (mean +7.9%, median +5.1%), and 2026’s −2.5% is the third-weakest of the 13 first halves, behind only 2025 (−5.1%) and 2019 (−3.0%). The second half was positive in 10 of 12 years, with a mean of +8.0% and a median of +7.8%.

Is 10 of 12 impressive? It depends on the base rate

Bitcoin rises more often than it falls over any short window, so a “10 of 12” needs a yardstick. Across all 4,124 rolling 15-day windows that began between September 2014 and the end of 2025, bitcoin rose in 56.6% of them, with a mean gain of 3.0% and a median of 1.4%. Against that, the second half of October (10 of 12, or 83%, with a mean of +8.0%) is stronger than the typical fortnight. A simple binomial test, treating each year as an independent coin flip with a 56.6% chance of an up-move, gives a probability of about 5.3% of seeing 10 or more up-years out of 12 by luck alone. That is borderline by the usual 5% convention, and it comes with a large caveat: we looked at several slices of October (first half, second half, full month) and reported the most striking one, which inflates the apparent significance. We would treat it as suggestive, not proven.

There are other reasons for caution. The 12 observations are not independent in the way a coin flip is: several of the years (2015, 2017, 2020, 2021 and 2023) were years in which bitcoin rose strongly over the whole twelve months, which lifts the odds for any single month. The standard error of the second-half mean is 3.0 percentage points on a standard deviation of 10.5, so the mean of +8.0% is distinguishable from zero but the range of plausible values is wide. And a pattern that is widely publicised can lose its edge, since traders who position early for “Uptober” pull the gain forward.

The video above was published on 2 October, when bitcoin was near $84,500; we include it as an example of the optimistic Uptober framing at the start of the month, not as a forecast. Two weeks later the price is about $3,000 lower.

Does a weak first half predict a weak second half?

This is the question that matters for 2026. We split the 12 years by how the first half went. In the five years when bitcoin fell during the first half (2014, 2018, 2019, 2022 and 2025), the second half returned −11.6%, −4.2%, +14.3%, +6.4% and +1.3%: three up, two down, a mean of +1.2% and a median of +1.3%. In the seven years when the first half rose, the second half was positive every time, with returns of +19.5%, +9.2%, +13.0%, +21.7%, +0.7%, +21.6% and +3.8%, a mean of +12.8% and a median of +13.0%.

So the data are consistent with a momentum reading: a good start has been followed by a good finish, and a weak start has been followed by a toss-up. The difference in mean second-half return between the two groups is 11.6 percentage points with a standard error of about 5.5, a t-statistic of roughly 2.1. With groups of 5 and 7 years that is thin evidence, and it is another slice we chose after looking at the data, so we would not trade on it. What it does say is that, in this sample, the best Octobers started well, and 2026 did not.

What “Uptober” does not tell you

Seasonality studies describe the past; they do not describe the causes of the current week. This week, bitcoin’s moves have been linked more to inflation (September CPI rose 3.7% year on year, per our 15 October report), to oil near five-month highs and to a stock-market selloff in debt-funded AI companies (today’s market report). Those factors do not read a calendar. The same applies to the 15 days left in the month: key events include the Federal Reserve’s 27–28 October meeting and the 29 October PCE report, and any of them can matter more than a seasonal tendency.

There is also the structural change since 2024: US spot bitcoin ETFs, which now trade alongside the spot market and whose flows we track daily. October’s flows so far are net negative (see our latest ETF flows update), which is not what a seasonal inflow story would predict. Only two of the 12 Octobers in our table (2024 and 2025) took place after the ETFs launched in January 2024, so the sample says little about how today’s market behaves.

The Yahoo Finance segment above is from 18 October 2025, when a similar “October looks different” argument was being made. In that year bitcoin did fall in October (−3.9% on our table), and the second half returned only +1.3%, a result that sits in the weak-start group above.

What the numbers support

Taken together, the data support four statements and no forecast. First, bitcoin has risen in the second half of October more often than in a typical 15-day window (10 of 12, against a 56.6% base rate). Second, that edge is borderline in statistical terms and comes from a small, bull-market-heavy sample. Third, in this sample years that began weakly did not show the same strength as years that began well. Fourth, 2026 has begun weakly, down 2.46% against a first-half median of +5.1%. Whether the next 15 days look like the median second half (+7.8%) or the weak-start median (+1.3%) is not something this history can settle.

Method: bitcoin prices, highs, lows and closes are Coinbase Exchange candles (UTC day or UTC hour) pulled by Bitcoin Mastery at about 06:15 UTC on Saturday 17 October 2026; the 17 October candle is still open at that time, so any closing figure quoted is the 16 October close unless stated. US index, ETF, stock and Brent closes are Yahoo Finance daily data or the Associated Press figures cited; other third-party figures are attributed to their source and date. Yearly returns for 2014–2025 use Yahoo Finance BTC-USD daily closes (UTC); the 2026 figure uses Coinbase Exchange daily closes. Means, medians, standard errors, the binomial probability (p = 0.566, 12 trials) and the Welch t-statistic are our own calculations; the 4,124 rolling windows are 15-calendar-day windows starting on each day with data between 17 September 2014 and 31 December 2025 and overlap one another, so the base rate describes a long-run frequency, not independent samples.

Disclaimer: This article is for informational purposes only and does not constitute investment advice. Bitcoin and other cryptocurrencies are volatile and you can lose some or all of the money you put in. Nothing here is a recommendation to buy, sell or hold bitcoin, any exchange-traded fund, any listed security or any other asset, and the technical levels, probabilities and scenarios discussed are descriptions of published data, not forecasts. Do your own research and consult a licensed financial advisor before making investment decisions.