Ether Historical Data Brief

Hashrate

While refreshing some old charts with new data, I recreated the following plot.

The relationship between Block Size and Hashrate appears to forming a normal distribution, or inverting at a minimum. This is different than what I have observed in the past. In a previous chart, I pulled data up until 11/09/2017 and noted the following trends:

The chart on the left shows the full history up until November and the right one a redacted history excluding data points when the Block Sizes being mined were relatively undeveloped. Both indicate that a linear relationship was developing at the time.

Based on the new behavior I see forming, I posed questions to others and was pointed to the relationship of Hashrate to Price. By changing the Block Size to Price, I observed a similar dynamic of a distribution that was beginning to invert.

A normal distribution has not developed yet, nor would it make intuitive sense unless there is a reflection of a significantly different technology running the Blockchain. To investigate this possibility, replotting this data into time series help identify if any patterns are developing at certain dates.

The chart above maps the two variables Hashrate and Price over time. It is not a perfect relationship, nor would I expect that one variable drive the other. However, the suggested pairing of these two points to possible divergences based on the summer of 2017 and start of 2018.

Substituting the original hypothesis of Hashrate and Block Size, the time series also appears to show some interesting behavior at the start of 2018. Also, there was a significant drop in Block Size in the 3rd Quarter of 2017.

As an aside, the chart below also shows that there appears to be more effort needed to meet consensus around the same time in 2017.

It is not clear to me why the Block Size would be significantly different towards the end of 2018 (I put forth hypothesis on experiments with new mining hardware, possible effects of Cryptokitties at end of 11/2018, etc). It will be a topic matter that I’ll continue to ask about.

One relationship that I was told to monitor at the same time is that of Transactions to Block Size. This relationship seems to be fairly straightforward at present. However, this is subject to change with the new consensus methodology and Proof of Work experiments. I’ll see if the data corroborates any of these theories with time.

Wrapping up these individual correlations, it serves as reference to see the actual numbers and rough shapes as correlations amongst all of these variables. I focused on the stronger correlations shapes over time, but certainly how relative the values are amongst others are important as well.

Day of the Week Evidence

One of the more interesting relationships the evidence supports is how there is different behavior for transacting on specific days of the week.

I’ve been asked about this observation before and I have no definitive advice on when to trade. I’ve made previous record of the amount of transactions on each day and this could be helpful for miners. However, to develop any Blockchain intuition as far as price, a box plot for Price Change over the course of the week may be helpful.

While this chart above is helpful, cutting off 1% for each tail focuses the frame to remove those particularly large data points (while still keeping the quartile calculations and mean intact).

This chart draws attention to the days of Tuesday to Thursday having potentially higher values given 75% percentile rates are highest for these days. However, it is hard to look at the actual mean Price Change. Producing a quick table can get us easy to view numbers.

##   Day Price_Change.mean Price_Change.sd
## 1 Fri      0.0072611216      0.07588645
## 2 Mon     -0.0003232503      0.11600210
## 3 Sat      0.0023863822      0.09109366
## 4 Sun      0.0093164065      0.06763465
## 5 Thu      0.0183920385      0.09383893
## 6 Tue      0.0101873341      0.08143783
## 7 Wed      0.0100742838      0.08847924

What this output shows is that Tuesday - Thursday do produce the highest mean price change. However, the mean and standard deviation for Monday is skewed. This is probably because it shows an anomaly in the data which included data for the date of 08/11/2015 in which the price of Ethereum went back to 0 in the first 2 weeks thus introducing a -100% and throwing off the scale of other means. If we subset the data to exclude those first 2 weeks, we can show another set of statistics with more accurate data.

##   Day Price_Change.mean Price_Change.sd
## 1 Fri       0.007369497      0.07644963
## 2 Mon       0.007079368      0.07838533
## 3 Sat       0.006899612      0.07532550
## 4 Sun       0.009455457      0.06813157
## 5 Thu       0.018664513      0.09450953
## 6 Tue       0.010262241      0.08173415
## 7 Wed       0.010148359      0.08880207

The table above cleans up the data and gives a numerical output for each day. While the visual Box Plot helps guide the user to possible trends, the numbers substantiate the hypothesis on a more specific basis. Accordingly, I put forth the following takeaways:

• price change on Tuesday, Wednesday, Thursday are on average higher than the rest of days of the week
• price volatility on Tuesday, Wednesday, Thursday are also higher as measured by standard deviation
• the volatility as measured in standard deviation is lower for the weekends than the weekdays
  • returning to a past observation, this corroborates that the weekends tend to have less volatility in prices

This days of the week analysis has continued to be a cute observation on when people tend to write on the blockchain and whether it has effect on Ether. I imagine that with time, this effect will wash away, but it has been unique. In most global finance markets, there is not an active marketplace on weekends so watching Ethereum, which is 24/7, continues to be novel.

References

Original Dataset - https://www.kaggle.com/kingburrito666/ethereum-historical-data
Current repository of dataset - https://etherscan.io/charts
Necessary primer to understanding Cryptoassets - https://www.amazon.com/Cryptoassets-Innovative-Investors-Bitcoin-Beyond/dp/1260026671
Glossary - http://ethdocs.org/en/latest/glossary.html
More Technical Glossary - https://bitsonblocks.net/2016/10/02/a-gentle-introduction-to-ethereum/
Price Reference- https://coinmarketcap.com/currencies/ethereum/

I foresee updating this analysis every month or so. I am curious to see if these patterns, behavior and relationships hold over time as this data set continues on. Reach out to me at @RichLopezNY or richardjameslopez[at]gmail if you have any comments or ideas on what I may be missing.

Full Disclosure - this analysis is not meant to be used for investment recommendations in any way. This is simply exploring data.

Donations kindly accepted at: 0x76276793c2A8dEB6F51eb75C43a9aDfCBb78C195