Hourly Active Addresses - Zooming In to Optimize Intraday Trading
Many of you may who have been following our analyses and discoveries revolving around active addresses and their correlation to price fluctuations these past few months may have already refined your mid and long-term trading strategies based on our models and previous insights like these. It is quite clear that the amount of active addresses occurring on a network on a daily basis have a long-term effect on a crypto asset's ability to thrive or dive. The amount of unique addresses moving tokens, as well as the amount of unique tokens moving themselves, are arguably the two most important on-chain factors for price movement, period.
This being said, we're well aware of the bullish and bearish divergences that form when an uptick or downswing in the amount of addresses interacting occur on a day-to-day basis. Now that Santiment has hourly and even minute-by-minute address activity to go along with the daily variety, it was only a matter of time that we listened to the suggestions and researched whether this same daily phenomenon can show signs of promise by comparing how address activity fluctuates for Bitcoin on an hour to hour comparison.
The first thing we need to do is download the freely available .csv data that can be pulled any time for any asset for any time interval as long as you're a Sanbase PRO user.
In this case, I wanted to pull the hourly active addresses tracked for a three-month time period. We could use a larger sample size such as the previous year, or even previous five years, but I wanted to start with a relatively recent sample that didn't include any major BTC events like March's Black Thursday or May's lackluster halvening. With this data now in numeric rows and columns, we'll need to organize all hourly data into common hours. Dates don't matter. We simply want ~90 days worth of the fluctuations that occurred between 12:00am to 1:00am UTC, 1:00am to 2:00am UTC, etc. After some sorting and formula dragging, we end up with a table that looks like this:
As many would suspect, as we can see in the middle column, there is a distinct drop-off in unique from 5:00am UTC all the way until 2:00pm UTC. This coincides with the most common sleep hours for the United States, and other metrics like trading and on-chain volume follow the same sensible time-based peak and valley fluctuation pattern.
Now with this table created, we can look at this same data in a more visually, easy-to-digest manner with a simple two-axis chart:
From July 4, 2020 to October 4, 2020, the pattern really shows itself through this chart. The address activity on Bitcoin peaks around 9:00pm to 10:00pm UTC before declining just slightly upon market close from 11:00pm to 12:00am. The lowest hourly address activity occurs between 12:00pm to 1:00am UTC (3:00am to 4:00am for those on the United States west coast).
Price fluctuations do see an interesting trend, but it's not really the notable correlation to hourly address activity that we would really be hoping for. That being said, it's not absent of any correlations whatsoever, either. If we simply split the first and second halves of each trading day, there is a clear better average performance during the second half. From 12:00am to 12:00pm UTC, the average price return is -0.006%. But from 12:00pm to 12:00am, the average return is +0.023%.
The amount of average address activity between the average market day's first half (967,184 per hour) and second half (970,942) is nearly a wash, but it does favor the latter by a very subtle 0.39%. The two best performing hours over this three month data set does indicate that the early afternoon hours (UTC) have two respective great performing hours (+0.1% per hour at 1:00pm to 2:00pm and 3:00pm to 4:00pm) just as address activity begins to increase from its low point of its 24-hour cycle.
So what conclusions can we draw from our first real deep dive into hourly address activity correlations? Well, it does appear that there is an average mini-surge in price that could be taken advantage of via buying in the early to mid-morning US hours. And the final hour of Bitcoin's 24-hour market cycle being the 2nd worst performing hour on average is also quite fascinating. But this final hour is also the 5th highest amount of address activity, making it hard to really make a connection to active addresses transacting on the network.
Since we aren't just satisfied with very, very minor correlations, we'll definitely be releasing a 'Part 2' to this study where we will have a much larger sample size to work with after this 24-hour address activity metric being integrated with our highly popular Sansheets tool. We'll also be looking into how these hourly fluctuations differ on the Ethereum network, and perhaps a few other ERC-20 tokens. Stay tuned, as we figure out how to optimize intraday trading using one of the most reliable leading indicators available on Santiment.