There is a popular belief in markets - " Sell in May and go away" - It seems like there are some more like this and I realised a new one today when I was discussing this with my friends. So, thought to share this cool data with you all.
Year
June Lows
July Highs
% return
2006
2595.65
3208.85
23.62%
2007
4100.80
4647.95
13.34%
2008
3790.2
4649.85
22.63%
2009
4143.25
4669.75
12.07%
2010
4961.05
5477.5
10.41%
2011
5195.9
5740.40
10.47%
2012
4770.35
5348.55
12.12%
2013
5566.25
6093.35
9.47%
2014
7239.50
7840.95
8.3%
2015
7940.1
8654.75
8.99%
2016
7927.05
8674
9.42%
2017
9450
10115
7.04%
2018
10550
11335
7%
2019
11625
11946.75
2.76%
2020
9544
11300
18.39%
2021
15450
15924
3.06%
If we look at the historical data of june and july from last 16 years, We can find some interesting observations :
Some Findings
Average return from June lows to July highs is 11.19%
In last 16 years, 8 years (50% have given double digit returns)
6 years (37.5% ) have given returns ranging from 7 to 9%, So, we have almost got close to 7-20% in 8.7 times out of 10 in last 16 years.
Only 2 years, the difference btwn highs and lows is 3% or below but those 2 years have come in the last 3 years.
Disclaimer and some other points :
This data may not mean much as there is no guarantee that past returns can be replicated. I have only checked june to july data. It needs to be seen what returns we have got if we compare 2 monthly returns for rest of the months (maybe we can do that in this thread for reference in the future)
Maybe i misunderstood, but this seems to be forward looking test. No one will tell you what is June low and what is July high. You could do that for any two consecutive months and probably you will see better than avg returns.
Now if this was buy June1 / June 30, Sell July X then maybe you can have something ( even that i doubt as seasonality has very low sample size).
What this does give you is range of the two months, so maybe you can do that for multiple pairs and see what is likely to be more volatile or less volatile. So maybe mid year is less volatile so range is lower. I have heard that march can be trending due to eoy flows, same with dec, so maybe those periods are more volatile.
This needs to be done. Will try to do it and share it in this thread. will be interesting to see how the bi-monthly returns are for rest of the year.
Yeah…You are correct. This data would make better sense at the end of june. :
At the end of the day, TA or even data analysis is a matter of probability. The whole task is to place ourselves in high probability low risk high reward setups.
Yes but this is a mistake - TA or data analysis. This is not predictable as we are choosing lows and highs in hindsight. There is 0 predictive value from results derived by looking at the future. This kind of mistake happens to all and sometimes it can be devious. There might be other inferences possible looking at the range, but not in terms of returns as we cannot execute without a time machine.
Ways to mitigate
While these results are ok, sometimes you might get something that seems too good to be true. In that case - it probably is, so must have some doubt and take extra effort in verifying. Has happened to me a few times.
There is no edge in randomness. One nice way of testing for future data poisoning results is to run same test over randomized market data ( Just reorder % changes from actual market data). If your theory has an edge in random data - then you have likely made a mistake.
The intention is to keep an eye on june lows and see how the trade pans out from those levels. Buy nearing lows and selling near highs is something only Gods and twitter traders can do
Rest of your points are absolutely spot on and are useful for anybody who wants to do some data crunching…
Lowest of one month and highest of next month. What genius! Definitely useful for time travelers Idea should go into the next edition of “Backtesting for Time Travellers”
Arre sir. Just testing waters with this type of data. It is just to get a broad picture as to how much our markets have bounced from lows in june in general historically.
Please understand the intention is not to buy low and sell at the top doing this analysis…I just wanted to check how the returns from a month’s low to next month’s highs are …I will check the data for rest of the year and share the findings here.
Nifty up from 15191 to 16550++ in July (9%++ returns)
As you all rightly mentioned, if we look at lows of previous months and compare with highs of next months, the returns are mostly going to look good)
I didnt yet check the average returns data for rest of the months…but july being a bullish month in general historically…I tried this unusual experiment.
This data would make better sense at the end of june. :
Idea should go into the next edition of “Backtesting for Time Travellers” 
