Is one weekday better than the others for crypto? To find out, we took five years of daily closing prices for six large coins, from November 1, 2021 to October 31, 2026, and sorted every daily return by its UTC weekday. The coins are bitcoin, ether, solana, XRP, dogecoin and Cardano's ada. The averages do differ from day to day; the harder question is whether the differences are larger than chance would produce. The crypto hub covers what each coin is and the risks of owning them.
Does the day of the week affect crypto returns?
Not in a way the data can confirm: only 2 of 42 coin-and-weekday averages reached an absolute t-statistic of 2, against 2.1 expected by chance. Volatility tells a clearer story, because weekend returns swung less than weekday returns for every coin.
- Weekday averages look like noise. Two combinations crossed the line of an absolute t-statistic of 2: bitcoin on Thursdays (t = −2.02) and dogecoin on Thursdays (t = −2.12). At a 5% significance level, 42 tests would be expected to produce 2.1 such results by chance.
- Thursday was weak, but not convincingly. It had the lowest average return for five of the six coins, from −0.24% for solana to −0.54% for dogecoin. The exception was XRP, whose lowest day was Sunday at −0.17%.
- Weekends were calmer for every coin. The weekend standard deviation averaged 0.79 times the weekday figure, ranging from 0.61 for bitcoin to 0.99 for ada.
- Bitcoin's pattern did not hold still. On Bitstamp data, its weakest day was Saturday in 2012–2016 (−0.18%), Sunday in 2017–2021 (0.06%) and Thursday in 2022–2026 (−0.28%).
- Averages and medians often disagreed. Ether's best-average day, Wednesday at 0.38%, had a median of −0.12% and only 47.9% up days.
| Measure | Value |
|---|---|
| Window | Nov 1, 2021 to Oct 31, 2026 |
| Coin-weekday combinations tested | 42 |
| Combinations with |t| of 2 or more | 2 |
| Expected by chance at a 5% significance level | 2.1 |
| Average weekend ÷ weekday volatility across the six coins | 0.79 |
How were the weekday returns measured?
We split five years of daily closes for six coins by UTC weekday, then compared each weekday's average return with that coin's average across all days.
| Measure | How it is calculated |
|---|---|
| Daily return | The day's closing price divided by the previous day's close, minus 1. Days run from midnight to midnight UTC, so Saturday's return is the move during Saturday, from Friday's close to Saturday's close. |
| Average return | The mean of all daily returns that fall on that weekday. |
| Median return | The middle daily return for that weekday, which a few extreme days cannot drag around. |
| Standard deviation | How widely that weekday's returns spread around their average, the usual measure of volatility. |
| Share of up days | The share of that weekday's returns above zero. |
| t-statistic vs all days | The weekday average minus the coin's all-day average, divided by the weekday's standard deviation over the square root of its number of days. |
| Weekend ÷ weekday | The standard deviation of Saturday and Sunday returns divided by that of Monday to Friday returns. |
| Prices | Coinbase Exchange USD daily candles for BTC, ETH, SOL, DOGE and ADA. Bitstamp daily candles for XRP and for the long bitcoin comparison. |
| Window | November 1, 2021 to October 31, 2026, giving 261 of each weekday from Monday to Saturday and 260 Sundays per coin. The bitcoin periods run 2012–2016, 2017–2021 and January 2022 to October 2026. |
t = (weekday average − all-day average) ÷ (weekday standard deviation ÷ √number of days)
XRP comes from Bitstamp because Coinbase did not list it from January 2021 to July 2023, a gap that overlaps the study window, so its figures come from a different order book than the other five. Coinbase's candle documentation stamps each daily candle with the start of its UTC day, which is how each return is assigned to a weekday.
Time zones matter more than they seem. Midnight UTC is 8 p.m. in New York during daylight saving time and 7 p.m. in winter, so this study's Saturday begins on Friday evening New York time. The NYSE's core trading session runs from 9:30 a.m. to 4 p.m. Eastern on weekdays, which means a UTC Saturday or Sunday contains no regular US stock trading at all, while a UTC Monday contains Monday's full session. Our guide to stock market hours and holidays has the full calendar.
An absolute t-statistic of 2 is a common rule of thumb for significance at the 5% level; NIST's table of critical values puts the exact two-sided cutoff at 1.984 with 100 degrees of freedom and 1.960 for very large samples.
Worked example: How big a weekday gap needs to be before it counts
Bitcoin's Thursdays had a standard deviation of 2.81% over 261 days. Dividing by √261 ≈ 16.16 gives a standard error of about 0.17 percentage points, so a Thursday average needed to sit roughly 0.35 points away from bitcoin's all-day average to reach a t-statistic of 2. The actual Thursday average of −0.30% produced t = −2.02, just over the line.
Which weekdays had the highest and lowest average returns?
Thursday had the lowest average for five of six coins, while the highest fell on Monday, Wednesday or Saturday, depending on the coin.
| Coin | Highest average | Lowest average | |t| ≥ 2 |
|---|---|---|---|
| BTC | Monday (0.30%) | Thursday (-0.30%) | Thursday |
| ETH | Wednesday (0.38%) | Thursday (-0.39%) | None |
| SOL | Saturday (0.39%) | Thursday (-0.24%) | None |
| XRP | Monday (0.48%) | Sunday (-0.17%) | None |
| DOGE | Saturday (0.56%) | Thursday (-0.54%) | Thursday |
| ADA | Saturday (0.21%) | Thursday (-0.34%) | None |
The gap between a coin's best and worst day ran from 0.55 points for ada, 0.21% on Saturdays against −0.34% on Thursdays, to 1.10 points for dogecoin, 0.56% on Saturdays against −0.54% on Thursdays. Bitcoin's gap was 0.60 points, from 0.30% on Mondays to −0.30% on Thursdays.
Those gaps are small next to a normal day's swing: bitcoin's Thursday standard deviation was 2.81%, more than nine times the size of its Thursday average.
Worked example: What a 0.30% average day means on $10,000 of bitcoin
Bitcoin's Monday average of 0.30% works out to $30 a day on a $10,000 holding. Its Monday standard deviation of 3.28% works out to about $328, more than ten times the average gain, and a typical Monday could land on either side of zero.
Medians tell a more cautious story than averages. Ether's best-average day, Wednesday at 0.38%, had a median of −0.12% and only 47.9% up days. Bitcoin's Wednesday showed the same shape, with an average of 0.23%, a median of −0.14% and 47.5% up days. On both coins more Wednesdays fell than rose, and a handful of large up days lifted the average. Shares of up days stayed in a narrow band, from 41.2% for dogecoin and ada on Sundays to 57.5% for ether on Saturdays.
Why don't the weekday averages hold up?
Because the gaps are small next to daily swings, and testing 42 combinations makes a couple of chance results likely even if no weekday effect exists.
The multiple comparisons problem is easy to picture with ordinary coin tosses. Toss 42 fair coins 261 times each, and a couple will land on heads, or on tails, often enough to look biased, purely by luck. A test at the 5% significance level is built to raise a false alarm about once in 20 tries when nothing is going on, so 42 tries should produce about 2.1 false alarms. This study found 2.
NIST's statistics handbook makes the same point about multiple comparisons: across many comparisons, the chance of at least one false positive is far higher than the significance level of any single test, which is why comparisons should be chosen before the data are examined.
Worked example: Counting false alarms across 42 tests
Suppose no weekday effect existed and the 42 tests were independent, each with a 5% chance of a false alarm.
- Expected false alarms: 42 × 0.05 = 2.1.
- Chance of at least one: 1 − 0.95⁴² ≈ 88%.
- Chance of two or more: about 63%.
A Bonferroni correction, which divides the 5% level by the number of tests, sets the bar at 5% ÷ 42 ≈ 0.12% per test, roughly an absolute t-statistic of 3.2. The largest in the study was 2.12, for dogecoin's Thursdays.
Correlation weakens the evidence further. The six coins tend to rise and fall on the same days: bitcoin's and ether's daily returns had a correlation of 0.81 from 2021 through 2025, according to our bitcoin vs ethereum comparison. A rough patch of Thursdays for bitcoin is likely to show up in the other coins too, so five weak Thursdays are closer to one observation repeated than to five independent confirmations. The same correlation makes the 88% figure above an approximation, since it assumes independent tests.
Choosing what to test after seeing the table adds one more bias. Thursday drew attention here because it came out lowest; had Tuesday come out lowest, Tuesday would have drawn the same attention. The full table lists all 42 combinations, with the t-statistic in the last column.
| Coin | Day (UTC) | Days | Average return | Median return | Standard deviation | Share of up days | t-statistic vs all days |
|---|---|---|---|---|---|---|---|
| BTC | Monday | 261 | 0.3% | 0.1% | 3.3% | 52.1% | 1.24 |
| BTC | Tuesday | 261 | 0.0% | -0.1% | 2.6% | 48.7% | -0.28 |
| BTC | Wednesday | 261 | 0.2% | -0.1% | 3.0% | 47.5% | 1.01 |
| BTC | Thursday | 261 | -0.3% | -0.3% | 2.8% | 44.4% | -2.02 |
| BTC | Friday | 261 | -0.0% | -0.1% | 3.0% | 48.3% | -0.51 |
| BTC | Saturday | 261 | 0.0% | 0.1% | 1.4% | 53.3% | -0.40 |
| BTC | Sunday | 260 | 0.1% | 0.0% | 2.1% | 50.8% | 0.69 |
| ETH | Monday | 261 | 0.3% | 0.3% | 4.1% | 53.3% | 0.88 |
| ETH | Tuesday | 261 | -0.2% | -0.1% | 3.4% | 47.5% | -0.90 |
| ETH | Wednesday | 261 | 0.4% | -0.1% | 4.1% | 47.9% | 1.37 |
| ETH | Thursday | 261 | -0.4% | -0.3% | 4.0% | 44.1% | -1.70 |
| ETH | Friday | 261 | -0.1% | 0.0% | 3.8% | 51.0% | -0.32 |
| ETH | Saturday | 261 | 0.1% | 0.2% | 2.3% | 57.5% | 0.61 |
| ETH | Sunday | 260 | 0.1% | 0.0% | 3.0% | 50.0% | 0.17 |
| SOL | Monday | 261 | -0.1% | -0.2% | 5.1% | 47.5% | -0.66 |
| SOL | Tuesday | 261 | -0.0% | 0.1% | 4.4% | 50.6% | -0.41 |
| SOL | Wednesday | 261 | 0.2% | -0.1% | 6.0% | 49.8% | 0.37 |
| SOL | Thursday | 261 | -0.2% | -0.6% | 5.0% | 45.2% | -1.04 |
| SOL | Friday | 261 | 0.2% | 0.1% | 5.1% | 51.3% | 0.37 |
| SOL | Saturday | 261 | 0.4% | 0.1% | 4.1% | 51.3% | 1.23 |
| SOL | Sunday | 260 | 0.2% | -0.2% | 4.1% | 47.3% | 0.31 |
| XRP | Monday | 261 | 0.5% | 0.5% | 4.5% | 56.3% | 1.39 |
| XRP | Tuesday | 261 | 0.1% | -0.1% | 3.8% | 49.0% | -0.19 |
| XRP | Wednesday | 261 | -0.1% | -0.2% | 4.2% | 47.1% | -0.66 |
| XRP | Thursday | 261 | -0.1% | -0.5% | 6.2% | 44.1% | -0.46 |
| XRP | Friday | 261 | 0.2% | 0.1% | 4.5% | 50.6% | 0.36 |
| XRP | Saturday | 261 | 0.3% | 0.1% | 3.1% | 52.1% | 0.88 |
| XRP | Sunday | 260 | -0.2% | -0.4% | 3.7% | 45.8% | -1.14 |
| DOGE | Monday | 261 | -0.1% | -0.1% | 5.2% | 46.7% | -0.38 |
| DOGE | Tuesday | 261 | -0.1% | -0.2% | 4.7% | 47.5% | -0.37 |
| DOGE | Wednesday | 261 | 0.0% | -0.4% | 4.8% | 46.7% | -0.03 |
| DOGE | Thursday | 261 | -0.5% | -0.7% | 4.3% | 43.7% | -2.12 |
| DOGE | Friday | 261 | 0.5% | 0.5% | 4.6% | 57.1% | 1.55 |
| DOGE | Saturday | 261 | 0.6% | 0.2% | 4.6% | 54.4% | 1.83 |
| DOGE | Sunday | 260 | -0.1% | -0.5% | 4.3% | 41.2% | -0.57 |
| ADA | Monday | 261 | -0.1% | 0.0% | 4.6% | 50.6% | -0.26 |
| ADA | Tuesday | 261 | -0.2% | -0.4% | 4.0% | 45.6% | -0.87 |
| ADA | Wednesday | 261 | 0.2% | -0.4% | 4.8% | 46.7% | 0.64 |
| ADA | Thursday | 261 | -0.3% | -0.3% | 4.5% | 45.6% | -1.14 |
| ADA | Friday | 261 | 0.2% | 0.0% | 5.2% | 49.8% | 0.71 |
| ADA | Saturday | 261 | 0.2% | 0.1% | 3.0% | 52.5% | 1.28 |
| ADA | Sunday | 260 | -0.1% | -0.6% | 5.8% | 41.2% | -0.12 |
Did bitcoin's weekday pattern last across periods?
Only partly: Monday had bitcoin's highest average in all three periods, but the weakest day moved from Saturday to Sunday to Thursday.
| Day (UTC) | 2012–2016 | 2017–2021 | 2022–2026 |
|---|---|---|---|
| Monday | 1.0% | 0.7% | 0.3% |
| Tuesday | 0.6% | 0.2% | 0.0% |
| Wednesday | 0.5% | 0.4% | 0.3% |
| Thursday | 0.3% | 0.1% | -0.3% |
| Friday | 0.4% | 0.3% | 0.0% |
| Saturday | -0.2% | 0.4% | 0.0% |
| Sunday | 0.2% | 0.1% | 0.1% |
This table uses Bitstamp's longer bitcoin history, split into 2012 to 2016, 2017 to 2021 and January 2022 to October 2026. The last period overlaps the main window but starts two months later and uses a different exchange, which is why its Thursday figure of −0.28% differs slightly from the −0.30% in the main table.
Several days traded places:
- Thursday slid from 0.35% to 0.07% to −0.28%, ending as the weakest day.
- Saturday went from the weakest day in 2012–2016 at −0.18% to the third strongest in 2017–2021 at 0.36%, then to 0.00% in 2022–2026.
- Tuesday ranked second in 2012–2016 at 0.63% and averaged 0.00% in 2022–2026.
- Monday stayed first throughout, but its lead over the next-best day shrank from 0.36 points (0.99% against Tuesday's 0.63%) to 0.22 points and then to 0.07 points (0.33% against Wednesday's 0.26%).
Six of the seven averages were lower in 2022–2026 than in 2012–2016, Saturday being the exception, so part of every change is simply a lower overall level of daily returns. The table carries no significance test, and with only three periods, Monday's steady first place could still be luck; in the main window its t-statistic was 1.24.
Patterns that fade once noticed are familiar from stock markets. G. William Schwert's review of market anomalies for the National Bureau of Economic Research found that the stock weekend effect and other calendar patterns weakened or vanished after research documenting them was published.
Is crypto calmer on weekends?
Yes: for all six coins, weekend returns swung less than weekday returns, with the weekend-to-weekday volatility ratio averaging 0.79.
| Coin | Weekday standard deviation | Weekend standard deviation | Weekend ÷ weekday | Weekday average | Weekend average |
|---|---|---|---|---|---|
| BTC | 3.0% | 1.8% | 0.61 | 0.0% | 0.1% |
| ETH | 3.9% | 2.7% | 0.68 | 0.0% | 0.1% |
| SOL | 5.1% | 4.1% | 0.80 | 0.0% | 0.3% |
| XRP | 4.7% | 3.4% | 0.72 | 0.1% | 0.1% |
| DOGE | 4.8% | 4.5% | 0.94 | -0.0% | 0.2% |
| ADA | 4.6% | 4.6% | 0.99 | -0.1% | 0.1% |
Bitcoin showed the biggest difference. Its weekend standard deviation was 1.79% against 2.95% on weekdays, a ratio of 0.61, and its Saturdays were the calmest coin-and-weekday combination in the study at 1.41%. Ether at 0.68 and XRP at 0.72 followed. At the other end, dogecoin at 0.94 and ada at 0.99 were almost as volatile on weekends as on weekdays, and ada's most volatile day of the week was Sunday, at 5.78%. Saturday was the calmest day for five of the six coins; dogecoin's calmest was Sunday, at 4.33%.
Worked example: Bitcoin's typical daily swing on $10,000
With a weekday standard deviation of 2.95%, a typical weekday move on $10,000 of bitcoin was about $295 up or down. On weekends, at 1.79%, it was about $179, and on Saturdays alone, at 1.41%, about $141.
The data do not say why, but the calendar of other markets offers clues. US stock exchanges are shut on weekends, and the spot bitcoin exchange-traded products the SEC approved on January 10, 2024 trade only when those exchanges are open. CME Group's regulated crypto futures did not trade on weekends until May 29, 2026, when the exchange moved them to 24/7 trading. Nearly all of the study window therefore comes from a market in which large regulated venues paused every weekend, and later data will show whether the weekend calm survives that change.
Calmer does not mean better. Weekend averages were higher than weekday averages for five of the six coins, XRP being the exception, but no Saturday or Sunday t-statistic reached 2; the largest was 1.83, for dogecoin's Saturdays. For a sense of scale against stocks, see our study of how volatile the biggest stocks are.
What can't this weekday data tell you?
It cannot say why a pattern appeared, whether it will continue or what happened within each day, because it rests on one closing price per day.
- The day boundary is a choice. Midnight UTC puts Friday evening in New York into Saturday. A different cutoff, such as the 4 p.m. New York close, would move every weekday boundary and change every average.
- Five years is a short sample for small effects. With about 261 observations per weekday, a real difference of a few hundredths of a percentage point would be impossible to detect. The study cannot rule out tiny effects; it can only say that no large, stable one showed up.
- The tests overlap. The 42 comparisons share the same calendar days and closely linked coins, so the count of two significant results is not the outcome of 42 independent experiments.
- Extreme days carry a lot of weight. A handful of very large moves can shift a weekday's average or standard deviation, as the gaps between averages and medians show, and the t-statistic is only an approximation when returns are this uneven.
- Exchange prices differ slightly. Results reflect Coinbase Exchange and Bitstamp closes, and other venues would give somewhat different numbers. Fees and spreads are not included.
- Survivors only. The six coins are among the largest today. Coins that faded during the window are not in the sample.
- Price only. Returns leave out staking rewards on coins such as ether, solana and ada.
What do the results mean for anyone buying crypto on a schedule?
They suggest the choice of purchase weekday matters far less than the price swings around it, and that weekend calm describes averages, not any particular weekend.
Past performance does not predict future results, and a weekday pattern found in one five-year window is weak ground for any plan. Three readings follow from the numbers:
- An average daily return is not a cheap price. Bitcoin's 0.30% Monday average describes the move during Monday. Capturing it would mean buying at Sunday's close, and a 3.28% Monday standard deviation means any given Monday could easily go the other way.
- Timing within a week is small next to timing within a year. In our bitcoin DCA backtest, weekly and monthly purchase schedules ended with median 12-month returns 1.3 points apart, while a one-month change in the start date could move the lump-sum result by more than 100 points.
- Costs are certain; weekday edges are not. Fees and spreads are paid on every purchase, while the weekday gaps in this study cannot be told apart from chance.
Our guide to dollar-cost averaging in crypto covers how recurring purchase plans work, and the standard deviation glossary entry explains the volatility measure used throughout this page.
How do you cite or reuse the weekday dataset?
You may copy, adapt and republish the data under CC BY 4.0, commercially or not, provided you credit the source, link to this page and note any changes.
The weekday CSV contains the 42 rows of the main table, one per coin and weekday, with returns and standard deviations in percent and the t-statistic as a plain number. Please cite it as “Source: Sonic Wallet analysis of Coinbase Exchange and Bitstamp daily candles, data through October 31, 2026.” The license deed sets out the terms in full, and our editorial policy describes how these figures are produced, checked and corrected. When the dataset is refreshed, the dates at the top of this page change with it.