Pre-Proposal: Decentralizing Network Stake via Parameter Revision(s)

NOTE:
For the purpose of this discussion, any numbers below ignore the addition of a couple of new ‘large’ Indexers who cropped up in the past day, as the effects of their presence has not yet had much chance to play out with stakeholders. These are:
0x85fe868adf7f5950b052469075556fb207e5372d
0x38e6b1dc04087f69e69f10d7318643accc9854f5


Introduction

After watching the network grow during the current bootstrap phase for the past 2 months and observing the behaviours of the various stakeholders and members of the Graph community, there’s been a growing indication that we may be hurtling towards a very centralized distribution of GRT across the network.

There’s many reasons and variables that can be attributed to this issue. It may be partially due to how Delegators are behaving and the metrics they are choosing when picking an Indexer. It may be how those Indexers are behaving, the broad gap between cuts/APY on offer, or the amount of marketing/promotion they do.

However, I feel that some core parameters within the network are also contributing to this pattern, and maybe even guilty of steering the behaviours of both parties.

Those parameters being - “Delegation Ratio” and the lack of an “Overdelegation” cap.


Issue #1 - Overdelegation Cap

I’ll cover this one first, as it’s quite simple. The lack of a cap on ‘overdelegation’ allows Delegators to continue delegating to an Indexer that is already above capacity (on their 16x ratio).

This is not a common behaviour, yet, but it does happen. And the reason it may happen, or more importantly why it’s likely to happen more in the future, is because that if an Indexer has a low enough fee, the APY on offer from a Delegator’s perspective can still be very competitive, to a certain extent. This ‘extent’ grows along with the size of an Indexer’s stake, as it then takes more GRT to make an impact in terms of reward dilution.

The ability to overdelegate in the first place, from what i gather, was to allow some kind of free market behaviour. Though the ‘vision’ as i understand it was never for this to become some common tactic as described above, though it may.

Personally, i feel it serves very little purpose, but it’s a smaller aspect of this issue, overall.

Issue #2 - Delegation Ratio

Currently, we can see there’s quite a number of extremely large Indexers/stakes on the network.

The current (16x) delegation ratio means that even the top 3 by stake weight could already accommodate 50% of the fully diluted (10bil) GRT supply, before even worrying about overdelegation/dilution of APY.

There seems to be very little purpose in even allowing this amount of capacity in the first place, because while it may never happen, there’s no scenario in which we should want half of the network to be able to be held under only 3 nodes.

And while a low cut/high APY seems to be what draws in the majority of delegations, a large stake (and possibly brand?) also seems to help get you noticed, with the top 3 by ‘delegated amount’ also being amongst the top 5 by ‘stake weight’.

It seems that delegations trend towards large stakes and low cuts, and hence we see approximately 57% of the current (1.7bil) delegation weight, across those 3 Indexers.

Summary of the Issues

Due to how ratio works, being able to attract delegations allows you to lower your ‘cut’, while still pulling in the same ‘effective cut’ (fee) as you grow. At large scale, this effective % translates in to large amounts of GRT, while overheads remain quite linear.*

*This may change in the future when large stakes need to keep up with the query load sent their way, but the amount of traffic required to offset the incoming rewards would be substantial, to say the least.

And when delegations are concentrating amongst the few, it’s not possible for the many to continue lowering their cut (to be competitive) while remaining profitable. And I believe that if this continues and there becomes only a marginal difference between running a ‘small’ indexer with a low amount of delegation, or just becoming a Delegator yourself, then we could eventually see many smaller Indexers switch sides.

In addition, the dilution of rewards received by a Delegator when their Indexer is overdelegated, requires a lot more in terms of GRT to make an impact against a large stake weight. This can mean that a large stake with a low fee remains a preferable option in terms of APY well beyond 100% capacity when compared to what smaller Indexers could offer (while dealing with the same overheads as others). This only further exacerbates the problems faced by smaller Indexers.

All of the above seems like it has and may continue to centralize the stake on the network. This is clearly not an ideal situation for anyone who cares about the network health in the long-term.


Intended Outcomes

By highlighting the current situation and where we seem to be headed, i’m hoping to start a discussion that will hopefully lead us towards a proposal/solution in the near future. I say “near”, as i feel this is already quite a critical issue that is snowballing fast and should be addressed sooner rather than later. With some of the unlocks coming in the near-term, we could see many more Indexers with multi-billion capacities appearing.

Making changes to steer the network towards a more broadly distributed set of delegations would ensure that smaller Indexers (in context - the vast majority) have a more likely opportunity to remain involved in the network, enabling healthy competition (also a positive for Delegators) and ensuring a more likely outcome of true decentralization of the network.

Delegators could still enjoy attractive APYs, but this would instead be provided across a larger set of Indexers through the ‘trickle down’ effect of network stake being passed down.

Feedback, criticism, and further suggestions are all welcome!



Please note:
I’ve been over a lot of numbers and projections with many people regarding this topic, but have not included them initially, as i was interested in getting this discussion started ASAP. We can of course discuss numbers, projections and expectations during the course of the thread. Any contributions of this nature are also highly appreciated!

19 Likes

My personal opinion on how we might solve these issues, to get the discussion rolling:

  • Revise ‘overdelegation’ in general.
  • Implement a ‘hard’ delegation cap.
  • Revise the delegation ratio.

Overdelegation

Remove the ability to overdelegate, or at least set it extremely low. Is there any (network) benefit to this feature? There’s a concern that Delegators could be left with nowhere to go if all Indexers were ‘full’. However, we’re very much far away from worrying about that scenario, and the likelihood of that playing out in the future would depend on how other parameters (below) are balanced.

Hard Cap

Some numbers I’ve discussed with others have been along the lines of 200mil being a good ‘hard’ cap, for the purpose of initial consideration/discussion.

Whichever number is discussed here, it should neither be too low to remain profitable to smaller Indexers (EG being drowned by ‘gas’, etc), nor too high as to centralize whatever % of the supply we may expect to be actively delegated, in the future.

And while large stakes could simply spin up additional nodes, despite any cap, they will see diminishing returns in their revenues as their expenditure on gas will increase in a basically linear fashion, even if their infra or other opex costs do not.

Delegation Ratio

Some initial thoughts on this topic were that lowering the Delegation Ratio would be the answer to the issues. However, if a hard cap is implemented, then there’s no further reason to lower it and doing so may only really hurt smaller indexers. If anything, it could even be increased. Maybe rounded up to 20x?

The minimum cap for an Indexer is 100k GRT, but at such low stake amounts, they will need to be able to onboard a lot of delegation weight in order to cover their future opex.

20x, purely as an example, would allow them to onboard up to 2mil GRT. While still incentivizing all Indexers to increase their own stake, up to at least 10mil GRT, to maximize their delegation capacity and onboard more delegations.

8 Likes

Anyone out there who has a talent for modelling, your input would be highly valued as we explore ideas that are suggested here. For me, the most important factor beyond the discussion is understanding the second order effects of making such changes. Thanks @Fattox for taking the lead on this.

2 Likes

Great thread! Large indexers are a threat to the ecosystem and the decentralization of the project. I think you laid out a few good options to mitigate this, however, it really comes back to the balance between indexing rewards and query performance. It will be interesting to see if these large indexers will actually have an interest in indexing many subgraphs and deliver great query performance or only to reap indexing rewards on some select few subgraphs. We’re still early in this process. The game of curation is not even here yet.

1 Like

Great write up. :slight_smile:

How would this be “cleanly” implemented given the .5% tax on new delegations?

If a 200M GRT cap gets implemented, which GRT delegators would be forced to choose a new indexer to delegate to and would they still pay a tax for redelegating?

Using this indexer as an example: 0x5a8904be09625965d9aec4bffd30d853438a053e

It has 350M more GRT than proposed cap. 17.5M GRT would be burned or $35M worth of GRT on just this indexer alone if all of those GRT delegators redelegated to another indexer.

3 Likes

A few things to consider:

  1. I think it would be unwise to implement something like this retroactively, but if there is no other option, you need to consider how to do so while being fair to all indexers and delegators already in the network.

  2. I think a hard cap on new initial indexer stakes, and a hard cap on delegation ratios for new delegations, would be reasonable and solve some of these issues. But the network probably needs to allow for overdelegation of existing indexer stakes and delegation ratios because rewards are automatically restaked/redelegated. Further, if some indexers withdraw their rewards (when eventually allowed) but their delegator rewards continue to be automatically redelegated, that will naturally lead to overdelegation over time.

2 Likes

Great topic and definitely worthy of discussion as this problem looms over the network.

Just thought I’d throw out there the suggestion of a delegation ratio that scales according to indexer-owned stake. The smaller the stake the more delegation that can be accepted.

Perhaps something like:

Indexer-owned: 100k → 1m → 5m → 20m → 50m → 100m
Delegator stake: 20x → 16x → 12x → 8x → 4x → 1x

Wait, how is this equitable IN ANY WAY SHAPE OR FORM? Since indexers make money by charging delegators a cut, why in the world would indexers who have staked the least capital be entitled to a larger potential cut of rewards?

Like, seriously, what the heck?

As a delegator (you can choose whether or not my opinion remains valid in your eyes after this disclosure) I am absolutely sick and tired of this constant stream of proposals that purport to “fix a problem with the system” that don’t spend a paragraph or two addressing potential unintended consequences of implementation, or even support their claims of a “big problem” with much reasoning behind it. Is it possible that the system is working more or less as intended, and that some indexers overplayed their own hand, when it comes to how they’ve chosen to deploy their capital?

The answer to the “centralization problem,” such that it is even a problem at all (the proposal doesn’t actually address what issues would arise from some indexers being larger than others) is not to enact protectionist policies for the indexers who have staked the least, it is to entice more large entities to become indexers.

All of this reeks of “delegators delgate to the wrong indexers because they are dumb.” This is absolutely laughable. Delegators are smart enough to realize that until query fees are a larger percentage of the rewards on offer, that it is in our best interest to back indexers who are not claiming they are about to run out of capital to pay their expenses.

It seems to me that smaller indexers (less than 1m in staked GRT) have put very little on the line, and ought to consider becoming delegators instead if they cannot afford the overhead requirements of their position. Why would I ever let someone who is crying poor become a custodian of my GRT?

This constant stream of fixes for things that are not broke is incredibly tiresome. Over the last few days, a large player has entered the scene and staked 140m GRT. What is this argument that indexing is somehow not profitable? Perhaps it isn’t for you, but it is for someone, otherwsie they wouldn’t enter the fray. I do not hold enough GRT to be a profitable indexer, so you know what I do? I delegate. I don’t cry foul that the system is designed in such a way that doesn’t perfectly suit my situation.

Indexer entitlement is gross. If it isn’t working out for you, go find another decentralized network to participate in that does. I for one suspect there will be no shortage of entities that would like to take 10% of my rewards, some of which can even afford the investment cost of a year’s worth of operating expenses.

If this issue boils down to “delegators are too dumb to do basic math” then let me first say, F*** THAT, and also let me say, perhaps indexers should have done their own ROI calculations when they decided to stake their capital into this network.

Seriously, how can anyone say with a straight face that, 2 months in, the system is flawed. That is such bad process I feel I shouldn’t be required to say much more.

I just, I just hope that the actual leadership of the Graph Council has a steadier hand than some of you guys. Thinking that you’ve spotted some fatal flaw in the system while it is literally 2 months into the rollout phase is ridiculous. Now is not the time for course corrections, now is the time to continue the work of migrating subgraphs to mainnet. An index is an investment, and if you’re overextended I’ll give you the same advice I give first-time GRT purchases- do not overextend, give yourself plenty of wiggle room. If it is not possible to make your investment work while doing both of these things, it is not the correct investment vehicle for you.

I’m not suggesting this, I’m just throwing out an idea to even up the balance of delegation to promote a more decentralised network, in line with the original post.

I suggest you go and take a listen/watch of the latest “office hours” recording for more understanding of the problem that Fattox is addressing here. I don’t think you have an appreciation of the problem that is being tackled here, think of what would happen to the network if 90%+ of Indexers decided to pack up shop because delegation was more profitable for them to do with their tokens.

You do realise that Fattox (original poster), and cryptovestor are both council members, right?

2 Likes

Going to have to agree with derkhersh. It’s way too early to predict how incentives will play out for delegation behavior. Ideally the network will be decentralized, but I don’t think we can force that through parameters just for the sake of decentralization. Processing enough queries to attract anybody is going to require some serious machinery and indexers with the bare minimum might not even be able to keep up with that if they can’t generate the capital to draw in a larger delegation pool. Maybe just delegate for now until we see what query traffic and rewards will look like. I do wonder if the big pools are even interested in the querying portion of indexing or if they’re just camping the network for the inflationary rewards. But the health of the network would see them lose their pool if the query rewards start being worth it.
Basically it’s so early but that anybody delegating is just fishing high APY and feeling out the system. Once queries pick up maybe it starts to make sense to find smaller indexers or maybe new players jump into the mix. We shouldn’t artificially stack the deck in favor of those with the least skin in the game though. Not sure what the situation is for the indexers but perhaps switch to delegating until we get a clearer picture of things.

2 Likes

Actually I had an idea. What if query processing had an impact on the inflationary delegation rewards distributed to an indexer? That could help prevent large indexers from just camping the network, hogging delegators, while not really performing the actions of an indexer.
Some sort of structure that would expect that a certain size of indexer should perform an equitable portion of queries in the network in some given window and should be rewarded relative to that contribution. An indexer that just exists to be large while not processing an equitable amount of queries would get less rewards and people would naturally go to other indexers. It would likely also encourage smaller indexers without necessarily “punishing” an indexer for the crime of being large. Hardware limitations might make it hard to just have infinite money and get 99% of the network stake if you can’t actually process 99% of the queries.

Yes, I understand fattox’s position. I assume he speaks for himself and not the council writ large.

And I’m sorry, I promise I don’t mean to direct all my ire at you in particular, but if all the actual discussion regarding reasons/consequences of proposals happens elsewhere, what is the purpose of this particular forum’s existence? I had assumed this space was meant for public review/comment. If so, the idea that “we talked about some of this stuff elsewhere, where were you?” does not fly. A proper proposal must provide adequate summary of the problem it wishes to address, as well as a statement addressing the potential negative impacts of its implementation-- regardless of the status of the one proposing.

I reject the notion that 90% of the indexers would pack up shop. But lets say that happened. What percentage of staked GRT would that amount to? Certainly much less than 90%. Further, if they turned to delegation, that GRT would remain in the network, so, would the network suffer, or would those particular indexers just lose out on a bit of extra ROI?
Further, it’s a bit near sighted to suggest that if a wave of underfunded indexers were to drop out that it would preclude new, better funded players from entering the fold.

You guys don’t even have query fees generating yet and you want to dramatically change the incentives. It’s been 2 months. In the context of 28 day cooling periods for me to withdraw my funds- a measure that impacts me both positively and negatively, a measure that was clearly advertised and I made a personal financial calculation to accept- proposing big changes after 2 months seems… quick triggered.

It would also seem to me that there’s a fundamental misunderstanding as to the goal of the network, at least among some indexers. The goal of the network is to serve data queries, and to distribute that load efficiently. Decentralization as a goal is NOT the same as “the more indexers the better, regardless of their financial ability to stake themselves.” If queries can be served efficiently with only 25 indexers, so be it. If you think that’s at all likely, you aren’t that bullish on The Graph’s use potential. Chances are there will be many, many willing to put up with short term capital outlays in order to secure themselves a piece of this network.

Delegators are a piece of that, in the sense that we vote with our feet. Start generating query fees and we might start seeing a reason to back someone who isn’t compounding our rewards daily / able to offer adequate assurances that we won’t need to undelegate sooner than we’d like to.

By constantly crying poor, the small indexers are shooting themselves in the foot. Delegators talk, and we talk about how obnoxious it is / how disconcerting it would be to delegate to these folks.

1 Like

Agreed, the original post has been posited as a “pre-proposal” to invoke discussion on the topic, so thanks for engaging and expressing your opinion. I think the original post adequately describes the issues at hand well enough to have this discourse.

I don’t think such a rule would really be necessary. As a delegator, I’m only going to get my fair share of rewards, less the indexers fee. So the indexer being large won’t net me anything extra in return.

Once query fees account for a generous portion of the return potential for delegating, I for one plan on switching from my big-box indexer to whoever is generating the best query traffic. I suspect other delegators will make a similar decision. If my big-box indexer winds up generating good query traffic, I’ll stay with them, but in this scenario they can’t be accused of camping.

Market forces should make this happen without any artificial engineering. The primary focus of anyone with any technical ability should not be changing the incentive structure, it should be getting mainnet up and running. This is misplaced energy, you know?

Really, the only reason I can see for needing to implement new policies is that “delegators are too dumb to figure this out and back the right people” and I just really resent the implication.

Hi @Derkhersh, thanks for participating in this debate.

I think your position definitely deserves to be voiced, but I would ask you to please be more respectful in future posts. The same points can be gotten across without the dramatics. Everybody who’s here is here to try to make the network successful. There will be a lot of decisions to make through governance and it will be a more enjoyable ride if everybody is kind to each other and focuses on the nature of their arguments.

Thanks!

6 Likes

Awesome, thanks for that! I appreciate the desire to understand the deal you’re getting into and not have things change right under your feet. I also appreciate the various proposals that are starting to come in. You can rest assured that any proposals will have plenty of time for discussion. No core protocol economics will be changed without a fair amount of deliberation, modeling, etc.

So this is all part of the process and there are plenty of people reading and processing these discussions with an open mind looking to be convinced. Thanks for taking part!

2 Likes

@Derkhersh I sure hope your post is a troll post because this line is so ironic it blew my mind. I, too, am a delegator and I recognize that the original intent of the protocol was that people who have no technical ability and/or the required initial capital to run an indexer to still be able to participate by delegating.

Now from my understanding, that means that delegators (people with low skin in the game) could make money off indexers (people or organizations with yuge skin in the game) with little effort. I participated in the testnet as an indexer and it’s clear to me that it’s pretty much a full-time job, especially on mainnet. As an indexer one assumes the infrastructure + maintenance costs, manpower costs, risk of being slashed, keeping up to date with new subgraphs and curator signal (in the near future), constant delegator community management, etc.

So to come round to what I first said, I’m a delegator because I did not want to assume the risks stated above. It saddens me that big players caused a race to the bottom in terms of fees and enabled delegators to be entitled, hence the irony in your post. From what I understood by attending Brandon’s talks, this was not the originally intended protocol dynamics, so it makes sense for people who actually care about the protocol and put hard work into making delegators money to initiate such proposals to ensure the protocol will not become a centralized mess; also, is it really that outrageous that someone cares about the little guys too? After all, without them, there isn’t a protocol to speak of.

I hope I didn’t come out as rude, but watching all this from the sidelines makes me wonder who’s really in control here.

4 Likes

What if there is only 3 McDonald’s in the world and you are forced to travel across the globe in order for a bite to eat. Don’t forget the consumer.

Don’t forget the original intent of the proposal - network decentralisation.

1 Like

Now this here is a fundamental difference in opinions - I will never be in favor of corporations using monopoly tactics to drive out the smaller businesses without them having even a fair chance. Also, in this case, the analogy doesn’t hold because you only need a couple of McDonald’s types in the network and it’s already massively centralized. What if the only McDonald’s in the world is in the USA but still it’s mailing $0.00001 burgers to any customer? Not in favor of big businesses undercutting then placing the blame on the small guys.

Going back to the beginning of your post - if indexers aren’t owed delegators, then why are delegators owed indexer rewards?

I understand you view this through a free market lens, it’s obvious that you’re only in it for the APY, and I don’t blame you for it. But maybe take into account that as much as we want to preach decentralization, this is a controlled environment in its infancy with parameters that can be tweaked over time in order to ensure the survival of the protocol.

1 Like

So if miners on Ethereum increased the gas limit on blocks, that’s not really decentralization?