Why we measure crypto on weekdays: 365 versus 252

Every price on this page is invented. The coin and the index fund are two series generated for this lesson from a fixed seed: they are not any real coin or fund, and none of the figures is market data. The coin is not the Orbit (ORB-USD) that the example portfolio of this site holds: it is a separate invented series, so the figures here are not that portfolio's.

Seven days against five

A share has a price when its exchange is open. A coin has one every day. In 2025 the invented coin had 365 daily closes and the invented index fund 250 sessions: 365 − 250 = 115 days on which the coin has a close and the fund has none, 104 of them weekend days and 11 weekday holidays.

A portfolio is rebuilt on one table with one row per date, and every figure on this site is computed from that table. A portfolio that holds both therefore needs one calendar, and this product uses the shares': the days the exchange was open, or every weekday when the portfolio holds only coins.

Folding the weekend onto Monday

On each row, a coin's price is its last price on or before that date, set on the calendar before the series are joined. A Monday row therefore holds the whole weekend, and the return of that row runs from Friday's close to Monday's:

• P_Mon, P_Fri — the coin's closes on the Monday and on the Friday before it.
• r_Sat, r_Sun, r_Mon — its own daily returns on the Saturday, the Sunday and the Monday.
• R_Mon — the return of the Monday row.

Nothing is invented or lost, only spread over fewer rows, so the total return between any two dates of the table is the raw one, P_end / P_start − 1. The shares are not touched.

Why not simply use all seven days? Because a share has no price on a Saturday: its last price would be carried over and the day would count as a return of exactly 0.0%. From 1 October 2025 to 30 September 2026 the invented fund has 250 daily returns on its sessions, and 115 of the 365 rows of a seven-day table would be zeros. Annualised with 252, its volatility reads 13.5% on the sessions and 11.2% on the seven-day table, about 17% lower, because of days on which nothing happened. On the weekday table that cannot happen, so a share's volatility is the same with or without a coin in the portfolio.

A worked example: the weekend of 11 to 14 May 2018

The invented coin's closes, each day ending at 00:00 UTC:

On the weekday table the three days are one row, Monday's: 2,169.69 ÷ 2,491.86 − 1 = −12.9%, and (1 − 0.024) × (1 − 0.072) × (1 − 0.038) − 1 = −12.9% as well. It is the largest fall of the 522 Friday-to-Monday moves since January 2015. On that same Monday the invented fund has one row too: +0.3%.

What it costs

1. Timing. A weekend's move shows on the Monday, not on the day it happened. A fall on a Sunday evening is in Monday's row.
2. Extremes inside the weekend. The invented coin's highest close of early 2018 was on Saturday 13 January 2018: 8,250.00. A weekday table does not have that day. Its high is Tuesday 9 January 2018, 8,140.05, 1.3% lower, and the fall that followed reads −86.4% instead of the −86.6% measured on all seven days.
3. The hours between closes. The coin's day ends at 00:00 UTC and US shares close at 16:00 New York time, three to four hours earlier depending on daylight saving. Two series that close at different hours look less correlated than they are, so the daily correlation between a share and a coin reads slightly low. It is not corrected: that would mean inventing a closing time the data does not have.
4. The annualisation, which is the next section.

The annualisation: 261, not 365 and not 252

A portfolio of coins alone is annualised with 261 periods a year. In the twelve months to 30 September 2026 its weekday table has 261 rows and a seven-day table has 365. The same coin, the same year, four ways:

The first two are estimates of the same quantity from different samples, so they are close and not equal. The third is 18% too high: a row of the weekday table stands for 1.4 calendar days on average, a Monday for three, so treating each row as one calendar day overstates the year. The last is 1.7% too low: 252 is the shares' count after removing exchange holidays, and a table of weekdays has no holidays removed.

• σ_row — the standard deviation of the table's daily returns.
• N — the number of rows the table has in a year: about 252 sessions when the portfolio holds any share, 261 weekdays when it holds only coins.
• σ_year — the annualised volatility.

N is simply the number of rows because, with independent returns, the variance of a year is the sum of the variances of its rows. A year has 365.25 ÷ 7 = 52.18 weeks of five weekdays, 260.9 ≈ 261 rows. If every calendar day had a variance s², a Monday row would have 3s² and the other four weekdays s² each: the mean row has (3 + 4) ÷ 5 = 1.4s², and 261 × 1.4s² = 365.4s², the variance of the year. Using 365 would give 511s².