I loved Itn and Cardano, but I may not love mainet because of a0!

I loved the Incentivized testnet. It was great because anyone could run a pool and validate some blocks, with chances proportional to their stakes,even with as few as 100kAda you’d get a block per month. I ran a pool, and I loved that, honestly it did not get me that much of earnings so far,
but I was very happy with the idea to participate directly to the security of the network. I was looking forward to mainet and a PoS crypto that we truly control.

But this may be coming to an end because I just discovered that in the mainet, big pools will apparently be artificially advantaged over small ones giving them higher RoS and ability to validate blocks. (Note that they were already because a big pool had more rewards for the similar operating costs, but I’m talking about something else). This means that we will artificially give more (power and money) to those who already have much, and less to those who already have little. This can only lead to centralization in the long run. Decentralized is not only about how many pools there are, it is also about the fact that anyone should have the potentiality to participate directly to the network in proportion to their means, no more, no less.

(See the more technical post about this: Understanding shelley reward formula (and k, a0 parameters) )

But why change the way itn worked, when switching to mainet? Because it was susceptible to Sybill attacks. Because if people delegate randomly their stake to pools, then someone could create lots of different pools and pretend to be different people. And then either hope that people will delegate randomly to their pools on their own (just because there are so many), or make little advertisement to attract people to their pools. If the attackant could then cheaply control 51% of stakes he would then control the whole network.

There are 2 possible strategies to avoid such sybill attacks (both can be applied at the same time):

1- Incentivize people not to delegate randomly:

a-Have them stake to a pool that they directly operate themselves
b-Have them stake to a pool operated by a friend or someone they know directly
c-Have them stake to a pool of someone they know indirectly but hard to fake : youtuber, university, association, well-established provider of service, well-know member of community, ect…

Of course for this first strategy to work you want to allow small pools to be at leat as competitive as big ones so that many people operate small pools on their own, and subsequently many people have friends who do so. But you want also to have an expected RoS which is pretty much the same for every correctly operated pool, so that the sole criteria for a user to choose a pool is what pool he trusts the most. If you introduce variable RoS then you steer people to delegate randomly to the one with higher RoS even if they have no idea about who operates it. So overall the chosen formula worsen things in that side.

2- Limit the ability of one actor to create many pools that can potentially become big, by limiting the amount of delegated stake you can get proportionally to the pledge/stake you directly own.

It would be very easy to enforce that for every ADA I pledge directly in my pool, I can have at most 2,4,5, 10 or 20 ADAs delegated. (It would not advantage big pools over small ones.)

What has been chosen instead may seem similar but it’s not: the amount of pledge would influence RoS of the stakepool, to give and Incentivize to have a large enough pledge, but not enforcing anything. This implies quite complex and unpredictable consequences, not very well know impact of a0 and k, and need to try and estimate all that with complicated models. The formula chosen also have the very nasty side effect to advantage big pools over small ones as I said before.

Also if you find problematic to enforce a limitation, note that k is already enforcing maximum stake the exact same way.

I may be wrong or miss something, if so please correct me. If this is truly indispensable that big pools be advantaged over small ones, to prevent sybill attacks, at least we should have good public proofs or evidence of that. Please show me that simply limiting the ability to receive delegated staked ADA proportionally to pledged ADA, is not a good idea.

If I’m right, I hope the reward formula will be changed, for the best.

28 Likes

Also it is to be noted the following thing: if users are incentivized to delegate according de pools RoS, how long is it going to take before a robot will be available to change automatically delegation according to optimal strategy; switching to higher RoS? Maybe there will even be a fork of Daedalus doing that. And if everyone starts using this robot/algorithm, then it will completely nuliffy the whole delegation system, because we could have calculated the exact same thing without people in the loop, by applying directly this algorithm.

This brings us to why do we give rewards to people who delegate in the first place? And the answer is to secure the network. And how do they do that? By choosing carefully a trustworthy stake-pool! (i.e that will not attack the network) And therefore they should be informed that this is their responsibility, and incentivized to do so, rather that incentivized to delegate randomly to higher RoS pools.

If we let happen this automatic delegation with robots and optimal strategy, then this will probably be the end of Cardano.

7 Likes

Interesting topic. Hope that we can have a good follow up on this.

2 Likes

Have you seen this episode from TCE where Lars, Duncan and Kevin explain in depth about pledge? This may answer some of your questions.

2 Likes

@ Katsumoto Yes I did, and it does not address the issues I talk about here.

Ps: To answer in the video space and have visibility I’ll probably try to make a video myself explaining my point in few days if none shows me it is nonsense since then.

But what I am seeing at the INT is that, at first with 400k, I got at least 1 or 2 slots a week with my pool MICRO.
Now I have (had) 1m-800k stake and for the last 5 weeks I got 1 slot. If I watch other pools they also are getting less and less slots. Probably because there are a lot more small pools.

And because little pools don’t get any slots they lose stake, that stake goes to the bigger pools that does get slots to produce. And that way you get centralization of the network.

So I am really looing foward to main-net shelley staking. The 100m stake saturation/pool in INT is way too much at the moment.

Ok you are talking about k value, I’m also in favor of a higher value (maybe around 300-400). But for now there are no guarantees that k will be increased in mainet. Also if you read my posts you will see that if you are a small pool, new reward formula will not advantage you in comparison to itn, on the contrary it will disadvantage you, and the more a0 is high the more small pool will be disadvantaged.

Yes I understand that the smaller pools will have disadvantage. But pledging more ADA can alter that a bit if I understand correctly.
But if smaller pools still got a chance to produce blocks because the k factor is higher, you can compete with larger pools because your stats are good (you can produce blocks) even if you just make one.
At the moment I produce 0 day after day and stake holders will not delagate because you get 0 slots for weeks so 0% ROS.

You can have a small pool with 100% of your stake as pledge it will have lower RoS than a big pool with 5% pledge, (and you will almost have an RoS almost identical than if you had 0 pledge in your small pool) . If you want to discuss it more in details the formula and its implications, please let’s do it in the technical post.

It depends delegators that know they delegate to a small pool know that they won’t necessarily get reward everyday, but when they get rewards they are much higher, so overall it does not change expected RoS. Personally it does not disturb me as long as there is at least one block per month let’s say. When you are lending money usually you don’t get interests daily so…

I can tell you that at some very early point someone delegated 70k to my pool and did not get reward for more than 15days, but when he did he had a very good overall RoS. It all comes back to why people delegate to you? Because they trust you or for higher RoS at a very specific moment?

But even pools that validate 1block a day will be significantly disadvantages by the new formula.

An interesting proposal is that instead of a0, an emission curve token model was implemented for each pool. With each pool having its own token and with its delegates receiving these tokens as rewards.

These tokens can be used to exchange products and services for each pool, marketed … among hundreds of utilities that can encourage a certain pool to create a transparent community so that its tokens have more value and their pools become more attractive.

The emission curve for these tokens can apply several incentive parameters, such as an emission curve that encourages delegating to smaller pools with promising projects that have great potential for valuing your token.

All of this can bring another level of possibilities to the ecosystem, such as encouraging the creation of DApps related to the pools, new ways of interacting with the treasure.

In short, the solution can be simply tokenize the pools and align incentives.

All of this is well studied in the personal token space and an extensive game theory behind, if many are interested, we can create a CIPs about.

1 Like

I believe this will naturally occur by pool using the multi-asset DSL. Pools could even issue a synthetic derivative so that stakers could access “Synthetic ada” specific to that pool, and use it as collateral within any defi dapp they wanted, all while still staking/earning on their ada amount in that pool.

Thank’s for the reply, though a little bit more explanation and arguments would be nice.

I what I say makes 0 sense don’t you think I and the community altogether should understand why and stop being that “ignorant”? ( 8 people liked my post I’m sure they will be interested as well). The point with crypto is that we should all agree on a protocol and know why we should agree on it, right? (And I’m quite sure to know how selection of slot leaders works, but maybe you can explain more to show I misunderstood).

1 Like

This is your comment I was replying to

if users are incentivized to delegate according de pools RoS, how long is it going to take before a robot will be available to change automatically delegation according to optimal strategy; switching to higher RoS? Maybe there will even be a fork of Daedalus doing that. And if everyone starts using this robot/algorithm, then it will completely nuliffy the whole delegation system, because we could have calculated the exact same thing without people in the loop, by applying directly this algorithm.

You completely ignored how staking/delegation/un-delegation works in Cardano to come up with this construed reality of a world where a bot could auto-swap my stake between pools. simply not how the selection process works (which I feel like you should know if you were a pool operator) but sure. Smarter people than you and I have thought about Cardano’s crypto incentives since 2015 and produced numerous papers. Each of your points on ao are answered in the recent TCE episode or were answered in this thread. This on the other hand is just pure speculation on a bot/algo that would not even work according to your argument in the Cardano network.

1 Like

I don’t understand how you hope that if different stake pool have different RoS and subsequently for a delagator to earn as much as possible it is usefull to re-assess decisions on a daily basis ( like was briefly mentioned in the cardano effect episode), people who have better to do than look at pools stats all day would not write a bot to do it for them…

Smarter people than you

Come on please… Are we really here for a competition of who’s the smartest and ego things, or to make sure Cardano succeeds? If the point of this discussion is ego, not sure it is worth going any further.

Each of your points on ao are answered in the recent TCE episode or were answered in this thread.

Absolutely not, as I said it does not answer any of my points. And if it did people can link to what timing addressed what point.

4 Likes

people who have better to do than look at pools stats all day would not write a bot to do it for them

Yes tell me how that bot that’s trying to predict a future synthetic lottery with data from the past will work when it takes 1-2 epochs to even re-delegate to a new pool. Also tell me why literally no one in Cardano (or any other PoS coin) has come up with this “Super Stake Algo-bot” yet? genuinely curious.

This has nothing to do with ego bro, we are all here for the long term success of Cardano. You are simply speculating on variables you clearly don’t fully understand, are being presented with information/rebuttals (TCE Pledge episode, the ao formula responses in this thread) but continue to belligerently hold onto your belief that a single variable (which is literally a data point decided by the median pool operators ada pledge) is going to push out all small pool operators. This is not the case, several clear explanations as to why this is not the case have been keenly articulated in TCE and this thread.

1 Like

Very simple: mainet is not live yet, and in testnet there is no significant difference in exepected RoS that would make the “Super Stake Algo-bot” very relevant. (And even testnet is less than 6 month old, so it may just come).

That’ll be all for me with you, because we do not have have a sane and constructive debat (and probably cannot). I’m not running away from any debat but there is no debat only authority arguments and affirmations that I am stupid and ignorant. I cannot answer to that.

Good bye, and I hope that this account you created few hours ago, will be usefull to something else than simply discrediting my posts! :wink:

2 Likes

interesting conversation here. good points on both sides :wink:

and, like the currently vocal, i am very much for NOT having a scenario where the matthew effect (rich get richer) becomes the predominant play. though i think there are those, like @philpa, who see things differently.

i’ve got two points i’d like to highlight.

the first:

this is a great point as one of the already existing mitigations, i’m curious to hear thoughts on how this affects the hypothetical eventuality proposed in the original post - and if confirmation bias perhaps is why it’s not brought up. have other currently existing mitigating factors been excluded to support an emotion based argument?

the second point is:
humans always look to optimise, so it won’t take long before machine learning and automation is applied to this area. so we can acknowledge it as an inevitability and act accordingly or the do the opposite.

personally, i don’t see how this would be any different to the self-serving behaviour we witnesses with the manifestation of the 1% pools diverging at places like Frankfurt. we have acknowledged the phenomenon and are now taking steps towards remedying it by changing the propagating time so that latency doesn’t become a thing that disproportionally advantaging some while disadvantaging others.

i see ML/automation the same exact way. we the community will look at it like we always have - as a catalysing opportunity to make the protocol and community even more robust!

keen to hear anyones views on this! :nerd_face:

2 Likes

This does not change anything to my point of automation! Even if changing pool would mean lost epochs (which it don’t think it will, will it?), then it’s just a cost parameter to take into account to calculate when it’s worse changing pools or not. And it does not change the fact that it’s worse re-evaluating that on a daily basis with latest data if you want to maximize your gains, and that a bot will do that very well, probably better than humans.

the second point is:
humans always look to optimise, so it won’t take long before machine learning and automation is applied to this area. so we can acknowledge it as an inevitability and act accordingly or the do the opposite.

Yes it is inevoidable if we design the protocol such as different pools have structurally different RoS, and then if you think about it, it subsequently voids the security of the PoS protocol. But we may decide that all well-operated pool should have the same expected RoS, which then would change the way delegators choose a pool to something more human.

ok, but you realise that all the analysis could happen in the world - but you can only get rewards at the end of an epoch for the pool that you have delegated to at the beginning and only the beginning of the epoch which happens every five days.

this is a very key part that you can’t just disregard, quote:

if you’re actually looking to have a reasonable rational conversation about this.

Yes I do. But I mean the data I get today will give me a great deal of information on what will be in 5 days. Anyway every delegators will have to make their delegation choice on data they have today, even if their choice is applied only in 5 days, and it does not mean they will make random choice disregarding today’s data.

For exemple if a pool increases significantly it’s pledge today, I’ve got good reasons to think that it will still be there in 5 days, so I have interest to take action today, in regard of this information even if my decisions is only applied in 5 days.