Summary:
The PDAO (via the guardian for now) should reduce the maximum RPL collateralization rate from 150% to 75% on steady, but slow schedule over a period of multiple rewards cycles.
Rationale:
The upper levels of collateralization are not adding much in the way of protocol security while are causing lower collateralized folks to have their rewards diluted. This dilution could cause folks to only run a bare minimum of collateral, and potentially reducing the security of the protocol (this probably needs some analysis).
This also helps folks who migrate to LEBâs to be able to withdraw excess collateral sooner if RPL appreciates.
This should be implemented slowly, so as to not cause too much of a price impact on RPL at once. For example, a draw down schedule could be executed as such over 10 cycles, alternating between 5% and 10% reductions:
Cycle 0: 150% (Baseline)
Cycle 1: 145% (-5%)
Cycle 2: 135% (-10%)
Cycle 3: 130% (-5%)
Cycle 4: 120% (-10%)
Cycle 5: 115% (-5%)
Cycle 6: 105% (-10%)
Cycle 7: 100% (-5%)
Cycle 8: 90% (-10%
Cycle 9: 85% (-5%)
Cycle 10: 75% (-10%)
Considering the role of RPL as a form of collateral is changing dramatically with Atlas (LEBs), I wouldnât support a change that would add greater risk (less collateral) until we at least know how LEBs will impact the overall security model.
Itâs also not clear to me that this would be beneficial at all, but I especially think itâs risky to overlap with other tangentially-related changes to the base collateral design.
My initial stance for any change is âHow does it help the protocolâ â if I donât get a clear answer, I strongly prefer to leave it alone. Changes requires analysis ahead of time, and our analysis can be wrong leading to unintended consequences, so I think itâs proper to lean against changes.
The discussion in terms of NO advantages
Here I donât see an advantage to the protocol suggested. I see 2 conveniences to NOs suggested (less dilution for low collateral folks, being able to withdraw with less growth via appreciation or yield).
For RPL yield: thatâs double-edged. High-collateral folks would lose out - so this is a benefit to some NOs at the direct cost of other NOs.
For being able to withdraw with less growth: thatâs also double edged. Right now, if NOs want to rebalance, theyâd have to exit. If NO demand is long-term greater than rETH demand, as is expected by many including myself, then this would drive some turnover and let others participate. In this case, itâs a benefit to incumbent NOs at the cost of potential NOs.
I included the spoilered discussion there, but itâs really not my main point. My main point is that we should focus on protocol level benefits, health, risk, etc, which hasnât been addressed here. To me, weâd be taking on a risk for no benefit.
As Iâve mentioned in discord, having a huge maximum value of staked RPL per node is of limited usefulness to the protocol past a certain point and very large values result in a risk of undesirable RPL ratio volatility when minipools begin to exit.
I would tentatively support a re-examination of the maximum collateral %, especially in the context of the range for upcoming LEBs.
However security &reward dilution for those at low collateral are poor arguments for doing so, as Valdorff pointed out, you are simply moving rewards from one group of node operators to another.
Some better reasons would be a less daunting lockup for wary node operators or the prospect of otherwise staked RPL being instead used for additional minipools and for providing RPL liquidity.
Whether a lower cap would actually result in the RPL being used in this manner is very speculative though and there is some indication from the current amount of ineffective staked RPL that many node operators would prefer to simply hold it for appreciation.
Thank you everyone! Certainly a lot to consider here beyond what I was thinking of.
Coming from a security standpoint - is there an objective way to determine what a sweet spot for collateral might be? And would it make sense to concentrate rewards somewhat at that level?
From what i can tell, the beacon chain has ~13,500K ETH, and has had ~215 ETH of slashings (.00159% insurance needed)
rETH has ~125K ETH, but has 125,000 of NO ETH and 114,000 ETH worth of RPL (190% insurance provided)
It is evident that we are overpaying on security.
And these two statements cannot both be true:
10% RPL is appropriately collateralized
150% RPL is appropriately collateralized
However, I canât support this proposal, because I think that once tokenomics are in place for a project, changing them in a punitive way creates more problems than it is worth- hurt feelings and uncertainty for current/future investors. I think finding other ways to put the excess staked RPL to work would be beneficial to the protocol and to RPL holders.
My rocket pool roadmap (which deals with excess collateral), in order of importance:
UEB (best for RPL): NOs use RPL over ~50% collateral to spin up minipools entirely of protocol ETH. Because RPL can be used 1:1 with ETH for minipools, there is little reason for a NO to sell RPL to ETH. This also gives a way for RPL rich/ETH poor whales to earn returns without speculative selling or LPing. Supply and selling pressure goes down.
LEB (best for NO): bring many new entrants into the space with by offering outsized returns with a lower RPL bond (letâs say 10% of NO ETH); as liquidity will be largely absorbed by UEB, a small RPL influx per NO will have outsized price impact.
Lowered commission (best for rETH): UEB allows capital efficiency, as these NOs could receive a small commission (letâs say 5% of the 32 minipool ETH) instead of 0 commission currently on their RPL. LEB with low RPL bond allows capital efficiency, so a lower commission would still increase NO returns by multiples. Pay back this increased capital efficiency to rETH holder in the form of lower commission because it will amp up rETH TVL and because it is the right thing to do.
I think this is an interesting idea and yet Iâd like to have a better understanding of how all of this changes with LEB8âs. Can we quantify the security provided to the protocol versus max collateral limit? I would want to see an analysis of this before supporting this proposal.
In terms of the value proposition, I think the most compelling argument is that a lower cap may result in a perceived lower price of entry as a NO because they will not have to stake as much capital without the opportunity to withdraw it prior to being over-collateralizedâŚthis may draw in more NOâs and provide them with more comfort regarding investment in and staking of RPL knowing that their ability to withdraw isnât as âfar awayâ ratio-wise as it once was with a 150% cap. That said, supply of NOâs is not the issue right now.
I tend to agree with Uisce in that it is speculation as to whether a lower cap would result in NOâs actually converting their excess RPL to eth for more minipools. The data available right now suggests notâŚ
Sure. Based on the LEB discussion, consensus is that 8+2.4 ETH per validator is very safe. Nobody can honestly argue that 16+12 or 8+16 would be unsafe. Especially since RPL collateral is not penalizable in the context of MEV stealing.
Rewarding NOs for absolutely unnecessary collateral levels is by far the highest spend of inflation and at the same time we are apparently struggling with funding basic stuff like marketing and an appropriate team size. Iâd support this or any other proposal to fix NO rewards as long as it comes with a reallocation of RPL savings.
This proposal doesnât have any savings I see. Itâs a redistribution to a different set of NOs, but isnât proposing changing the 70% of inflation going to NOs.
I agree with the several voices that have pointed out that 150% cannot be needful for security if 10% is what we need for security. That doesnât imply 150% has no benefit, it implies that isnât it.
One plausible benefit is adding friction to sell RPL, especially in large amounts. This likely reduces price volatility. High volatility is bad because it makes the secondary collateral usage (the 10% minimum) less reliable.
In any case, the burden of showing a clear benefit remains with the proposed change, not with the status quo.
Yep. Iâm just saying thatâs the direction Iâd take this and that makes sense to me. I took OP more as a conversation starter.
If we go down that path, I think the burden is kind of on the status quo to justify spending 50%+ of inflation on collateral that is not needed for security.
Very happy to see this conversation started, thank you @NonFungibleYokem.
I tend to agree with the sentiment that now isnât the best time for this change, but I do agree with @knoshua that inflation and budgetary needs for the pDAO are a good place to focus, and reducing waste is a positive. Iâd like to begin analyzing the possibilities for future adjustments more deeply.
Hereâs my feedback on the proposal itself:
How narrow of a range can we have without regularly pushing a significant number of NOs out?
The answer to this will tie into the fundamental ratio floor which @boodle has researched before here (note to @peteris, a realtime version of this for RocketScan would be wonderful!).
In the long run, I prefer to settle on a tight RPL collateral range which is derived from a protocol-level security payment standpoint since a wide range is uneven and therefore inefficient. Even a 10-75% range is fairly wide, but we do need some range to account for NO UX since RPL/ETH value fluctuations are inevitable.
The complication of a collateral range in the first place is another direct consequence of the protocol having its own token instead of using ETH. Itâs unrealistic to change this any time soon, but if thereâs ever any way to transition away from RPL, we should⌠or we should at least aim to prove that itâs impossible without destroying protocol participantsâ wealth. This point could be a whole research paper in itself.
As @epineph pointed out, implementing UEB minipools first would allow for a more graceful transition to a lower maximum collateral, so I also prefer to see this implemented before collateral changes. Otherwise, I would expect to see some instability in the system as NOs sell off RPL in anticipation of tokenomics changes, which might ultimately result in a downward spiral. Even though itâs not a death spiral due to the fundamental ratio floor, a large portion of RPL/ETH price is speculative, and a downward spiral would still adversely affect the RP mission. With UEBs, however, NOs can simply move any excess collateral into new minipools and keep the system more stable.
Do we need to continue with inflation payments increasing at a 1:1 rate across the whole range of collateral levels As Iâve stated elsewhere, RPL needs to remain at a higher APR than ETH in the long run or weâll encounter instability as everyone migrates to the lowest level of collateral for efficiencyâs sake. Maybe once we settle more on a specific target APR for the long run, we could adjust the slope of this payment range to reduce the disparity between low and high collateral levels, though we obviously canât completely eliminate it.
As Iâve mentioned before, thatâs higher ROI including both yield and appreciation.
While I think thereâs a massive downside if everyone decided to rebalance at once, doing it gradually as different folks hit their personal thresholds based on their personal guesses about expected appreciation shouldnât be much of an issue. Itâs also worth noting leaving and re-entering has a cost (both in terms of gas and an opportunity cost for the time in queues). In the extremely long run, I do think we expect RPL price will stabilize somewhere and most folks will opt to be on the low end of RPL.
Why is this a goal? When you say a wide range is inefficient, what do you mean? Inefficient for who? The protocolâs cost is the inflation part, not who gets it or how much RPL they hold.
This is a goal because extra payment for security is inefficient for the protocol. If 10% collateral is sufficient, why bother incentivizing more than this?
Perhaps the best answer from a pure efficiency standpoint at a protocol level is to eliminate all NO RPL inflation entirely and only set a minimum threshold for collateral to unlock commission. Realistically, though, I think the current UX of a an incentivized range leads to a more stable system overall since NOs arenât incentivized to ride the minimum collateral line for the sake of their own efficiency.
Ok - it sounds like youâre with knoshua here. Youâre not just saying we need to narrow, but âwe need to narrow and redirect the RPL that would have gone to the wider range to somewhere OTHER than the NOs in the narrow rangeâ.
Not necessarily â it depends on whatâs meant by âredirectâ. We could simply compress the range without changing inflation since that would still be more efficient from a protocol standpoint; fewer RPL are locked up unproductively in this case. To preserve system stability in the short term, this option makes the most sense to me after UEB minpools.
That said, Iâm not against the idea of rethinking inflation allocations either. In the past, Iâve advocated for a dynamic inflation model which is intended to make NO inflation dependent on network conditions (i.e. need for new NOs) in addition to posted collateral.
Alternative solution: There is no range. RPL rewards are paid on a per minipool basis. NOs still need to put up 10%+ to start a minipool and people can manage the trade-off between dropping below 10% and too much collateral however they like.
I think this is the optimal solution from a security/efficiency standpoint. But what to do with all that overcollateralized RPL except for an immediate market sell on implementation?
I donât support this idea, for the same reason as @epineph: the high yields are exactly what incentivize node operators to reach high collateralization rates in the first place. Reducing the max collateralization limit could reduce RPL buy sizes and allow for a slow yet persistent selloff.
Some math on this idea -
If the per-minipool collateralization incentive budget was initially targeted at a flat 4 RPL per actively staked minipool, right now that would cost the protocol about 32K RPL per cycle. Or about 2/3 of what is currently being spent. Then as more minipools come online, that budget would be diluted, across more minipools. Especially with LEB8 conversions. It would probably make for a solid incentive to convert 16 eth minipools as demand allows. And given the increase (in absolute quantity terms) in RPL collateral could minimized the market impact on RPL.
That opens up about 16K RPL/cycle for âother stuffâ.
But if we really want to preserve the security aspect from overcollateralization, it might be worth driving those excess rewards into a separate and additional collateral tranche of node operator supplied rpl/eth liquidity instead of just locking up pure RPL - as @Xer0suggested a while ago.
Under the current setup, if there were a very wide scale slashing event (about the only reason right now that RPL collateral would be touched), open market RPL liquidity could very easily dry up faster then the collateral could be auctioned and RETH holders could take losses. But if node operators were to supplement their RPL stake with RPL/ETH liquidity tokens (Balancer or Curve, etc), and then provide additional incentives for that, that extra capital would be doing a service for the whole protocol by being RPL buyers of last resort in case of a serious black swan event. And in nominal market situations, there is a lot more liquidity for onboarding new buyers to the protocol.
That obviously presents operators who want additional RPL rewards some more substantial risks (divergence loss, liquidity pool contract risks) but it would give the protocol more of a service then just locking up supply.