DFR2 – Voting analysis
With the results in (read more at https://deepfunding.ai/deep-funding-round-2-results/), it is time for a deeper dive into the results. The goal of this analysis is to provide transparency and to gain insights that help us to improve and prioritize new features.
Within the scope of this article, we are limiting ourselves to the Deep Funding Vote and are not doing a separate analysis of the Loyalty award vote that happened at the same time.
Changes in the voting process
Before we dive into the details, it is good to recollect the changes we implemented in the voting process since the previous Deep Funding round:
- Quadratic voting: We introduced Quadratic voting, meaning instead of one token one vote, we calculated the square root of each person’s token balance, as an equalizing measure. It still gives you substantial extra weight to have more tokens but this measure ensures a more balanced weight between small and (very) large token holders.
- Reputation rating: We introduced another weighting mechanism, based on someone’s reputation score that was generated by being active and constructive on our proposal portal. We first experimented with this in our first Governance round and this gave us sufficient confidence to roll it out in an actual Deep Funding round. Besides giving active members more impact, it also serves to protect the process against strategic ‘wallet splitting’. Without reputation weights, it will be advantageous to split your wallet into multiple smaller ones to reduce the effect of the Quadratic balancing. It will however be hard to build a good reputation over all these smaller identities.
- Multichain voting: For the first time we enabled voting with both AGIX on ADA and AGIX on ETH.
- Wallet linking tool: We offered the users an option to link multiple wallets to their IDs. This enables the users to apply all their AGIX (on multiple chains and multiple wallets per chain) Without the hassle of going through the voting process twice or having to transfer their tokens to a single wallet.
- One signature per pool: In order to make the process easier, the users could cast a vote for all projects in a pool, and sign once to submit them. This is significantly easier than having to sign for each separate proposal vote.
There were also process changes that may have impacted the overall results such as peer reviews and the eligibility process, as well as some website changes, but although interesting in themselves, we’ll leave them out of this analysis.
Statistics Round 1 vs Round 2
Before going to the insights and conclusions let’s have a look at some numbers: and compare them to round 1 where possible. (limitations: there were fewer votes in round 2 because we signed per pool and not per project. There may have been people in round one that voted with multiple wallets whereas in round 2 they might have linked their wallets and voted only once, representing multiple wallets.
Round 1
Round 2
Total amount available
1 Million USD in AGIX
0.5 Million USD in AGIX
Total amount awarded
$ 741.000
$ 449.212
Nr. of pools
2
5
Nr. of submissions
47
89
Nr. of eligible projects
28
57
Nr. of awarded projects
12
17
Nr of wallets that voted
158
187
Nr of collections
n/a
144
Summarizing this data we see that while halving the total amount to be awarded, we doubled the number of submitted proposals!
At the same time the number of wallets that voted has gone up as well, but offset against the potential inflow of new Cardano wallets this is a relatively modest growth.
Collections and AGIX balances
If we compare the balances before Quadratic voting this is the result.
There is clearly one gigantic whale at play. If we take out the one whale the result looks like this:
If we do the same with the quadratic voting taken into account what we see is this:
Including whale:
Excluding whale
We could conclude that this graphic, including quadratic voting and excluding the whale, shows a reasonably healthy balance of weights.
The whale
When we zoom in on the behavior of the whale (Btw for those that suspect that this whale is somehow related to the Foundation; this is NOT the case.) We say that its voting behavior does not really differentiate substantially from the average. We can see a few projects that would have benefitted if we had not taken quadratic voting into account, but there is no clear pointing finger here. However, if we take out the votes from the whale completely there is one significant difference. Overall the outcomes would have been almost the same, except for just a few changes. However, there is a large difference in pool C. Without the whale not just one but 5 projects would have been awarded in Pool C! This shows that we have to be careful about assigning conclusions to the outcomes of the vote alone.
There are (only) a few other wallets displaying the same behavior, in pool C, but the impact of the whale on the outcome in this respect is significant.
Voting graphs
We can get a different perspective on the dynamics of voting by reviewing the infographics made by Robert Haas with his open-source graph visualization library gravis.
In each graphic at the top row you see the wallets/collection IDs and on the bottom row the proposals. Green proposals are awarded.
– The thick red lines represent ‘1’ votes.
– The orange lines represent ‘2-6’ votes.
– The light green lines represent ‘7-9’ votes.
– The thick dark green lines represent ’10’ votes.
The collections that have a reputation weight are pink, and the size of the circle represents the final voting power of the collection.
(The order in the row of wallets is determined by the plain token balance)
Pool C: Projects in the ideation phase
Main insights from these graphics:
We see an overall mixed distribution of red and green lines coming from the wallets. We interpret this as a healthy indicator of authentic voting behavior.
We see significantly fewer extreme votes (ones and tens) compared to round 1. We made a point beforehand that this behavior was not desired and this seems to have paid off.
Most collections are voting for a large number of projects, with a distribution of votes. We can take this as an indicator of clear engagement from the community.
Perhaps most significant is the low density of ‘pink dots’ this shows there are relatively few collections with a reputation score. Zooming into this we see:
- We have 276 users in our export from the proposal portal (Swae)
- 71 have a reputation score > 0
- But out of these 71, only 6 voted, or at least in a recognizable way.
This is rather disappointing and asks for deeper investigation. We have not yet found a good reason for this and some sample tests of known users showed no strange deviations.
Proposal portal engagement
We received an export from Swae containing 276 users. From these users, the majority was just incidental, but 71 had a reputation score that is higher than 1, based on these predefined variables:
Variables scoring
Comments and feedback created
3
Points per comment created
0
Points per reaction created (thumb up or down)
0
Point per emoji reaction given (positive or negative)
Comments and feedback received
2
Points per positive comment reaction received (thumbs up)
-3
Points per negative comment reaction received (thumbs down)
2
Points per positive reaction received (Genius/Happy/Hot/Clap /laugh/Love)
-2
Points per negative reaction received (sad, angry)
We calculated the scores and distributed the users in different tiers like this:
Voting weights
Percentile
Tier
Weight factor
96-100
A
5
81-95
B
3.5
61-80
C
2.5
31-60
D
2
11-30
E
1.5
<0-10
F
1
no points
G
1
Loyalty rewards
Back in December when we started round 2 we allocated 100.000 AGIX as a reward for engagement. The total Dollar value of the rewards at that time was something like $5,400,-
At the moment of writing this article, this value has increased to $44,000, which is very substantial! This amount will be distributed over 81 wallets, making some of the rewards a bit more substantial than we anticipated. We will however not change this situation, since this is what was agreed upon beforehand. For the next round, we may have a different way of calculating the awards, in order to avoid large swings in dollar value.
Some statistics on the Reputation rewards:
The bulk of the rewards is allocated to the constructive activity. Only 5% of the total rewards are allocated to IDs in the proposal portal that also voted.
Total awarded amount for reputation:
100,000
Reserved for voting by contributors
5,000
95,000 AGIX
Percentile
Nr of IDs per tier
Loyalty rewards p.p.
Relative rewards p.p.
Tier
D
15%
96-100
4
3,563
0.41
Tier
D
25%
81-95
12
1,979
0.23
Tier
D
25%
61-80
17
1,397
0.17
Tier
D
25%
31-60
22
1,080
0.11
Tier
D
10%
11-30
16
594
0.07
Tier
D
0%
<0-10
13
0
0.00
Tier
D
0% (no activity)
0
On top of these amounts, we have allocated 312.5 AGIX for those that voted (16 accounts, but only 6 of them had a reputation score > 0)
Conclusions and improvement points
Improvement areas
We have onboarded on this journey with a learning mindset. Be agile and make changes, and keep what works. Sometimes we are implementing processes as an experiment without yet having optimal tooling for this. We think this is the right way of doing things, for speed, out of a sense of urgency, and also to ensure that any tooling decisions and requirements are grounded in reality. The inherent risk is that this experimental approach comes with manual tasks and thereby some risks of errors. Unfortunately, this did materialize in 2 instances: We manually migrated the proposal content from the portal to our regular deep funding site for future reference and administration purposes. The downside was that we didn’t spot a few last-minute changes to some of the requested amounts. Another error was made in calculating some of the reputation rewards. Some users have received an amount that was too large due to an incorrect variable in a very large and complex spreadsheet. (We have contacted the related people and friendly asked the to return the additional amount).
Also without these errors, it is clear that the process is becoming more and more complex and harder to maintain, especially with the very few people assigned to these tasks. Therefore one of our main objectives is to create or enhance features that will enable automation, reliability, and transparency.
Another improvement point is related to the number of engaged people on the portal that can be linked to voting activity. For future health of the system and high-quality outcomes, we will continue to motivate both portal contribution and voting behavior. By better integration of our wallet linking tool, we hope to be able to make more connections between these two activities.
Finally, a remaining question is on the topic of transparency. Until now we have tried to be very transparent in explaining the results and the processes and decisions leading to these results. Where we have drawn the line, up until now, is releasing information that is potentially traceable to specific identities. Going forward this is a topic to review.
- Should we expose all wallet addresses and voting behavior?
- Should we show the reputation rating for each profile on the portal
- Should we even show the wallets associated with profiles on the portal?
All this will lead to greater transparency, which is positive in itself. A potential risk could be, however, that some individuals that value optimal privacy will prefer not to participate anymore. A classic case of privacy vs. transparency which are both important values in the blockchain work (and beyond). Should we go for full transparency, we will give any of the current users the opportunity to opt-out, be forgotten, or change their public ID. This topic could be interesting to explore in more depth in the next, upcoming, governance round.
General results
Overall we are looking back to a very successful round 2!
We are extremely happy with the number of submissions, and the engagement around them. Some processes and tools need to mature, but we are confident that we are on the right track and are looking forward to pushing these further. While we strongly believe in quality over quantity, there is room for further growth in the number of engaged community members and voters. Larger numbers will make the system more resilient and harder to game, providing we also have the right processes in place.
All these changes go hand in hand (processes, tools, and number of contributors). And they will need time to mature and come to full blossom. But to be in this place already, after only 2 rounds, and with many known improvements on the radar, we see a VERY bright future ahead for Deep Funding.
Therefore our gratitude goes to all the teams that submitted a proposal, whether awarded or not, to the awarded teams and community members that are taking charge of some special tasks, and last but not least to all the community members that were active on the Proposal portal and that voted for one or more of these great projects. We could not have done it without you all, and we are looking forward to many more, increasingly successful Deep Funding rounds!
Thank you all and see you in Governance Round 2 and subsequently in Deep Funding Round 3!
With the results in (read more at https://deepfunding.ai/deep-funding-round-2-results/), it is time for a deeper dive into the results. The goal of this analysis is to provide transparency and to gain insights that help us to improve and prioritize new features.
Within the scope of this article, we are limiting ourselves to the Deep Funding Vote and are not doing a separate analysis of the Loyalty award vote that happened at the same time.
Changes in the voting process
Before we dive into the details, it is good to recollect the changes we implemented in the voting process since the previous Deep Funding round:
- Quadratic voting: We introduced Quadratic voting, meaning instead of one token one vote, we calculated the square root of each person’s token balance, as an equalizing measure. It still gives you substantial extra weight to have more tokens but this measure ensures a more balanced weight between small and (very) large token holders.
- Reputation rating: We introduced another weighting mechanism, based on someone’s reputation score that was generated by being active and constructive on our proposal portal. We first experimented with this in our first Governance round and this gave us sufficient confidence to roll it out in an actual Deep Funding round. Besides giving active members more impact, it also serves to protect the process against strategic ‘wallet splitting’. Without reputation weights, it will be advantageous to split your wallet into multiple smaller ones to reduce the effect of the Quadratic balancing. It will however be hard to build a good reputation over all these smaller identities.
- Multichain voting: For the first time we enabled voting with both AGIX on ADA and AGIX on ETH.
- Wallet linking tool: We offered the users an option to link multiple wallets to their IDs. This enables the users to apply all their AGIX (on multiple chains and multiple wallets per chain) Without the hassle of going through the voting process twice or having to transfer their tokens to a single wallet.
- One signature per pool: In order to make the process easier, the users could cast a vote for all projects in a pool, and sign once to submit them. This is significantly easier than having to sign for each separate proposal vote.
There were also process changes that may have impacted the overall results such as peer reviews and the eligibility process, as well as some website changes, but although interesting in themselves, we’ll leave them out of this analysis.
Statistics Round 1 vs Round 2
Before going to the insights and conclusions let’s have a look at some numbers: and compare them to round 1 where possible. (limitations: there were fewer votes in round 2 because we signed per pool and not per project. There may have been people in round one that voted with multiple wallets whereas in round 2 they might have linked their wallets and voted only once, representing multiple wallets.
Round 1 |
Round 2 |
|
Total amount available |
1 Million USD in AGIX |
0.5 Million USD in AGIX |
Total amount awarded |
$ 741.000 |
$ 449.212 |
Nr. of pools |
2 |
5 |
Nr. of submissions |
47 |
89 |
Nr. of eligible projects |
28 |
57 |
Nr. of awarded projects |
12 |
17 |
Nr of wallets that voted |
158 |
187 |
Nr of collections |
n/a |
144 |
Summarizing this data we see that while halving the total amount to be awarded, we doubled the number of submitted proposals!
At the same time the number of wallets that voted has gone up as well, but offset against the potential inflow of new Cardano wallets this is a relatively modest growth.
Collections and AGIX balances
If we compare the balances before Quadratic voting this is the result.
There is clearly one gigantic whale at play. If we take out the one whale the result looks like this:
If we do the same with the quadratic voting taken into account what we see is this:
Including whale:
Excluding whale
We could conclude that this graphic, including quadratic voting and excluding the whale, shows a reasonably healthy balance of weights.
The whale
When we zoom in on the behavior of the whale (Btw for those that suspect that this whale is somehow related to the Foundation; this is NOT the case.) We say that its voting behavior does not really differentiate substantially from the average. We can see a few projects that would have benefitted if we had not taken quadratic voting into account, but there is no clear pointing finger here. However, if we take out the votes from the whale completely there is one significant difference. Overall the outcomes would have been almost the same, except for just a few changes. However, there is a large difference in pool C. Without the whale not just one but 5 projects would have been awarded in Pool C! This shows that we have to be careful about assigning conclusions to the outcomes of the vote alone.
There are (only) a few other wallets displaying the same behavior, in pool C, but the impact of the whale on the outcome in this respect is significant.
Voting graphs
We can get a different perspective on the dynamics of voting by reviewing the infographics made by Robert Haas with his open-source graph visualization library gravis.
In each graphic at the top row you see the wallets/collection IDs and on the bottom row the proposals. Green proposals are awarded.
– The thick red lines represent ‘1’ votes.
– The orange lines represent ‘2-6’ votes.
– The light green lines represent ‘7-9’ votes.
– The thick dark green lines represent ’10’ votes.
The collections that have a reputation weight are pink, and the size of the circle represents the final voting power of the collection.
(The order in the row of wallets is determined by the plain token balance)
Pool C: Projects in the ideation phase
Main insights from these graphics:
We see an overall mixed distribution of red and green lines coming from the wallets. We interpret this as a healthy indicator of authentic voting behavior.
We see significantly fewer extreme votes (ones and tens) compared to round 1. We made a point beforehand that this behavior was not desired and this seems to have paid off.
Most collections are voting for a large number of projects, with a distribution of votes. We can take this as an indicator of clear engagement from the community.
Perhaps most significant is the low density of ‘pink dots’ this shows there are relatively few collections with a reputation score. Zooming into this we see:
- We have 276 users in our export from the proposal portal (Swae)
- 71 have a reputation score > 0
- But out of these 71, only 6 voted, or at least in a recognizable way.
This is rather disappointing and asks for deeper investigation. We have not yet found a good reason for this and some sample tests of known users showed no strange deviations.
Proposal portal engagement
We received an export from Swae containing 276 users. From these users, the majority was just incidental, but 71 had a reputation score that is higher than 1, based on these predefined variables:
Variables scoring |
|
Comments and feedback created |
|
3 |
Points per comment created |
0 |
Points per reaction created (thumb up or down) |
0 |
Point per emoji reaction given (positive or negative) |
Comments and feedback received |
|
2 |
Points per positive comment reaction received (thumbs up) |
-3 |
Points per negative comment reaction received (thumbs down) |
2 |
Points per positive reaction received (Genius/Happy/Hot/Clap /laugh/Love) |
-2 |
Points per negative reaction received (sad, angry) |
We calculated the scores and distributed the users in different tiers like this:
Voting weights |
|||
Percentile |
Tier |
Weight factor |
|
96-100 |
A |
5 |
|
81-95 |
B |
3.5 |
|
61-80 |
C |
2.5 |
|
31-60 |
D |
2 |
|
11-30 |
E |
1.5 |
|
<0-10 |
F |
1 |
|
no points |
G |
1 |
Loyalty rewards
Back in December when we started round 2 we allocated 100.000 AGIX as a reward for engagement. The total Dollar value of the rewards at that time was something like $5,400,-
At the moment of writing this article, this value has increased to $44,000, which is very substantial! This amount will be distributed over 81 wallets, making some of the rewards a bit more substantial than we anticipated. We will however not change this situation, since this is what was agreed upon beforehand. For the next round, we may have a different way of calculating the awards, in order to avoid large swings in dollar value.
Some statistics on the Reputation rewards:
The bulk of the rewards is allocated to the constructive activity. Only 5% of the total rewards are allocated to IDs in the proposal portal that also voted.
Total awarded amount for reputation: |
100,000 |
|
Reserved for voting by contributors |
5,000 |
95,000 AGIX |
Percentile |
Nr of IDs per tier |
Loyalty rewards p.p. |
Relative rewards p.p. |
||
Tier |
D |
15% |
96-100 |
4 |
3,563 |
0.41 |
Tier |
D |
25% |
81-95 |
12 |
1,979 |
0.23 |
Tier |
D |
25% |
61-80 |
17 |
1,397 |
0.17 |
Tier |
D |
25% |
31-60 |
22 |
1,080 |
0.11 |
Tier |
D |
10% |
11-30 |
16 |
594 |
0.07 |
Tier |
D |
0% |
<0-10 |
13 |
0 |
0.00 |
Tier |
D |
0% (no activity) |
0 |
On top of these amounts, we have allocated 312.5 AGIX for those that voted (16 accounts, but only 6 of them had a reputation score > 0)
Conclusions and improvement points
Improvement areas
We have onboarded on this journey with a learning mindset. Be agile and make changes, and keep what works. Sometimes we are implementing processes as an experiment without yet having optimal tooling for this. We think this is the right way of doing things, for speed, out of a sense of urgency, and also to ensure that any tooling decisions and requirements are grounded in reality. The inherent risk is that this experimental approach comes with manual tasks and thereby some risks of errors. Unfortunately, this did materialize in 2 instances: We manually migrated the proposal content from the portal to our regular deep funding site for future reference and administration purposes. The downside was that we didn’t spot a few last-minute changes to some of the requested amounts. Another error was made in calculating some of the reputation rewards. Some users have received an amount that was too large due to an incorrect variable in a very large and complex spreadsheet. (We have contacted the related people and friendly asked the to return the additional amount).
Also without these errors, it is clear that the process is becoming more and more complex and harder to maintain, especially with the very few people assigned to these tasks. Therefore one of our main objectives is to create or enhance features that will enable automation, reliability, and transparency.
Another improvement point is related to the number of engaged people on the portal that can be linked to voting activity. For future health of the system and high-quality outcomes, we will continue to motivate both portal contribution and voting behavior. By better integration of our wallet linking tool, we hope to be able to make more connections between these two activities.
Finally, a remaining question is on the topic of transparency. Until now we have tried to be very transparent in explaining the results and the processes and decisions leading to these results. Where we have drawn the line, up until now, is releasing information that is potentially traceable to specific identities. Going forward this is a topic to review.
- Should we expose all wallet addresses and voting behavior?
- Should we show the reputation rating for each profile on the portal
- Should we even show the wallets associated with profiles on the portal?
All this will lead to greater transparency, which is positive in itself. A potential risk could be, however, that some individuals that value optimal privacy will prefer not to participate anymore. A classic case of privacy vs. transparency which are both important values in the blockchain work (and beyond). Should we go for full transparency, we will give any of the current users the opportunity to opt-out, be forgotten, or change their public ID. This topic could be interesting to explore in more depth in the next, upcoming, governance round.
General results
Overall we are looking back to a very successful round 2!
We are extremely happy with the number of submissions, and the engagement around them. Some processes and tools need to mature, but we are confident that we are on the right track and are looking forward to pushing these further. While we strongly believe in quality over quantity, there is room for further growth in the number of engaged community members and voters. Larger numbers will make the system more resilient and harder to game, providing we also have the right processes in place.
All these changes go hand in hand (processes, tools, and number of contributors). And they will need time to mature and come to full blossom. But to be in this place already, after only 2 rounds, and with many known improvements on the radar, we see a VERY bright future ahead for Deep Funding.
Therefore our gratitude goes to all the teams that submitted a proposal, whether awarded or not, to the awarded teams and community members that are taking charge of some special tasks, and last but not least to all the community members that were active on the Proposal portal and that voted for one or more of these great projects. We could not have done it without you all, and we are looking forward to many more, increasingly successful Deep Funding rounds!
Thank you all and see you in Governance Round 2 and subsequently in Deep Funding Round 3!
Join the Discussion (0)
Please create account or login to post comments.