Calendar patterns that held up after testing · as of 2026-10-23
as of 2026-11-20
28,321calendar patterns tested
4,946looked significant, chance alone gives about 1,416
1,135held up after correcting for chance
114worth reading: the market’s 26, and 88 a sector or stock has that the market does not
A pattern here is a kind of day, a weekday, a month, the turn of the month, whose record beat every other day of the history by more than chance allows, after correcting for how many patterns were tried. It describes the past; nothing here says what a coming day will do. How they are tested →
Patterns that held up
47 shown · sorted by strength of evidence
The marketHigher average
Sessions falling on the 1st of a month
Date of the month · median-stock series
54.9%of these days up
40.3%of other days up
+0.42 ptsaverage day vs the rest
The marketLower average
Every Monday
Day of the week · median-stock series
42.5%of these days up
40.1%of other days up
−0.15 ptsaverage day vs the rest
The marketLower average
Every Tuesday
Day of the week · median-stock series
34.6%of these days up
42.0%of other days up
−0.15 ptsaverage day vs the rest
The marketHigher average
Every Wednesday
Day of the week · median-stock series
49.1%of these days up
38.4%of other days up
+0.35 ptsaverage day vs the rest
The marketLower average
Every 4th Monday of a month
Which weekday of the month · median-stock series
38.8%of these days up
40.8%of other days up
−0.29 ptsaverage day vs the rest
The marketHigher average
Every 1st Wednesday of a month
Which weekday of the month · median-stock series
55.0%of these days up
40.1%of other days up
+0.40 ptsaverage day vs the rest
The marketHigher average
Every 5th Wednesday of a month
Which weekday of the month · median-stock series
60.0%of these days up
40.4%of other days up
+0.50 ptsaverage day vs the rest
The marketLower average
Sessions in Q1
Quarter · median-stock series
12.5%of these quarters up
19.8%of other quarters up
−8.31 ptsaverage quarter vs the rest
The marketHigher average
The last two sessions of a month and the first three of the next
Turn of the month · median-stock series
47.6%of these days up
38.6%of other days up
+0.20 ptsaverage day vs the rest
The marketHigher average
Sessions in the 1st week of a month
Week of the month · median-stock series
46.4%of these days up
39.1%of other days up
+0.18 ptsaverage day vs the rest
The marketHigher average
Every 3rd Wednesday of a month
Which weekday of the month · median-stock series
48.2%of these days up
40.4%of other days up
+0.28 ptsaverage day vs the rest
The marketHigher average
April
Month of the year · median-stock series
68.8%of these months up
25.4%of other months up
+4.48 ptsaverage month vs the rest
Check a pattern
measured on the daily returns, against every other day
Describe some days, the 2nd Wednesday of August, in expiry week - and see what they did. Several values in one box mean any of them; different boxes must all hold, and every box you add shrinks the sample, which the answer prints first.
or start from a date:
An answer here is uncorrected: the patterns above were fixed before testing, while a combination you build is a search, and about one in twenty clears p < 0.05 on noise alone. The tool counts your tries and says so.
What is coming upmeasured on the returns
One row per session ahead, and one name on it: the single stock, sector or index with the strongest record for what that day IS, a 1st Monday, an expiry Thursday, a turn-of-month session. Chosen on how far its up-rate sits from its own record on every other day, never on the raw up-rate, and only from findings that survived correction and beat the index baseline. A session with nothing beside it had no survivor, which is a result. These days have not happened; nothing here is a forecast about the date.
Historical behaviourwin rate and average move, per bucket, against the rest of the sample
Pick one instrument and read DOWN it: every dimension, every bucket, in calendar order, in one table, including the buckets F.36 described and never tested, which for most stocks is most of them. Each rate sits beside the same rate over every session outside the bucket, because that comparison is the only thing that makes it mean anything. Buckets are never sorted by result: ranking a dimension by win rate puts the luckiest sample on top, which is the mechanism that makes seasonality tables look convincing.