The Gain Per Trading Day And Day Of The Week Since 1970 in S&P 500
Last Updated on June 19, 2022
Do you know the average return in the S&P 500 separated into trading days and day of the week?
In this article, we look at the average gain per trading day and day of the week. The best days are centered around the first and last days of the month. We also find out that the best trading days are Tuesdays and Wednesdays, something which might explain the “Monday reversal”.
Let’s start by measuring the trading day of the month:
The average gain per trading day of the month
Below you can find a bar showing the average return per trading day in the S&P 500 since 1970, excluding dividends.
Day 1 is simply the first trading day, not necessarily the 1st day of the calendar month.
Obviously, the months have a different number of trading days. The numbers are not compressed or prolonged to adjust for that. Sometimes day 20 is the last trading day of the month, in another month it could be the 21st.
The average gain per day of week
Looking at weekdays we get this bar chart (1 is Mondays and the gain is from the close on Friday until the close on Monday):
Â
Fine overview. Randomly we just happen to do some similar studies at an empirical finance class at NHH; and adjusted for risk (volatility, new month, holding-time over weekend etc), the differences did not pass any hypothesis tests, and the few patterns found significant in the past all disappeared in newer times (as the effects were shown to have become well-known). So I guess the search for findings must be increasingly inventive going forward, and probably held close to the chest for the hopeful trader.
Interesting. But personally I think the reason for all these market cycles is more to randomness than trader patterns.
Yes. A while back I found an analysis (can’t find it now, unfortunately), which showed that in fact it was “worse” than randomness(!). Meaning that patterns existed for such a long time that the underlying process certainly could not be random, but then suddenly and deterministic changed to a new unpredictable state(call it a new market cycle). The study used chaos theory and Lorenz curves to show these transitions; in a system which in one dimension appear random, but in another dimension deterministic governed by some (viewed from the prior state) exogenous law. Thus, one could conclude that it is (at best/least) a combination of randomness and trader behavior.
My experience with patterns is this: I start searching, find a pattern, paper trade it for some time to test it, and as soon as I start trading it, it no longer works. Either I find randomness (there will always be patterns), or too many traders see the same 🙂
Nice post and interesting comments. IMO there are two kinds of edges – short term and long term. By short term I mean edges that are fleeting i.e., work for some time and go away as more people notice or with changes in market. I think most calendar based and many other usual quant strategies come into this category. Plus side, they are interesting and while they work, provides great returns. On other hand, for example Dow theory is a long term edge that manifests even on intraday charts. But that old rag somehow manages to hide from most while still being open in front.