Crypto seasonality and statistically tested patterns

Day of week, weekend, turn of month and month by year seasonality for crypto, plus conditional patterns, each with sample size, 95% confidence interval, FDR significance and walk-forward checks.

The seasonality module looks for recurring tendencies in crypto prices: how a coin behaved on each day of the week, over weekends, around the turn of the month and in each month of the year, and what historically happened after a given condition.

Patterns are filtered against noise. Every result shows its sample size and a 95% confidence interval, significance is corrected for testing many patterns at once (false discovery rate), and walk-forward checks verify that a pattern also held on data it was not found on.

Seasonality describes the past. It is not a forecast, and the module says so next to every result.

What the module shows

How to read the data

Recurring, statistically significant historical tendencies in prices: day-of-week seasonality (e.g. "BTC Mondays") and conditional probability (what historically happened AFTER a given condition). Each row describes the past with anti-noise safeguards: not a forecast.

Base vs Conditional, n (sample size), 95% CI, Seasonality charts, Samples on a chart, Whiskers = confidence interval, FDR-significant, Walk-forward

The average daily price change (%) for each day of the week separately over the selected window (1–3 years). Each bar is the plain arithmetic mean of the daily returns falling on that weekday, and the whiskers show the 95% confidence interval of that mean. A historical tendency, not a forecast.

Mean change (%), Whiskers / 95% CI, n (sample size), FDR-significant, Stable over time (walk-forward)

The month × year heat map: each cell is the total price change in that month of that year, and the summary column is the month's average across all years in the window. Color encodes direction and strength (green = up, red = down). A month row only gets badges once it has enough observed years.

Cell (month, year), Month average, FDR-significant, Stable over time (walk-forward)

The list of detected conditional patterns: “after a day meeting condition X, the next day was up in Y% of cases”. Each row shows the base rate, the conditional rate, the edge in percentage points, n, the 95% (Wilson) confidence interval and the quality badges. Only patterns that passed the significance threshold are published.

Edge (pp), n (sample size), Whiskers / 95% CI, FDR-significant, Walk-forward confirmed

The table of numbers behind the day-of-week chart: for each day n (how many days entered the mean), the mean change %, the 95% confidence interval, the share of up days and the quality badges. The same daily mean as the chart, in full numeric form.

n (sample size), Whiskers / 95% CI, % up days, FDR-significant, Stable over time (walk-forward)

The turn-of-month effect: the mean change over the last and first days of the month (the turn window) versus all remaining days, with n, the 95% confidence interval, the share of up days and the quality badges.

n (sample size), Whiskers / 95% CI, FDR-significant, Stable over time (walk-forward)

A comparison of the weekend (Saturday+Sunday) with weekdays: mean change, 95% confidence interval, share of up days and badges for both groups, computed from the same daily series as the other models.

n (sample size), Whiskers / 95% CI, FDR-significant, Stable over time (walk-forward)

Relative trading volume by day of week: each day's average volume expressed as an index where 100 = the whole-period average. An index of 115 means a day historically 15% more active than average, while 85 means 15% quieter. A measure of market activity, not of price direction.

Volume index (100 = average), n (sample size), FDR-significant

The averaged week path: a line accumulating from Monday, where each point is the sum of the mean daily changes up to and including that day. The shaded band around the line is the 95% confidence interval. The model uses the same per-weekday means as the bar chart: the two always agree.

Cumulative (%), Whiskers / 95% CI, n (sample size), FDR-significant

Other modules