[Proposal] New Delegation Strategy

Proposal: New Delegation Strategy

What do you propose?

We propose a new delegation strategy for Marinade that involves PSR bonds and bidding for stake. It maximizes and fully protects yield for stakers, improves Solana network through better liveness and censorship resistance, and enables full redelegation of stake between validators in just one epoch.

What is the rationale behind the proposal?

Marinade delegation strategy can be viewed as an optimization task between four parties with different goals.

1, Solana: maximize performance, liveness and censorship resistance
2, Stakers: maximize expected yield and yield stability
3, Validators: maximize profitability, predictability and speed of the delegation strategy
4, MNDE holders: maximize long-term value of MNDE through maximizing Marinade TVL

Challenges with the existing delegation strategy

The existing delegation strategy still remains suboptimal. The most significant challenges we’ve identified, and aim to address with the revised strategy, are:

  • slow redelegation — validators often wait for over a month for a redelegation of stake
  • high concentration of delegated stake to handful of validators as a result of directed stake and MNDE voting
  • directed stake and MNDE voting decreases stakers’ yield as this stake is often being directed to high-commission validators

The new delegation strategy

To ensure readers can easily navigate the complexities of the new strategy, we think it is beneficial to initially examine the delegation strategy from various perspectives.

The new delegation strategy from validator’s perspective:

We believe the new process is simple, fast and predictable. The process for validator looks like:

1, Estimate how much yield you can generate for stakers using a calculator. If you can generate more yield than the least yielding validator in the current delegation set on some amount of delegated stake, you will get all the stake you want in the next epoch.

2, Send a refundable deposit into your PSR bond plus some more for bidding.

3, Place a bid with 3 parameters:

  • max_yield: the maximum yield you are able to generate while remaining profitable (max_yield must be ≥ 0.01% higher than the current realized_yield (= yield the current least yielding validator can provide))
  • min_stake_wanted: the minimum amount of SOL you want delegated (≥ 10k SOL)
  • max_stake_wanted: the maximum amount of SOL you want delegated (≤ 2% Marinade TVL)

Each validator in the Marinade delegation set will only pay as much as the least yielding validator in the delegation set. This strategy guarantees that as long as you are in the delegation set, you will be profitable. And if you are removed, your initial deposit into PSR bond will be fully refunded.

It is important for validators to understand that the best strategy is to place their max_yield (e.g. 8.5% APY) close to their true max_yield straight away, because:

  • Placing the max_yield parameter lower (e.g. 8% APY) than your true max_yield (8.5% APY) means that as soon as the least yielding validator in the delegation set yields 8.01% APY, Marinade will redelegate all your stake and you will miss out on 0.49% APY of profitability.
  • Placing the max_yield parameter higher (e.g. 8.9% APY) than your true max_yield (8.5% APY) means that as soon as the least yielding validator in the delegation set yields 8.51%, you become unprofitable.
  • Placing max_yield parameter higher than your true max_yield and lowering it later so that it causes a redelegation of stake away from you is disincentivized by effectively slashing your refundable deposit in validator’s PSR bond.

The new delegation strategy from Solana’s perspective:

The new delegation strategy aims to improve Solana’s performance, liveness and censorship resistance in three ways:

  • economically incentivize validators to have maximum uptime and minimum skip rate
  • sufficiently decentralize across ASOs (datacenters) in order to improve liveness
  • sufficiently decentralize across countries in order to improve censorship resistance

The new delegation strategy from stakers’ perspective:

The new delegation strategy is designed to likely offer the best yield on the market with 100% uptime protection. We are confident that the value proposition of staking with Marinade will become so compelling that it will be an unequivocal choice for most stakers.

The new delegation strategy from MNDE holders perspective:

The primary mechanism for long-term value accrual to MNDE should be a low fee derived from the yield realized by stakers through Marinade. Consequently, the accumulation of value in MNDE should be directly correlated with Marinade’s TVL and the associated yield.

Deep dive into the new delegation strategy

1, Eligibility requirements

We propose less strict criteria compared to the existing delegation strategy:

  • 1, Uptime for each of the last 3 epochs ≥ 80%

  • 2, PSR bond to cover for the sum of:
    – refundable deposit (RD) (e.g. 100 SOL / 100k SOL delegated)
    – one epoch worth of validator’s bid (e.g. 0.1 SOL per epoch per 100k SOL delegated)

  • 3, Effective commission ≤ 7 % (e.g. validator can have 100% inflation commission, but as long as it offsets it through MEV and bids so that the yield it generates is equivalent to ≤ 7% commission on inflation rewards, it satisfies the condition)

2, Scoring and delegation

Scoring of validators is based on a single parameter called max_yield.
Max_yield is the maximum yield the validator is able return to staker on some amount of stake (e.g. 8% APY on 100k SOL delegated). Max_yield is calculated as the sum of inflation rewards, block rewards, MEV and bidding that can be returned to the staker. Marinade should provide a calculator with easy-to-use GUI to help validators estimate their true max_yield based on which they can decide what max_yield to offer to ensure profitability.

Example of how true max_yield can be calculated:

  • Validator assumes 100% inflation commission, 100% MEV commission and 100% block rewards commission on min_stake_wanted and/or max_stake_wanted (e.g. 200k SOL).
  • Validator adds its server and other costs, expected block rewards, MEV tips and skip rate.
  • Validator would generate a net income of 16,950 SOL on 200k SOL, therefore true max_yield is 16,950 / 200,000 = 8.475 % APY.
  • Validator decides to place max_yield to e.g. 8.3 % APY, so that it always remains profitable even in a scenario where it becomes the least paying validator in the delegation set which means he has to return the max_yield in full (in other words, when validator’s realized_yield is equal to its max_yield).
  • How validator’s realized_yield is returned to staker depends on validator’s inflation commission and MEV commission settings. If validator has 100% commissions on both inflation and MEV (it keeps all yield to itself), then all of the yield to staker has to be returned in the form of bidding. If validator has only 7% commission on inflation and MEV, then bidding costs will be substantially lower.

Bidding: Each validator places a bid in a format of: cost per delegated SOL per epoch and specifies the minimum and maximum amount of SOL that can be delegated (e.g. I will pay 0.01 SOL each epoch for up to 85 000 SOL delegated). The right amount of bidding which leads to some concrete amount of max_yield will be displayed in the same calculator as in the previous scoring section. If validator gets full 85 000 SOL delegated, it will be paying 0.01 SOL every epoch as long as it retains those 85 000 SOL.

Constraints: At every point in time, it must hold that:

1, Marinade stake concentration per country ≤ 30%
2, Marinade stake concentration per ASO ≤ 20%
3, Solana stake concentration per country ≤ 30%
4, Solana stake concentration per ASO ≤ 20%
5, Marinade stake per validator ≤ 2% of Marinade liquid + native TVL
6, Marinade stake per validator ≥ 10k SOL

Last price auction:
1, Order all validators by max_yield.
2, Start filling all Marinade TVL from top to bottom while respecting predefined constraints.
3, The yield all validators who received stake must return to stakers is equal to the max_yield of the last validator Marinade delegated to.

Example:

Validator ID max_yield (APY) max_stake_wanted (SOL) stake_received (SOL) realized_yield (APY)
1 10.6% 95 000 95 000 8.12%
2 9.58% 1 000 000 200 000* 8.12%
3 9.4% 80 000 80 000 8.12%
166 8.12% 50 000 15 000** 8.12%
167 8.12% 200 000 15 000** 8.12%
168 8.10% 80 000 0 0%

*assuming Marinade liquid + native TVL is 10M and constraint #5 limits stake_received to ≤ 2% which is 200 000 SOL
** assuming there is only 30 000 SOL left to be distributed and there are 2 validators at the bottom max_yield competing for it, results in giving 15 000 SOL to each

3, MNDE directed staking and decentralization bonuses

The current delegation strategy in production splits Marinade TVL into three parts:

In this new delegation strategy we propose a new split of Marinade TVL into three parts:

  • Performance-based (85% of TVL): based on max_yield, constraints and auction described in section 2.
  • MNDE voting based (10% of TVL): no bidding is paid for this stake and it can only be delegated to validators in the performance-based delegation set that charge ≤ 7% inflation commission. Stake directed from MNDE voting must also have sufficient PSR bond coverage.
  • Decentralization bonus (5% of TVL): more decentralized validators in the performance-based delegation set with ≤ 7% inflation commission receive extra TVL for no additional bidding costs. The same decentralization score based on country, ASO, stake concentration as in the current delegation strategy is used. Stake received from decentralization bonus must also have sufficient PSR bond coverage.

4, Redelegation rules

Redelegation of stake between validators on Solana is costly for stakers because it takes one epoch during which it doesn’t generate any yield. The current maximum redelegation speed is 2% Marinade’s TVL per epoch, which causes to up to 2% less yield for stakers. Despite that, it sometimes takes over a month for validators to receive stake from Marinade.

The new delegation strategy solves this. It enables validators to receive all of their stake in one epoch and without incurring any long-term costs as long as validator behaves honestly.

It also reduces the need for redelegation through a novel call option mechanism. When a validator A in the delegation set is about to be removed the next epoch because some other validator B is offering a higher max_yield, the validator A can call validator B’s max_yield and remain in the delegation set.

The refundable deposit (RD) is equal to the amount of yield lost on the amount of SOL validator gets delegated for 2 epochs (e.g. 100 SOL / 100k SOL delegated). It composes of two parts of equal size (e.g. 50 SOL each):

  • delegation deposit (DD): is fully refunded when a validator exits the delegation set without lowering their max_yield and their stake is redelegated.
  • yield protection (YP): is fully refunded when validator exits the delegation set without rugging commission.

Both delegation deposit and yield protection are fully refundable to the validator as long as the validator behaves honestly.

5, The start of the delegation strategy

We propose a 10 epoch dry run in order to facilitate a price discovery of realized_yield and to make validators familiar with the new strategy.

Validators will need to have their PSR bond covered in order to participate in the dry run.

At the end of the dry run, the strategy would go fully live.

What is the expected positive impact of this change?

We are confident that this new delegation strategy addresses the key shortcomings of the existing approach and optimally satisfies the needs of all involved parties:

  • Stakers can expect one of the highest yield in the market, complemented by 100% uptime protection.
  • Solana will maintain robust decentralization, high uptime and censorship resistance.
  • Validators retain complete control over their stake amounts and are likely to receive it the stake in the next epoch.

Together with other features of Marinade, this delegation strategy positions it as the best staking product on Solana, poised to enhance Marinade TVL and significantly boost the long-term value of MNDE.

Any other considerations?

1, The potential of an alternative first-price auction model—in which validators pay their own bid amount, as opposed to the amount bid by the lowest-bidding validator—to produce a higher realized yield for stakers is currently under investigation. A thorough analysis, including detailed simulations, is actively being conducted to examine this approach comprehensively.

2, The proposed strategy is unlikely to cover all edge cases at this stage and the actual implementation can look different.

3, Validators should ideally be able to add multiple bids. We decided to omit this feature for now in order to reduce complexity.

8 Likes

Hey! Thanks for the detailed write up! Really appreciate it. I’m not the biggest fan of the current delegation strategy and welcome changes to it. I think the ideas behind this are more ideal, however I have some questions and points of feedback.

  • “yield”: Can you go a little more into detail on how you planning on calculating this? For something as literal as “yield” is, I think in Solana staking, this can actually be quite subjective. For example, it seems like MEV is included. Is that going to be an average over multiple epochs since it can be quite variable? “Yield” can mean a lot of different things these days and I think it’s incredibly important to define it. Validators can consider “yield” a combination of different things, including single-validator LSTs, and it needs a very concrete definition IMO. I would not advise to get into the games of yield calculation of single-validator LSTs and I don’t think you were planning to, but just calling that out as well. It’d be good to see a formula for this when you get around to it.
  • Bidding: Can you go into more detail on how you planned to do this? As a validator, how would I actually go about bidding? Would this process be on chain?
  • Gamification: If MEV is included, i’m slightly worried about the potential to “wash trade.” That could be an overuse of the word but there’s a potential to inflate your MEV rewards to make a validator’s APY higher to gain more stake from Marinade. I’m not an expert on the Jito stack, but a validator could in theory send a large tip to inflate their APY when they are the leader and “wash it” to get a higher realized_yield.
  • Superminority: Would validators in the superminority be eligible?

What I like about the ideas about this DS vs the current one is it encourages more competition, which I think is a good thing. But I can’t help but feel it’s a bit overly complicated. Reasons:

  • The change would require validators to pay constant attention to their bids. “Yield” is not constant and it’s something that would need to be monitored by validators frequently.
  • Calculating validator profitability is not an easy thing to do, especially with Marinade. Firstly, you have to calculate costs in fiat (server costs) and voting costs (SOL). Secondly, a validator’s cost would not be fixed either. A validator might have to move to remain eligible because of stake concentration and their costs could change drastically. Again, this would need to be monitored pretty frequently by validators.
  • The changes seem to be in the spirit of increasing return to stakers which I also think is a good thing. But I can’t help but think the same thing can be accomplished in a far simpler model. The DS could just stake to the highest yielding validators with some criteria in there to limit the possibility of commission rugs or downtime as much as possible. PSR already does a good job of limiting the impact of a downtime event to stakers. I’m not sure why the proposed DS would do a better job of generating a higher return than staking to the highest yielding validators with some criteria in there that a validator has to have been running for ~X epochs (a good value of X might be 50 as an example).

This strategy guarantees that as long as you are in the delegation set, you will be profitable.

Is this really true? There are probably strategies here where a validator could bid an amount to NOT be profitable on the Marinade stake, but still be profitable overall. This could present interesting sybil attacks and also squeeze out smaller validators.

4 Likes

Hey @maxkaplan ! Thank you for your thoughtful comments. Here go my answers:

How is yield defined?
Yield is defined as annualized return to stakers in the last epoch. It is the sum of inflation, MEV, block rewards, and bidding that validator gave to staker last epoch. Staker wants maximum yield (ensured by validator competition) irrespective of whether it comes from inflation, MEV, block rewards or bidding. Two examples of the same yield:

  • inflation (7.5% APY), MEV (0.1% APY), block (0.1% APY), bidding (0% APY) = 7.7% APY yield to stakers
  • inflation (0% APY), MEV (0.1% APY), block (0% APY), bidding (7.6% APY) = 7.7% APY yield to stakers

This definition of yield ensures alignment with what stakers want (maximum APY, together with yield stability through PSR) and maximum inclusivity for all Solana validators, irrespective on what the commission conditions for the rest of (non-Marinade) stake are like for the validator.

I unfortunately didn’t explain it properly in the proposal, but how bidding works is that it’s not constant every epoch, but it varies so that validator returns the realized_yield each epoch to staker. E.g. if MEV and block rewards were low this epoch, validator is charged more from bidding so that realized_yield is returned exactly, and vice versa – if MEV and block rewards were higher, validator is charged less from bidding this epoch.

Why is the bidding this complex and can the same yield be achieved through a simpler mechanism?
The bidding is essentially a hacky way how to achieve what isn’t otherwise possible: specifying different commissions on different sources of stake. It would be ideal if validators could specify different commissions on different type of stake, e.g.:

  • On my own stake, I want 100% inflation, 100% MEV and 100% block rewards.
  • On external non-Marinade stake, I want for example 0% inflation commission and 100% MEV and 100% block rewards commission.

If such conditions were possible to specify, it would make the Marinade stake auction marketplace very simple. Marinade could enforce 0% MEV and block rewards commission through eligibility criteria (so that it is not validator’s job to estimate future MEV and block rewards and this variability is just moved to stakers’ yield directly), and the only variable for scoring validators would become (potentially negative) inflation commission. In this way, we don’t even need to use the notion of bidding! Easiest to use UX and best profitability predictability for validators (MEV and block rewards is not a factor anymore) would be achieved.

Unfortunately, such splits per type of stake are not possible so bidding had to be introduced to allow all validators to participate in the auction irrespective of their inflation, MEV or block reward commissions on all their stake.

Is validator in the delegation set always profitable?
Since future MEV and block rewards are not predictable, it is validator’s job to best estimate them and place max_yield accordingly. If validator overestimates MEV and block rewards, sets max_yield too high, it can become unprofitable while in the delegation set. However, as explained in the previous part, I believe it is impossible to work around this problem unless all validators are forced to have the identical MEV and block rewards commission, which heavily limits validator flexibility and inclusivity.

Superminority: Would validators in the superminority be eligible?
In the current version they are, but we expect most validators in the delegation set will be small.

2 Likes

Thanks a lot for these answers! This clears a lot up for me. I understand much more now what you are trying to do.

The idea is quite interesting. I like the spirit of the proposed DS a lot better than the one currently implemented. It makes things much more competitive and makes the product more appealing for stakers, which I think is a good thing for Marinade.

I don’t have major issues with it, and if it gets implemented, I’m very interested to see how it plays out. In a perfect world, I think it can go quite well. The main worry I have is just the overall complexity for validators. Some great tooling / docs will likely have to be built so validators can manage their bids. I clearly haven’t put in as much time thinking about it as you have, but I almost foresee interested validators needing bots to manage their bids.

Thanks a lot for sharing this interesting idea and for answering my questions!

3 Likes

Thanks for a very detailed proposal. I agree with @maxkaplan that parsing it requires a lot of brain cycles. As I was going through it, I was wondering whether it could be simplified.

The first thing that caught my eye is the “decentralization bonus”. Why have an extra decentralization strategy, if the main strategy already has dispersion (ASO, country) limits baked in?

The decentralization bonus is a whole allocatoin branch that stakeholders have to think about in their mental models.

The new strategy seems to be strongly focused on building the best in class product for stakers to attract a higher TVL and distribute it to validators. I would be in favor of simply removing the decentralization bonus and re-allocating it’s TVL proportion to the performance-based allocation and increase staker yield as much as possible.

My 2c.

3 Likes

Thanks for the detailed proposal, though unfortunately I don’t think it’s the right path for Marinade to go.

Marinade used to have an easy to understand scoring that pioneered algo staking, however it has now become an increasingly complex and convoluted delegation ecosystem (dashboard, playgrounds, gauges, etc). Further complicating this with the need for validators to now bid for stake begins to make Marinade a less attractive option to validators.

I’d like to see Marinade return to a simplified formula that is data driven but easy to understand and returns it to being the king of algo stake pools, rather than subject to internal decision making or simple equal-weight staking.

EDIT: I realize this speaks to some degree mostly from a validator perspective, from a staker perspective I understand the need to protect returns but at the same time I’d encourage comparing with other pools and what returns they achieve for similar or larger validator sets. It is only the delta from that return that is essentially the “value” of these complex strategies.

3 Likes

This proposal opens up an opportunity for private MEV mempool participants to buy their way into Marinade delegation.
With the earnings such private mempools offer, mempool participants can share 100% of their inflation+block+MEV-bundle rewards using bidding mechanism and probably even some mempool side rewards on the top, while still staying profitable.
This could significantly disadvantage validators who choose not to work with these mempools and could increase exposure of Solana users to sandwich attacks.
In addition to that, the proposal would decrease MNDE utility and value.

2 Likes

Hello team! Hope you are doing well!

I’d like to share my perspective on why I have concerns about the proposed delegation strategy.

  1. It is already impossible to get Marinade’s delegation without setting up a 0% commission. The proposed solution will result in validators giving away at the very least 60-80% of identity+MEV rewards in addition to their commission, which is a huge surge in current rewards. (Unless I am missing something). Moreover there always will be the risk of your deposit slashing which will make your journey with Marinade completely unprofitable and unpredictable.

  2. All the Marinade achieves with this would be + 1-1.5% APR for stakers. Do you expect to increase your TVL with such a promo: «We improved our delegation strategy boosting your APR from 7% to 8%!»? Stakers are unlikely to ever notice such a small change in APR, but validators will certainly feel the impact as profits will be down significantly.

  3. With the strategy «Pay for the stake», you completely demotivate validators to work in your referral programs. As far as I understand, this approach leaves no room to engage with MNDE and Directed Stake referral programs, drastically reducing incentives. MNDE bonus is down to 10% and there is no Direct stake bonus at all. In my opinion, this is bad, since these 2, with a few adjustments, could bring a lot of external stake to Marinade through validator engagement.

  4. I think the proposed delegation strategy primarily favors larger validators, enabling them to outbid smaller competitors consistently. In the end, we will have a set of 50 larger vals with 200,000 stakes, who forwarded 100% of their rewards to stakers, just to not let new validators in the game.

Therefore, the proposed delegation program has a chance to boost network centralization even more. I strongly believe this strategy is the opposite of what Marinade is trying to achieve.

2 Likes

I wanted to contribute my perspective on this matter, but this comment sums it all up.
Stake pools should be fairly easy to understand.
What’s the point of creating a pool meant to assist validators in bootstrapping if no one can understand it? (or at least the one who need the bootstrap)
Overall, I am in favor of making the pool more accessible and more “liquid” in terms of restaking, but not as suggested above.

I think what’s being proposed is the simplest permissionless system that provides the highest and most stable yield to stakers on Solana. Would love to hear your thoughts on simplifying it!

This is a very good point and thanks for bringing it up! Marinade should indeed detect and disqualify private MEV mempools from bidding auction.

To me it sounds more like there is a huge value for stakers with the 100% downtime protection already in place + the best APY you can get on the market with introduction of bidding. And yes, I agree. If Marinade wants to position as a market leader with highest yield for stakers, it will definitely result in lower margin for validators, but I assume not the negative one.

I’m not saying that private mempool participant should be disqualified. Such activity is very hard to detect and there might be a lot of false positives. There is also no concrete proof that all private mempools are actually participating in sandwich attacks.
Just flagging that this economic model might disadvantage validators without a mempool revenue stream, potentially harming decentralization.

My suggestion for improving stakers returns and decentralization are 2 simple steps:

  • lower Marinade fee by 1%
  • lower maximum stake to a single validator to 3% of Marinade TVL
1 Like

The recently proposed changes to Marinade’s delegation strategy is very concerning to me.

The LST market is undoubtedly very competitive, especially with some pools trending towards the de facto standard of 0% commission only, it’s understandable that Marinade wants better APY for mSOL to stand a chance. Several parameters affect the APY of an LST, the most prominent ones includes inflation/MEV commission (lower is better), additional revenue sharing (higher is better) and pool management fees (lower is better).

Regarding management fees, jitoSOL has 4%, bSOL has 5%, most single validator pools are able to do <3%, mSOL has the highest among them all at 6%.

Marinade’s directed staked validators’ APY are lower according to the published numbers, however it does have its merit for retaining TVL that may otherwise flow to self-managed native staking or other LSTs since those stakers would receive kickback from directing stake to certain validators, it’s a mutually beneficial relationship. The removal of mSOL directed staking severs those relationships and will ultimately, decrease DAO revenue due to decreased TVL.

Slashing MNDE’s stake direction power by 50% also significantly decreases MNDE’s utility thus its value. With the recent inflation of MNDE from the reward seasons, the damage to MNDE’s value will be compounded.

What this proposal truly achieves is to encourage toxic competition - pushing both inflation and MEV commission to 0 and try to take a large share of validators’ block rewards, as well as revenue from SFDP stake, reducing validators’ profitability to 0. This just encourages people who can run a large number of nodes to compete for stake (economy of scale at work here), and possibly gather that stake for nefarious purposes, such as running private mempools for sandwiching (was approached by a few of those pools - turned them all down). Forcing validators to truly race to 0 also sets a very bad precedent since it may inspire other pools to follow suit as an attempt to stay competitive, while validators are the ones to suffer. It is also noteworthy that Marinade’s attempt to utilise SFDP stake to increase mSOL’s APY should be brought to the Foundation’s attention so that they could intervene if deemed necessary.

Additionally, even after all those shenanigans there’s no guarantee that mSOL will actually benefit from that due to the higher management fees.

Now if marinade wants to solve the APY issue, there are a few less drastic ways to achieve it:

  • Limit MEV commission, that directly increases mSOL’s APY and the effects are visible almost immediately (this does have the drawback of denying non Jito validators from Marinade stake, but if mSOL want to compete with jitoSOL in terms of APY, then high MEV revenue is a must);
  • Limit the amount of stake directed per sol, that encourages new participants to stake direction and allows it to achieve its full potential;
  • Lower fees, for obvious reasons
1 Like

Hi Robe! How are you?

I agree that psr mechanism introduced recently has a huge value for stakers. They must be protected from potential harm caused by malicious or underperforming validators.

At the same time, could you please explain to me why you think that adding 1-1.5% of APY will also become a «huge» value for Marinade stakers? Marinade is already the largest liquid stake pool in Solana. Stakers are choosing it because it is relatively easy to use, and they trust their funds will be safe for the staking period.

Can you provide any stats or cases when adding up such a small % of APR, led LSP, or similar DAO to the rapid growth of total staked assets? Do you believe that Solana stakers currently tend to choose Marinade less than some other competitors only because of the 1% APR difference?

I strongly believe that validators will lose far more than both stakers/Marinade will gain from this delegation policy. My point is that, by squeezing additional APR from validators’ identity+mev rewards and lowering their margin as a result, Marinade will directly affect the number of projects that validators are building around Marinade’s ecosystem. Validators will neither be motivated (the new program doesn’t assume it) nor have any money in circulation to build around mSol or contribute to direct stake attraction campaigns.

I believe that sustained growth and a robust ecosystem are essential factors in driving Marinade’s TVL over the long term. Lowering the validator’s margin will harm one of the ecosystem’s most crucial contributors. It risks alienating those who are deeply invested in Marinade’s success and are eager to grow alongside it.

In my view, this delegation policy isn’t the right path forward for Marinade.

Our 2 cents.

TLDR; Our team expresses negative feedback on current proposal.

Since we are Validator it is pretty clear that we will share feedback from that point of you rather than Stakers.

So far “race to bottom” process was limitted by 0% commission - where all (or almost all) validators who would like to get some fraction of delegation from Marinade should set 0% commission fee.

Current proposal opens “can with worms” where 0% commission is no longer a bottom line, and now Validators are in the position where they have to share even more revenue, i.e. to share their block rewards or/and MEV rewards with Stakers.

P.S.
We are a honest Validator and waiting for Delegation from Marinade since early January, but for some reason did not get any SOL so far.

As you mention in your proposal, there are a few constituents here: stakers, validators, Marinade, and the network itself. In equilibrium and assuming a homogenous markets, suppliers will bid away all of their economic returns at the benefit of the consumer. Auctions are a way of reaching equilibrium, so this proposal is good for stakers and bad for validators. This likely colors some of the above dissent.

The first question is whether it’s okay for mSOL to be better for stakers and worse for validators. I think it is: stakers are the ultimate lifeblood of Marinade. No stake, no Marinade. Most internet platforms are demand-side: it’s more important to deliver benefits for consumers than it is to deliver benefits to suppliers. Uber, for example, is notoriously bad for drivers. Juno, on the other hand, tried to be more driver friendly: giving higher rates, for example. Uber is currently the market leader and Juno is dead.

The second question is whether it’s worth doing this. Here, it becomes much more iffy. As others have mentioned, most stakers don’t seem to care much about a 1% difference in yield. Evidence of that is Marginfi’s $LST, which is yield-optimized (e.g., no Marginfi commission) but only has $87MM of TVL, compared to Marinade’s $1.4B. So maybe the juice isn’t worth the squeeze here.

The third question is whether this has any negative impacts on the Solana network. I think the answer is yes. The constraints you’ve listed would help with decentralization even in an efficient market, but not much IMO. For example, imagine that you depend on Marinade’s stake to run your validator and you’re in ASO xyz. ASO xyz is currently at the 20% mark. Then one of the other people in your ASO expands the stake that they can take, so they take your stake and you are forced to shut down. The main problem with constraints is that they are binary. If you’re at a constraint boundary, it’s difficult to know if you’re going to be left with any stake next epoch. This uncertainty may impede validators from participating. I wonder if you could smooth it out by either removing the constraints or making them looser and adding ‘decentralization-adjusted yield.’ So there’s an algorithm that determines how much a validator helps decentralization, which gets added to the yield in the auction. So if you bid 8% but Marinade gives you a 1% decentralization boost, it’d be like you had a 9% bid.

But yeah, no definitive thoughts on whether this proposal would be good or bad. I lean towards bad, unless revised, for the reasons mentioned above.

1 Like

Hey everyone. I’d like to address most of the concerns I see here.

1, This proposal incentivizes validators to run private mempools

I agree we should prohibit validators to extract users’ value from sandwiching and other malicious forms of MEV. We should ideally add this to the eligibility criteria. There are currently tools being built to detect such activites which Marinade could use.

2, Marinade has high management fee (6%)

As written in the proposal, the primary mechanism for long-term value accrual to MNDE should be a low fee derived from the yield realized by stakers through Marinade. I also believe this should be the only form of revenue to MNDE eventually, therefore I propose to reduce the mgmt fee to 0%.

3, Stakers will not notice the extra 1.5% yield anyways

Yield isn’t the only differentiator on the staking market, but it is one of the most important factors. Other factors are e.g. yield stability, ease of use, security, reporting, decentralization. This strategy will not increase yield by some small amount temporarily, but it provides a mechanism to make sure that Marinade will always offer one of the best yields on the market. Other Marinade features will be 100% uptime protection, ease of use, strongest security, great reporting and sufficient decentralization to improve Solana’s liveness and censorship resistance. I believe this combination will make Marinade v2 the best staking solution on Solana which will dramatically increase the TVL and thus increasing the value of MNDE.

4, The strategy is way too extractive from validators

I highly disagree with this framing. Marinade is only adding revenue and increasing overall profit of validators in the delegation set. The proposed strategy’s aims to make every validator in the delegation set profitable. Additionally, the strategy doesn’t introduce any non-refundable costs when the validator behaves as expected. I believe Marinade simply has to prioritize stakers in order to offer the best product on the market and grow the TVL which subsequently grows the value of MNDE.

1 Like

Greetings from the Deanslist crew on IslandDAO! :desert_island:

We took some time as a group here & had some time to study and discuss this proposal and the v. interesting comments below it.

TLDR on marinade proposal:

Too much complexity and extra workload for validators to give more yield to the staker.

This new DS ofc benefits the stakers and this is very positive, sth astutely commented by @metaproph3t .

Sth important to note is that one unforeseen/inadvertent effect could be that this new delegation strategy ends up promoting centralization by pushing out validators that do not meet the constraint sets. ie you allow the big ones w the most resorted anyway to stay and dominate.

Also very negative that this new proposal slashes the $MNDE voting power on DS.

We also love how this comment by @fisiroky nails it here:

That being said, we love the Marinade team so much that we trust your judgement and we’re hesitant to vote no.

1 Like

mSOL is the best LST in Defi by outreach, so it has an advantage. In regards to TVL, well, we are seeing already some tendency downwards, right? So something needs to change. All Validators benefit from hight TVL, so we should think about it as an important metric.

I appreciate Marinade showing skin in the game and proposing the 0% fee.

I would vote yes for this proposal in Realms, if the Management fee of 0% gets included in the proposal and if the MNDE allocation of stake remains at 20% and not 10%.

1 Like