@maliky At least not that I know of. Only the ISP’s can really trace an IP, and then it might also be a fake IP forwarding the packets.
You should keep in mind that not the number of pools are threatening the protocol but the pools with the same entity gained power.
As Cardano can have thousands pools but the total (voting) power i.e. total stakes are still the same. I am happy to see lot of small pools arising but unhappy that they are not incentivised for having only one pool.
3 Likes
That’s true, the protocol should favor running only one pool, but there is no real way to stop anybody from cloning their pool, except proper community knowledge of who runs which stake pools, or am I missing something?
2 Likes
Shameless plug regarding pool splitting… An Alternative to a0 and k
2 Likes
I have read your proposal, but I do not think that the reward calculation function needs to be changed.
The issue seems to me relating to the reward distribution functions and not to the reward calculation one.
I’m afraid I don’t follow. What do you mean when you differentiate between a “reward calculation” and “reward distribution” function? From one of your earlier posts I gather you imply that the reward distribution function = reward calculation function - stake pool’s cut where stake pool’s cut = margin fee + per epoch fee(340 for most pools)?
Yeah, it’s a bit misleading cos I think I referred to reward distribution functions, as there are two, one for member and one for operator.
how it calculated (roughly)
- R the total available epoch reward is calculated from monetary policy, tx fees and decayed deposit (has not implemented yet) minus treasury taxation, then
- that R is split between the pools and considering apparent performance called actual reward i.e.
pf(s,p)=R_pool. - the pool’s actual reward is split between the operator and pool members. I call it reward distribution as
actual reward splittingis confusing /w R (total) reward splitting.
From Design spec 5.5
These calculations proceed in two steps: First, rewards are split amongst pools. Next, each
pool splits its share of R amongst its operator and its members.
If you interested I created some Video presentation for some SPOs on Sunday you can check it here.
The video here
The desmos link here
And the PDF of the presentation here
Though there are some typos etc, but that’s ok, as it was a some kind of ELI5 for some SPOs.
Okay, I see what you mean now, but I don’t think that there is any issue with the operator reward. The way that they’ve written it is very confusing, but you can reorganize the terms to make it more understandable. Let’s say that d is the relative stake delegated to the pool so that s + d = σ. This means that the relative portion of delegator’s rewards to be taxed by the margin fee is d/σ = 1 - s/σ. Now (f-c) is just the reward after the cost per epoch is taken out. Breaking down the bottom terms of roperator we have the margin fee c + the remainder (f-c) multiplied by (m+(1-m)s/σ). But this last term is just m(1-s/σ)+s/σ = m d/σ + s/σ or simply the margin on the (relative) delegated pledge + the pool owners (relative) pledge. All together, roperator = c + (f-c) * (m * d/σ+s/σ) is just the cost per epoch + margin fee*the delegators stake + the SPO’s rewards of their own pledge.
I think part of the problem with your analysis is that you are plotting the SPO rewards relative to their pledge. If their pledge is 0 then the “return on investment” blows up to infinity as in your plots. But this ROI isn’t really what SPO’s are concerned with, what they want are total rewards. According to the plots, if I run a stake pool near saturation with 1000 pledge ADA and then up my pledge to 2000 ADA, my ROI drops by nearly half, but in reality the total reward that I gain is negligibly affected.
I did not want to response as I could not see that you were seeking for the objective truth in your response. Anyway, I am just picking up your two examples:
- pool nearly saturation /w 1K pledge and
- similar but /w 2K pledge
1K Pool’s operator
- Epoch reward: ~1,781.00 ADA
- Yearly reward: ~116,958.00 ADA
2K Pool’s operator
- Epoch reward: ~1,782.00 ADA
- Yearly reward: ~117,000.00 ADA
So, what would incentivize that 2K pool to do a full pledge (with 2K) instead of splitting its 2K pledge to two 1K pledged pools (to prevent this type of Sybil attack what we see /w 1PCT and others?), if he is good in marketing and/or any other tools?
- Meaning it would earn 117K ADA in a year (~58x i.e. ~5800%) if he/she would not do it
- but would earn 334K ADA in a year (~334x i.e. ~334000%) if she/he splits his/her pledge, (independently of the ROI as percentage or arbitrary values). And is it fair? Ofc, it depends ond the SPO point if view…
So, that’s what I wanted to explain mainly in my presentation and not explaining formulas in some ELI5 format, though it requires for the audience tor understand the 1PCT like issues.
If I am not responding then it’s not because I think I am smarter (hell no, I more often feel the opposite when I am talking/chatting to/with somebody), or I agree or disagree with you, but simply because we would have two different conversations here.
I agree with you on
- There is, at times, an incentive to split pools
- There is currently little resistance to Sybil attack
- We could go on endlessly in circles discussing the topic
I disagree with you on
- Your view that the incentive (and perhaps solution) for pool splitting lies in the distribution of pool operator/member reward rather than the optimal pool reward function itself
- The strong assumption that after splitting a pool one can easily attract twice the delegates with proper marketing and/or any other tools, which leads to…
- While nothing is incorrect with the ROI calculation, I don’t think it is a “good” metric to showcase why a pool would choose to split (without additional consideration of the costs associated with hosting a new pool, attracting new delegates, and other things that are much more difficult to mathematically pinpoint)
This, probably, can be mathematically proven whether it holds or not, based on some models, despite any (my, yours or others) personal opinions.
For example Lars’ model (which can be found in his personal github repo) seems to me very good candidate.
Which is all about science. I do not want to be right, usually the good attitude is to always seek for being wrong. But, be aware of that you’re a bit biased as you have some theory and if I am right then you probably wrong, which is inconvenient or even threat and our the human brain is very clever to eliminate these kind of inconveniences/threatens by using some tools (confirmation bias, cognitive dissonance etc. etc.).
I am biased too (and aware of it), though I have no any theory but some observations, and I came to conclusion, which ofc can be wrong, after by analysing both part of the reward calcutations (even analysing the Ranking and the monetary policy which luckily corrected by some great community members).
And as I told, I have no theory, I have simple observation what we all can observe.
I never have this assumption even I have never ever did any firm statements relating to this.
I just have some observations. My example above was just simply focus on the issue by an example.
And we can argue whether it’s the case or not, but isn’t a bit contradict when you’re stating
There is currently little resistance to Sybil attack
and
The strong assumption that after splitting a pool one can easily attract twice the delegates with proper marketing
Also, it does not need attract twice, it’s simply enough to gain some delegates as it would earn the minimum reward, based on his pledge, anyway.
As we are facing this type of Sybil attack (there are different type of Sybil attacks), when pools are incentivised to split their pledge, or even not split just create a new pool based on the rewards earned as 1PCT[0-X] as an example.
And I said: incentivised, means not everybody would do it (as it is based on the balance of moral and financial incentives) as not everybody would do for seeking to oversaturate its pool (by using some tools) if there would not be z0 capping implemented on pledge and stake, but luckily we have k limit which just simply ensure that those few (adversarial) would not be able to do even if they want to (the other type of Sybil attack is eliminated by this).
And this is where Nash equilibrium comes to play to ensure that does not matter how, we humans the weakest link of the chain, try to trick the system will inevitably converge to the equilibrium.
The rewards showing as percentage relative to the investment (i.e. pledge) is just some visualisation tool for those who likes graphs and functions, for explaining the issue which comes from a pledge value from which below (left side of the inflection point) an SPO is incentivised to create more pools to maximise it’s yield/profit (rational behavior).
I give you some detailed example by not visualising as ROI but as some reward, assuming inflections point at pledge 22,693,660ADA:
- 1 large pool who has 40M for pledge
Scenario1, assuming fully saturation.
- Pool would earn ~2,415,537ADA if he would do full pledge, assuming full saturation.
- Pool would earn 2* ~1,175,421ADA, therefore ~2,350,842ADA, if he would split to his pledge to 2*20M pledged pool, assuming full saturation ( can be thought as on the right of the inflection point meaning earn less than the fully pledged ).
- Pool would earn 4* ~623,465ADA, therefore ~2,493,860ADA, if he would split his pledge to 4*10M pledged pool, assuming full saturation ( can be thought as on the left side of the inflection point meaning earn more than the fully pledged ).
Scenario2, assuming no delegated stakes at all s=p
- Pool would earn ~2,225,558ADA if he would do full pledge, assuming no delegated stakes.
- Pool would earn 2* ~1,024,743DA, therefore ~2,049,486ADA, if he would split to his pledge to 2*20M pledged pool, assuming no delegated stakes.
- Pool would earn 4* ~501,367ADA, therefore ~2,005,468ADA, if he would split his pledge to 4*10M pledged pool, assuming no delegated stakes.
Keep, in mind these are not extremes as we can see on live data with 1PCT and similar, but simply was demonstrate that if a pool is on the left side of the local minimum of the function then it’s incentivised to split his pool or create a new one from the earned rewards (mean 1 entity multiple pools), and it’s up to him whether it would do or not as it there is not force to unincentivise them like the cap on pledge and stake in the reward function.
Also, keep in mind the highest pledge in the public pools currently is 10M (EDEN, even he split his pledge from 20M to 2*10M which is a rational behavior to maximise his profit, I know him and and he is a very good guy), so everybody is on the left side of the local minimum are incentivised to split their pledge or create a new pool /w the gained reward (like 1PCT or DIGI now have 4 pools now), meaning incentivised to create a new pool by one entity. with the average 35K pledge of the ~1500 pools.
But, it does not matter what examples, which is based on data, I would write and how many times, as unfortunately, my perception of your writing tone tend to me think that you want to be right and back your theory which would eliminate any constructive dialogs from any opposite side.
Sorry, I wasn’t trying to talk down to you or give you any sort of a tone. I genuinely just don’t understand what you mean when you say that the problem is with roperator.
I think your last two scenarios illustrate that point. In the first example, by pool splitting, the SPO is better off by only -2.6% or +3% after assuming they gain enough delegation to saturate 1 or 3 more pools, respectively. In the second example, they are always worse off by pool splitting.
In the end, it doesn’t matter, we can agree to disagree. I’m just glad that people are into Cardano enough to really try and think about it at a level deeper than “when moon.”
1 Like
No worries, probably I overreacted a bit.
The c is constant (f-c)m is relative to the delegated stake independently of the pledge, meaning if the pledge = stake, then the operator would get the max anyway relative to the z0 (full sat) i.e. somewhere between 4.98% to 6.5%, depending on the pledge (as it’s propotional).
If the pledge is lower than stake it means it would have propotionaly higher reward. It’s like if I have 100K and fully saturated then I would earn (max pool reward - pledge) * expeted reward as percentage * margin.
In summ, even if I have 0 margin, and only pledge 50K (like 1PCT and it is higher than the average 35K) and I can attract delegators (as we can see /w 1PCT and similar), I would earn 24K ADA in a year, with no margin even it does not need to be fully saturated but reach some threshold saturation in which the delegators won’t feel real difference in their reward), that is 48% instead of put all 1M of the 1PCT pledges to one pool and earning just around 50K a year (50K versus 480K atm /w 20 pools).
So, that is my issue /w the r_operator and even I have not calculated the 1% margin what comes top of that.
So, what would be the 1PCTs best strategy in the game to maximise his profit?
Create a new one when he earned 50K rewards from his other 23 pools and do it again and again until the game allows it.
Ofc, different pools can have diff strategies, but this seems to me the more rational one.
Hi guys, new here and deeply interested in Cardano. As background I’ve read a few books on blockchain a few years ago and quickly realized that blockchain has fantastic potential, but that Bitcoin is fundamentally flawed.
That out of the way, I’ve observed the same centralization and “non democratic” power shifts with Cardano which is quite distressing. I for one, considered become a SPO but the system is already skewed heavily towards incumbents, which is a disincentive.
Returning to first principles, what is the desired outcome?
- Capable Pool Operators - hardware, software and stable internet that meet the MINIMUM Cardano network requirements
- Pool operators incentivized to deliver 1.
- System geared to prevent decentralization
Suggestions
-
Per attached research paper in this thread, I’d propose that the eligibility to create a block be carried out locally and include a capability check. Hardware, software and connection quality. Perhaps that happens already? Successful stake attraction is to a large degree, marketing - the product doesn’t always meet the marketing promises.
-
Has anyone ever considered that the ADA be distributed equally among all SPOs? With 100% random selection of which node mints the block? I for one, as an example of an ‘average Joe’, would definitely participate in such a system. Right now it’s a popularity contest with little chance of earning. This is AFTER the requisite upskilling required to set up a pool. For many, that’s one step too far, and per thread links, pool operators are already dropping out.
-
I have read through this thread twice and it seems to come down to SPOs spinning out new pools when saturated (potentially with multiple accounts belonging to the same entity). It makes economic sense to do so but centralizes the system - which renders it worthless. Various theories would say that a loss in value should drive rational behavior to decentralize, but we haven’t seen it substantially in practice as far as I can tell. I guess all the big players tell the other big players ‘you go first’… Suggestion: As a South African, a big driver of blockchain on the African continent, besides currency, is digital ID. There seems to be a lot of work on this place with blockchain and could actually be a major driver for Cardano adoption. So perhaps as a start Pool Operators would need to create a blockchain digital ID, to which say 5 Pools max could be created? As to the “how”, I’m sure a lot of thought has already gone into digital identity, blockchain and how to prevent fraudulent IDs being created. In terms of hardware I think smartphones should be an enabler for frequent fingerprint rescanning, etc? Even in the developing world, smartphones and mobile have leapfrogged fixed line infrastructure, and is likely to accelerate with satellite broadband (thanks Elon!).
Anyway just some suggestions for discussion, I’m not an expert in this space by any means!
Cheers
Asogan
That’s not really true. What you can observe atm, is only the side effect of the initial phase, which would last I think for 2 to 3 years or more. As, the system, if the math hold in the protocol, would inevitably converge to a near Nash equilibrium.
Though, the system parameter would need to be tuned and perhaps even some of their part a little-bit changed based on the collected data, but that’s the scientists jobs and not an averages joes (like me) with all the answers who even cannot do proper Eastern-EU/Asian edu systems’ based primary and high school level math (algerbra, calculus) but make very firm statements about the protocol.
This multi pools arising is only because the delegators are behaving non-rational atm, due to their lack of the understanding and due to our by evolutionary wired brains, how the delegation works and due some other minor issues what the scientist, I think, are aware of and will fix it by the time. As a rational delegator would invest (delegate) where the highest profit is.
Unfortunately, your suggestion assumes human interactions what the system should avoid completely as we are the weakest link in the chain. So, if the protocol rules are set properly and the node’s software follows the proven rules, then we would not need these at all.
For example brief replies to your suggestions.
-
The pools will converge by the time to the best configs (HW) etc, because after the era of these current marketing bullshit won’t too much impact on the delegators, as their business will be in a very high risk that they could not afford, if they won’t do this. But, it assumes the we will grow and mature, which is very unlikely. It’s like evolution, which is a very slow process.
-
The system to be sustainable will need to reach near Nash equilibrium with no doubt, that means those
knumber of surviving servers will have almost the same stakes. This is like natural selection which is healthy for the system. Also, you would need to have some level of understanding in Cardano Ouroboros Praos (and the others, Genesis, Hydra, Chronos etc), why your suggestion is very unhealthy for the system. -
Yes, but it’s just for the initial phase and they have benefits and they (we) the SPOs are riding it.
Also, your suggestion seems just treating (slightly mitigate) the disease and not curing it (eliminating those incentives). Also, keep in mind, the system is getting more and more decentralized by the time, what we can observe even now.
Do not get me wrong. Our community needs a lot of critical thinkers (to compensate the probably the majority greedy/speculative ones), which is very rare in the generic population, despite everybody thinks that he is the most critical thinker in the word (but luckily it can be objectively measured).
But, the issue is that we ppl need to reach some level of understading, by very hard way by learning, learning and constantly learning, how Cardano (and everything around) works what can and cannot be done etc.
But that’s tough, and we as species are very lazy, and, therfore, unfortunatley, we are tribal, who needs idols, gods etc, who have all the answers, as that is the easy way instead of learning seeking for objective truth endlessly.
So, what I often see here, that the ppl have. I am not talking about you, as you seem reasonable, but to those who just jump in said things and even have no clue of their level of understanding (illusory_superiority). I.e. they overestimate their low ability (Dunning-Kruger)
I think this is the second mention of Dunning-Kruger here in the forum after my first in the early 2018.
If it would not be the case then there would not be scammers, as they would not be able to survive, but as we can see they’re blossoming in this era.
And do not get me wrong, I constantly critizise the things in Cardano protocol that have, imo, design flaws and/or unhealthy implementations etc.
1 Like
Thank you for taking the time to reply, and for your thoughtful answer. It appears that this issue has been raised repeatedly in other forums e.g. reddit, twitter so at least I’m not the only one with this concern.
Hopefully, delegators will be better educated and more rational in future. I don’t know. But considering the politics of the time in the USA (and where I’m from in South Africa), I guess I don’t have as much faith in people being rational 
I also agree, that it would be best if incentives were aligned to drive the right behavior, but I see no harm in doing both i.e. correct incentives + programmatic rules
I have slightly edited my proposals so it might make more sense?
- I was looking at MINIMUM requirements, not best configs. Otherwise it would have the same issue as BTC. Just the minimum HW + internet connection + correctly configured software, to be checked locally and confirmed with the network.
- If 1. is in place, then why not distribute the reward equally among operators, with 100% random selection to choose a node that can mint a block? Right now, the surviving servers would be the winners of a popularity contest versus a capable ‘builder’, or of the networking effect for incumbents like Binance.
- Agreed, that my proposal is to treat the disease vs curing it - but I see the benefits of doing both? It is just a suggestion, limiting the number of pools that a single entity controls is a good thing - And we’d need to look at good solutions to prevent anonymous cloning of pools by the same entity. Pledge is a decent idea at first glance, but won’t stop Elon, Jeff, Warren and Bill from creating 1000s of pools!
My background is mechanical engineering, though I’ve moved into finance over the years, so I’m not expert by any means!! I’m still reading and learning.
Thanks for the discussion!
IMO, Incentives are and should be aligned by programmatic rules.
-
There is a min requiremetns set by IOHK/Cardano. See here
But think about it a bit. We need memory, storage, network and computational resources. Storage needs for store all the chain’s blocks, which grows almost linearly by the time and 24GB is more than enough (~7GB is the chain, atm). Memory required for keeping some states in it, for example ~300K UtXOs, peer list, caches etc.4GB RAM is ok for that. Network, if we assume 100 TPS then we need some bandwitdh to handle the constant flow of the max 16K txes and 64K blocks. So, I would say 50Mb/sec as starter is fine ( they recommended 2Mb/sec), what a Raspberry PI at home can easily handle. And keep in mind that there is some slight differences between the pools and the other nodes (relay, full-node wallets) etc. The pools need only a little-bit of more extra resources due to their block creation process, which is way-more compensated by much more less peer connection then that relays have (assuming the relevant blokc-producer is behind multiple relays). -
Because it would lead to different type of Sybil attacks. The servers only able to mint a block by their invested pledge and acquired stake. And every pool can earn around ~5% relative to their pledge (independently their pledge value) yearly anyway. For example if I have a pledge for example /w 10K ADA, there is a probability that I will mint one block in a year and that time I will earn around 500ADA due to the apparent performance. So, that 's the beauty as an everyday mom at home can run a pool with her for example 50K investment and earn 2500 a year running a pool on a raspberry pi4 for example, assuming or pool forges the block when it’s scheduled.
-
Everybody can and should spin up as many pools as much ADA they have for full-pledge, the others should have only and only one (though it’s not incentivised by the protocol). Not the number of pools are the threat, but the number of pools /w low pledge gaining power. I.e like Elon creates 1000 pools with almost 0 pledge and would able to attract all (or at least >50% the delegators to hisr pools), meaning he would have 100% (>505:) says on the network, that is the issue. As you only have as much says as many pledge + stake you have. Which are mitigated by the protocolThese are mitigated
Agree, as an example, my pool was scheduled to create only 18 of the 24.2 expected epoch blocks (based on pledge+stake) in the last epoch due to the lottery.
Some delegators, monitor this and leave and delegate to an other pool in the hope they will earn more in the other pool. They did not think this through carefully, and their investment will probably converge to less gain by this behavior (not really true).
Why is it the case? Because they leave when the luck is less then the average, meaning the next few epoch probably will have more luckier than this, but we can say that in long term the luck averages out to nearly ~100%. And they move a much luckier pool for this current epoch, which probably will have less luck in the next few epochs.
But, keep in mind, this is not really true as it’s a gambler fallacy (but this is pseudo-random so who knows), as probability does not have any memory (i.e it’s not remembering the past event).
So, the best strategy is staying than leaving, not because of the probability, but they won’t spend money for the re-delegations etc. and does not make sense of moving, unless the other pool has better properties, meaning less margin and cost and higher pledge and stake.
3 Likes