Rethinking Retroactive Rewards

hudsonsims

Summary

With the end of the period approaching, it is time we established how to distribute retroactive rewards.

With the DAO expressing badges not being a priority, suggest simply voting after the final mint of the period on different categories where members can distribute their voting weight. The team would then tabulate the distribution according to the weights and allow the DAO to make a final approval.

Steps

Weighted vote on categories:

Days voted
number of fragments voted on
number of Boosts
number of Winners voted on before leaderboard (if possible)
VP spent in each round (doubles up on active rewards)
number of gov proposal voted on
buy back & burn $BOTTO
send to treasury

Divide up retroactive rewards pool among categories and weight by rounds participated in

E.g. If “days voted” got 10% of the vote, 10% of the retroactive rewards pool would be distributed to voters based on the proportion of days voted relative to everyone else.

Background

In BIP-14 that established the parameters of the second period, we introduced direct distributions of revenue. There was significant debate about how to go about it, and so two systems were established: active rewards and retroactive rewards.

Active rewards are 25% of revenue distributed each round to voters weighted by the amount of voting points they spent. This was coupled with a change to voting point generation such that everyone generated voting points linearly where 1 $botto = 1 vp generated per week, and removing a cap on voting points such that it could be easy to vote as much as you desired on one fragment.

Some members disagreed with this because of how it favored whale voting behavior to pick favorites rather than provide a diverse amount of feedback across many fragments. The counter to this was that there was still strong incentive alignment by whales to give Botto good feedback, and so active rewards for voting points spent won a good deal of support.

Retroactive rewards were established to give the DAO an opportunity to decide on the other 25% of revenue at the end of the 12 week round and the available hindsight of what was seen as valuable contributions and potentially reward more diverse behavior.

BIP-14 and other discussions proposed a badge system to help facilitate this such that it would be easy to see different types of contributions to inform what to reward.

That system was not set up or agreed on in BIP-14, and when a sketch of that system was put forward, the response of the community was that it was not a priority over other systems like setting up vBOTTO. (Side note: the feedback was not that it wasn’t wanted, only that other improvements were a higher priority. Further, an establishment of vBOTTO could allow for a badging system, soulbound or not, to be implemented permissionlessly.)

The DAO must now decide how to distribute retroactive rewards. It may also want to consider if this is the revenue distribution process it wants to keep going forward.

Specifications

A. Defining categories to reward

Distributing rewards will be most straightforward if we pick behaviors that can be proportionately rewarded. We can also only reward behaviors that have been tracked throughout the period, and so the onboarding bounties are a good reference to start with.

Days voted
number of fragments voted on
number of Boosts
number of Winners voted on before leaderboard (if possible)
VP spent in each round (doubles up on active rewards)
number of gov proposal voted on
buy back & burn $BOTTO
send to treasury

Divide up retroactive rewards pool among categories and weight by rounds participated in relative to others.

B. How to vote

We can choose to spread votes across multiple categories such, and portion out the revenue distributions accordingly. Snapshot allows for this mechanism.

The team can then provide an accounting based on the distribution of voting before executing.

This vote should happen after the Period ends so as to not allow easy gaming.

Timeline (tentative, may be pushed back to review scenarios)

Now-Jan 31 - Discuss and finalize prop, add/remove categories as needed
Jan 31-Feb 2 - Weighted vote
Feb 3 - Final Tabulation
Distribution prior to new Period

juanje

Looks good at first glance. The first thought that came to my mind was including something related to governance participation, so:
How about adding "# or % of proposals voted during the period" in the tabulation?
Someone may argue that this would not be really related to voting, but imo we should start thinking about incentivizing governance participation, and this could be a nice starter.
Sidenote: I wouldn´t likely benefit from this in this period

mazzi

juanje

I like the governance participation category as well

choobie

juanje How about adding "# or % of proposals voted during the period" in the tabulation?

I think this is a great idea too.

hudsonsims

juanje How about adding "# or % of proposals voted during the period" in the tabulation?

really like this, will add # of proposals, which will be proportionate to others anyways

mazzi Not a fan of "Days voted" and "Winners voted in"
Reason is that "Days voted" can be gameable having a bot voting each day with 1 VP, while "Winners voted in" encourages whales to boost a single piece instead of multiple ones.

don't want to jump on removing categories this early in discussion as people can choose to not vote for them, but noted these as "gameable" and to potentially remove. curious to see if others agree.

juanje

Something similar could be done in terms of forum participation, btw.

mazzi

Not a fan of "Days voted" and "Winners voted in"
Reason is that "Days voted" can be gameable having a bot voting each day with 1 VP, while "Winners voted in" encourages whales to boost a single piece instead of multiple ones.
"Fragments voted on" and "Boosts" are fine, encouraging the spread of votes towards lots of fragments.
"Round voted in" and "VP spent" are reasonable, since we want to reward constant (not defined by days though!) and decisive (by VP spent) curators.

I'll think about new categories as well

Cooper

mazzi I generally agree with Mazzi on the categories. Voting for winners is maybe not so important for the protocol. And rewarding days voting could be mentally negative because I don’t think it’s good to promote that kind of grinding behavior that make you feel like you loose rewards if you miss a day. Could take the joy out of it.

Number of fragments voted on also has the risk of becoming a grind, but I see the benefit in awarding training Botto. That’s missing now.
I would like to ask - how important is the 1/1 voting for the protocol? And how much do we need more votes?
The risk is I would maybe vote faster and not enjoy the art as much. Don’t like that.

In that sense vp spent is good, then you can spend as much or as little time as you want.

Can we award training Botto in a way that doesn’t lure some of us into grinding?

choobie

Personally I'd like to see a combination of the weighted vote categories with the least weight given to VP expenditure.

hudsonsims

added 2 more categories based on some comments in the gov discussion channel on discord:

buy back & burn Botto
send to treasury

quimp

+1 for governance participation

The 2-phase system is straight forward and it has my support.

I'd like us to be open to more novel ideas than ETH distribution, too. What if we rewarded active voters with art?

A raffle to win a Botto edition of a community favorite.

Artist grants for editions: Mia Forrest derivative of Complexity Incarnate, an Apocalypse Art derivative of Blossoming Cadaver, a Tjo's Witness Intensify, a Grant Riven Yun's Another Sector, etc.

oiseau

Agree with Cooper that it'd be useful to have a clearer idea of what we're aiming for, as well as some insight on voting patterns so far (maybe that'll be easier when we get the API though?). Even on an individual level, I've never really been sure what's a useful approach in the voting -- and have wondered if that was a decision from the team not to 'teach' us so we don't boss Botto around too much.

Doing the category voting seems like a reasonable solution for this round given everyone's different interests. After that, agree with quimp and personally would prefer spending on things that feel fun, creative and exciting. Fun is also an incentive, but it's harder to measure.

hudsonsims

Cooper I would like to ask - how important is the 1/1 voting for the protocol? And how much do we need more votes?

sharing here response from @Quasimondo:

I think botto does get enough votes, yes. Generally I would like to find a balance for rewards which does not only favor people with a lot of VPs.
Which is why I believe days voted is a fair measure that should stay in
I also think that number of different fragments voted should get a large ratio
I also think winners voted on is important [if we can easily track those who voted on a winner leaderboard]
I am not too concerned about bot votes at the moment at least not in this phase
number of boosts is something I feel is already covered by the active weekly rewards

Some updates based on discussion:

Voting Data - We can release the voting data so that people can better assess the categories they want to award, and scenarios of how different weight distributions would look in the payouts. We will likely push out the snapshot a few days to give people time to dig in. cc oiseau

stfnfhrmnn Could also consider variations of those, tied to picking winners or almost-winners:
...
How about further categories that deemphasize the distribution to whales (which already receive by virtue of spending large amounts of VP on winners, potentially)?

As for more variations on categories, I believe we should keep it simple for this vote. Categories need to be pretty self explanatory when listed in the snapshot as I wouldn’t count on everyone following closely. However, I think these suggestions could be good to consider for regular rewards going forward if we are to get rid of this retroactive rewards format, along with other fun formats like quip suggested.

Regarding use for burning, treasury, or LP, it is important that we follow through on what we’ve promised, which is that 50% of revenue would be distributed to participants. LP I think is off topic for this and requires its own discussion, which is ongoing in #governance-discussion on discord right now. Burning and treasury are indirect distributions, so can work as categories among others and provide good data on whether this is where people want to see revenue going.

Rounds multiplier - Will shift this to a “multiplier” rather than a category. We want to distribute according to round participation, which helps weight for time invested without biasing to people’s methods (voting over many days vs. focusing on one day each week, etc.). The way this would work is to simply count your rounds proportional to everyone else’s and multiply by your other qualifying categories.

sw5park Advocating removal of categories that incentivize high-frequency, short-term grinding behavior:

Days voted
Quantity of fragments voted on

Days voted and Quantity voted - These have gotten mixed support, so important to keep since people can choose to vote on something else. Days voted can be an incentive to keep coming and checking in and do some voting each day, and it is important for Botto to get feedback on many different fragments. I believe there’s been some criticism of “grinding” that is actually an element of simply supporting Botto. Returning each day can be good to continue to see new art and get some votes in.

General notes: As we just saw yesterday, sybiling is a real risk. The minimum 2000 $BOTTO staking requirement significantly mitigates that activity affecting these retroactive rewards, but we should absolutely be wary of things that are notable.

What’s useful about this format is that as a retroactive vote and weighting by rounds helps prevent gaming on this upcoming decision. And if you don’t want to reward a category, put your voting weight somewhere else.

However, given the repeatable nature of each period, these benefits of retroactive rewards probably won’t work as well going forward as they become more easily gameable. We can use the outcome of this vote to see where people want to see weights to determine how we want to reward in the 3rd period and whether to keep retroactive rewards at all. We can also consider other weighted methods and a staking lockup multiplier, but will leave that for a separate BIP.

Tldr updates:

Voting data release with scenarios + added time to analyze
Rounds become multiplier, removed as category

sw5park

Advocating removal of categories that incentivize high-frequency, short-term grinding behavior:

Days voted
Quantity of fragments voted on

stfnfhrmnn

I feel "Days voted" and "Quantity of fragments voted on" are both good ways to track participation. Yes, they could be rigged with the help of bots, but if those categories stay with a lower weight in the distribution, that might lower the danger of that happening.
Could also consider variations of those, tied to picking winners or almost-winners:

number of times voted on a fragment that made it into leaderboard (or top 3, top 5)
number of different days voted on a fragment that made it into leaderboard

How about further categories that deemphasize the distribution to whales (which already receive by virtue of spending large amounts of VP on winners, potentially)? Is it possible to track VP spent per address relative to (average) daily VP generation? This would enable things like:

VP spent on leaderboard fragments relative to VP generation (maybe excluding boosts?)
VP spent on leaderboard entries that did NOT win relative to VP generation

juanje

Another behavior/metric for the tabulation: # of times people vote in #staking-descriptions
Reason/background: few people do (just looking, fewer than 15 last time)

choobie

juanje Yeah I think this is an interesting thing to explore actually, it signals a higher level of participation. One caveat is maybe that the process for getting your wallet connected and verified (to get the Staker role) isn't the cleanest in terms of implementation, so maybe some don't know how to participate. I'd like to think the weekly announcements tagging stakers would spark enough curiosity there though.