Sure! The spec says that you take the square root of the total amount staked, multiply it by the BASE_REWARD_QUOTIENT, divide by four, and then divide the effective balance by that amount to come up with the base reward (snipped a pic in my answer). In essence the relationship between the total balance of staked ETH and the reward is described by an inverse square root function.
Per Vitalik’s PDF, this was done to reduce the potential reward (vs. rewarding a flat inflation amount no matter what) for performing an “epsilon attack” whereby you grief a subset of validators and take a larger share of the staking reward pot for yourself. Basically, the design goal is to make it less profitable to forcefully reduce the overall number of validators to concentrate rewards. The PDF was helpful in explaining this.
I was saying that this idea makes sense, but we’re still using certain constants (e.g. fixed numbers) like the BASE_REWARD_QUOTIENT which helps determine a baseline for how much Ether is rewarded. For example, adjusting the BASE_REWARD_QUOTIENT up would make all rewards smaller, while adjusting it down would make all rewards larger. Hence our questions for where it comes from.
For the record, it’s not like I have a “right number” in mind. I’m just curious how we think about these things.
Can you share the link to the PDF (if it’s public)? Does it maybe contain any practical explanation/example of a real-life epsilon/griefing attack? I’m asking because I couldn’t come up with one (but I’m not yet saying it’s unrealistic to expect it in reality).
The PDF describes an attack that occurs completely on-chain (don’t think the PDF goes into the exact mechanism), but it’s a plausible way to attack the network offchain as well simply by taking large swaths of the network offline. This whole argument is just a technical way of saying that, if it’s too profitable to take stakers offline and hoard the rewards, someone’s sure to try it.
One scenario: I’m imagining someone targeting somewhere like RocketPool or Coinbase to take their validator clients offline. If they can do this for extensive periods of time, they can reap significant rewards by becoming a larger percentage of the stake. As for attacking via on-chain censorship, I can’t speak to an example either.
Thanks @haokaiwu, I’ve just read the paper, it’s insightful.
Yes, I thought of this same thing while I was reading the paper, offline attacks are a way more realistic threat IMHO. Even existing miners/pools are more or less constantantly under DoS and other attacks (especially when important/controversial protocol-related decisions are being or about to be made), so I guess there’s no reason for this not to happen with (large) validators. For this and other (obvious) reasons, I think the client/validator software should be designed to make running large number of validator clients a challenging task (the complexity/resources/work required should, if possible, increase linearly with number of validator instances, not logarithmically).
This is with a fixed 1% yearly inflation, which is easier and looks more attractive imo. I would suggest inflating on a fixed 100 mil and not on total supply, to prevent runaway inflation after xx years. In simple terms: create 1 mil ether every year to reward validators. Inflation will keep going down this way and will only be 1% in year 1, and end up being less and less (0.5% after 100 years as supply will be approx 200 mil then).
Example C brings the economics more in line with the interest of competing networks - attracting larger players now willing to build services around their spread - would also bring in existing validators with professional experience that in theory are more reliable. On the retail side the interest now exceeds other ethereum based defi products. - Something between B and C could be the sweet spot - 8.5% at 10MM Eth
I think it is also very important that the initial inflation is high enough so that we can credibly say that we should/will not ever need to increase it in the future as it would set a precedent regarding ether’s store of value properties. I know a lot of people here do no support the whole SoV narrative but I believe a credible monetary policy is essential, fixed supply/inflation or not.
I’m was not especially pro SoV narrative before, but then @econoar made an important point and I changed my mind:
That said, your suggestion is in line with mine, I also think it’s better to set it reasonably high (especially because even that “high” still means <1%), and then hopefully reduce it in future. FWIW, that future reduction should have some positive psychological effect on the price (just like increasing the inflation would probably have negative).
I see two reasons for not setting inflation too high from the beginning. First, we’ll likely have a hard time adjusting it back down in the future. It’s unrealistic to expect stakers to vote against their own best interests and lower inflation “for the good of the network” when doing so is entirely voluntary. Yes, we’ve successfully reduced mining rewards in the past, but we relied on the community’s pre-commitment to PoS as a justification for steadily reducing PoW mining rewards. We won’t have such a convenient justification going forward.
Second, if the ratio between inflation rewards and transaction fee rewards are too far off, it encourages perverse incentives from the stakers. Assuming the transaction fee market doesn’t change under PoS, let’s use 500 Ether as the median transaction fees paid per day. Multiplied by 365, that means stakers earn 182500 Ether total, making it a 1.83% return assuming 10M staked Ether.
In @econoar 's Example C, inflation rewards more than 5x transaction fees. As a rational actor, stakers have a much greater incentive to keep the total staked amount low rather than increase the usage of the network. To illustrate, imagine we quadruple the transactions fees paid and double the amount of people staking. Transaction fees now return 3.65% from 1.83%, and we’ve doubled the “protection” in terms amount staked. However, stakers actually lose out 1.18% return in total since the inflation reward goes down from 10.15% to 7.17%.
As a result, at high inflation amounts, stakers could be incentivized to minimize the amount of other stakers at all costs, even if it results in less usage of the network. Price fundamentals of ETH would also largely track the amount of people staking, not network usage. It’s worth more if fewer people stake given the returns you can get, and it’s worth less if more people stake. This behavior is contradictory to the design goals of Ethereum, which seeks to align incentives with increasing network usage and increasing economic value in each transaction.
As an aside, I don’t believe we should treat staking participation as strictly an economics problem (e.g. how much inflation should we allow to give us a necessary return). After all, the biggest holders of Ether likely are people with a big stake in the success of crypto (exchanges, whales like Novogratz, etc.). They’re not diversified between asset classes in such a way that they could shrug off the failure of an ecosystem as big as Ethereum. Based on my experiences with the community, it’s also not unreasonable to expect that many existing community members will stake just to support the network in the short run.
Therefore, practically speaking, I believe the issue of staking participation is as much a question of community organization and rallying to get stakers in the short-term as it is a rational discussion about inflation targets for the long-term. As such, it’s best if we don’t get ahead of ourselves when talking about the necessity for a certain amount of inflation.
This is what is happening with lightning network right now. There is no real incentive to stake other than earning very small amounts on transactions and the ‘cool’ factor of it all. The overhead of running a node is very small, can be done on a Pi, so electricity consumption and investment is minimal.
Some differences between LN and Serenity is that it is possible to immediately withdrawal and there is little risk of losing your stake for not behaving properly, downtime, or bugs. I understand that you would rather avoid focusing on the economics, but it is for those reasons why, as a miner / staker who cares about the security of the network, I’m interested in the economics of staking.
In short, Serenity is risky business, which deserves economic incentives to offset the risk.