Vote Distribution and Voter Behaviour on ICON Network (ICX)
We hope you enjoyed the video. We also hope that you felt somewhat alarmed by the skewness of the vote distribution. It is important for the success of the ecosystem, and by extension, your investment, that contribution is rewarded, and doing nothing is not. So we would like to encourage everyone to make informed decisions when they vote. This will not only give funding to the best teams, but also increase the security of the network by furthering decentralisation.
The ICON Network has been decentralised since August 2019. The number of Public Representatives (P-Reps) have increased since then, and there are 126 P-Reps as of today (May 31, 2020). We aimed to look at the distribution of the votes to identify the behaviour of voters and get a glimpse of our progress towards the optimal decentralisation.
P-Rep distribution by country
First, we looked at the number of P-Reps by the countries they reside in. Note that these are self-reported. It was clear that the USA had the largest number of P-Reps (25%), followed by South Korea (10%) and the UK (8%).
When grouped by sub-region, Northern America had the highest number of P-Reps (N = 38; 30%), followed by East Asia (N = 24; 19%), Northern Europe (N = 15; 12%), Western Europe (N = 14; 11%) and South-eastern Asia (N = 11, 9%). Not surprisingly, the votes were concentrated heavily in the USA.
* ‘N’ represents number of P-Reps
P-Rep vote distribution
Next graph must look very familiar to Iconists. This is the current state of votes for each P-Rep. We cannot stress enough that the vote distribution is heavily skewed. ICON Foundation alone has more than 15% of all votes. Nearly half of the total number of votes (49%) in ICON ecosystem are delegated to Top 10 P-Reps.
To gauge the effects of potential large vote holders (so called ‘whales’), we removed outliers (3 standard deviations away from the mean using z-scores — if delegation per voter is 3 times larger than the average of total votes per P-Rep). These outliers made up nearly half of all the votes (46%; coloured bars).
We also looked at the vote distribution using the number of voters. The rankings for the number of voters were slightly different from those by votes, however, it was evident that top 5–6 P-Reps were still very popular among the voters.
In case we did not capture large vote holders for those P-Reps that may have been made up of ‘whales’, we looked at the ratio of the votes (total votes / total voters) to see what the average number of votes were per voter in each P-Rep. Large vote holders were more prevalent within Sub P-Reps, with the largest vote-per-voter ratios found in Hourglass, followed by Orangebloc, and Cardinal with almost no outliers! Bara-node was an interesting one, it had a number of small vote holders (the tiny red bar), and one large vote holder (which is considered an outlier — a potential ‘whale’).
Look! It looks like we, Transcranial Solutions, also have a whale voting for us! The average votes per voter for us were around 40K, but after removing one outlier, it reduced to around 10K. It turned out that the ‘whale’ was the Pilot Delegation Program, much like what you see in ICON Guide Star who received the delegation by this initiative.
Trends in vote behaviour
We then focussed on the behaviour of the voters. We could see that the number of voters participating in the P-Rep voting has been increasing every month. There was also a strong correlation (r = 0.95, p < 0.001) between the number of votes delegated and the number of voters over time, indicating that the growth has been dependent on each other, and the same voters re-staking were not the only reason for the growth.
We next explored the spread of vote allocations — how many P-Reps does one voter delegate their votes to on average. It was evident that more voters have been voting for only one P-Rep to date, compared to the number of other allocations.
Has it always been this way? We decided to look at the historical data since the decentralisation. The results showed that the median (midpoint of a frequency distribution; white dot in the plot) has decreased since February 2020, from 4 P-Reps-per-voter to 3 P-Reps-per-voter. The density widths were widest at 1 (a greater number of voters voting for only 1 P-Rep), and it widened further gradually. The result suggest that voters delegating to only 1 P-Rep has increased over time, and it makes up for the largest proportion of all voters. It was interesting to see that voters voting for 10 P-Reps have also increased, and the vote allocations appear to have become more and more bimodal (1 or 10), but leaning towards 1. We need to shift this trend so that the top portion of the violin gets ‘fatter’ while bottom part ‘thinner’.
Who do these 1-P-Rep voters vote for? More than 25% of them were voting for ICON Foundation, and most were found in Top 22. It was interesting to see ICX Australia (current ranking #56) amongst them. The reason is unclear at this stage, but perhaps our next report on the analysis of vote stagnancy might shed some light into this.
How has this changed over time? While the proportion of ‘Other’ P-Reps have remained rather steady, there has been an increase in the number of 1-P-Rep voters delegating to ICON Foundation. We could also see a decrease in the proportions for other larger P-Reps such as Ubik Capital and VELIC.
The results suggest that new voters may be more inclined to vote for ICON Foundation. It is great that we are getting new voters to ICON Network, but it is just as important to educate them to spread their votes.
Here is the table for the above data for the reference.
Summary
The votes are heavily skewed towards Top-10 P-Reps, with nearly half of all the votes delegated to them. ICON Foundation has been getting the most votes and the rate is increasing. Although we have more voters than before, there are more voters voting for only one P-Rep than any other number of allocations, and as many as 25% of them were delegating to ICON Foundation. Good news is that 10-P-Rep voters are also increasing, but we think we could do better. Although many Iconists tend to scrutinise P-Reps to ensure that a positive ecosystem is built on the ICON Network, all of us have a role to play. After all, this is a decentralised project and we have the power (votes) to change it for the better tomorrow.
What’s next?
There were many questions that arose during the exploratory analysis of the data. Vote stagnancy is one example and we will be diving into this in the next article. Stay tuned!
Acknowledgement
We would like to thank Insight for their help in setting up our node. We would not be a P-Rep without their help. Please consider shifting some of your votes to Insight so that they could be at top 22, they deserve it!
Also, we used Everstake’s Vote Monitor website (https://iconvotemonitor.com/) to gather the data via webscraping. It is a great data source, please have a look if you haven’t. They were also very helpful with any queries we had. Thank you.
Resources
The code used for webscraping, data manipulation and calculation, analysis, and visualisation can be found in our GitHub Repository. It also contains the data we used, as well as other images (P-Rep logos). Please note that some P-Reps did not have any logos, and we designated what we thought were best suited. We will monitor the logos and update them periodically.
