To be honest, I was surprised by these results myself. I recently saw this link which indicated that there is a "best day of the week to buy bitcoin":
https://steemit.com/bitcoin/@datageek/best-day-of-the-week-to-buy-bitcoin-data-analysis
The message of that post: buy BTC on Sundays. Obviously, this can't be that easy? If something has been known for years, it should be taken into account in future price developments. I decided to apply a statistically rigorous method rather than simply eye-balling the averages.
Best time to buy
First, we start with data going back to 2013 and generate a Bayesian generalized linear multivariate model which take into account the weekday data. We modelled the daily low which can be seen here:
It doesn't look like much, which shouldn't surprise us. If there were very obvious differences everyone and their mother would probably buy on that specific weekday. However, when I do a statistical comparison to a simpler model with no knowledge of the date. The complex model performs 33.5 SDs better which is remarkably strong (5 SDs indicates that you can be 99% sure that the complex model is better). This means that there is some signal in knowing the weekday if you want to predict the lowest price. According to the model, the lowest price can be found on Fridays.
As this takes into account many years of data, I tried to do the same calculation for just the last year. Here is the modelled data:
Again a formal model comparison showed that the complex model is 43.03 SDs better. Again we found that the best time to buy is Fridays.
Best time for selling:
We again start with data going back to 2013 modelling the daily high this time. Here is the modelled data:
Again, it doesn't look like much, but our model is once again better than a simpler one with 32.56 SDs. The best date to sell is Thursdays
I tried the same calculation for the last year:
Again the complex model is much better with 43 SDs. This time, however, it seems that Mondays are the best day to sell.
You can probably not go out and become a millionaire with this knowledge but there is a difference amongst days that is big enough to be detected with modern statistical tools. Maybe I will change my DCA to buying on Fridays just for fun. The difference within days, however, is bigger than the difference between days as you can see on the images.
As always, questions and suggestions are more than welcome. Good luck to all of you and take care.
If you found the above useful and wish to support me:
BTC: 1DxfCPqQQqaFuqjfdFFa4QayDskaEPmQKF
ETH: 0xF5027c04FD54a41d28Cd25a38d26883F317E26d8
LTC: Lby8oYdYhFy3uqqBHYFCpmZrcmhBQUJL4o