THE STATE OF THE UNION
At this very moment, the BLS/Validator ratio is about 10:1 → Most of the elected BLS slots in the Harmony ecosystem are in the hands of a few big players. The overwhelming majority of holders decides to stake instead of validating → the network tends towards centralization
KEY PROBLEMS THAT FAVOR CENTRALIZATION
- Staking is more profitable than running a validator, except for big entities (eg. businesses)
- Small validators lack the flexibility to stay within the effective range, pushing new delegators to big validators out of fear of missed rewards
- Stake centralization is detached from holdings. Professional staking businesses can scoop up delegations without a big investment in the ecosystem (just 10k ONE needed)
Validators earn the most rewards by becoming as big as possible. Established staking providers can use their reputation and infrastructure to outperform small validators and attract delegations. Small validators struggle to stay within the effective range and suffer from lost rewards and an unfavorable dashboard-statistics, further perpetuating the cycle of delegator migration to big validators. Big ONE supporters/investors, that don’t happen to hold exactly the effective range, are better off staking and therefore disincentivized to validate.
According to the rich list, there are only about 30 addresses who happen to hold a stake that would enable them to be within the effective range (effective stake x1, x2, x3, and so on). Every other ONE holder is better of staking. Considering the workload, node cost, low staking fees, and the lack of flexibility as a self-validator, staking might be a better choice for most of these 30 as well.
Ideally, we’d like to get those big holders (and the many who are slightly above or below) to spin up their validators instead of choosing the easier (and better rewarded) way of staking. How do we go about attracting a broad network of as many unique and personally invested validators as possible?
INTRODUCING: ELECTED VALIDATOR REWARDS
In short: EVR is a fee paid by stakers to all elected ONE validator addresses. 10% of all staking rewards are distributed amongst all elected addresses each epoch, as a fee for running the network.
I’ll use simplified numbers to illustrate this proposal:
Total Stake: 3,500,000,000 ONE
Median Stake: 7,000,000
APR: 15%
Elected Validators: 50
Elected BLS Keys: 500
The total stake yields 2,2m in rewards per epoch. 10% out of all staking rewards (220k ONE) will be distributed to the elected validator addresses on a per epoch basis. (~220k ONE per Epoch/50 addresses→ 4400 per address per Epoch)
With 50 elected validators:
7m ONE staking rewards/year: ~ 945,000 ONE
7m ONE validation rewards/year: ~ 2,000,000 ONEWith 100 elected validators:
7m ONE staking rewards/year: ~ 945,000 ONE
7m ONE validation rewards/year: ~ 1,470,000 ONEWith 200 elected validators:
7m ONE staking rewards/year: ~ 945,000 ONE
7m ONE validation rewards/year: ~1,200,000 ONE
Similar to EPOS incentivizing overall staking participation by exponentially increasing rewards at a low total stake, EVR would incentivize validator participation with vastly increased rewards at low validator numbers. By having a noticeable difference between validation and delegation rewards, ONE holder within (or close to) the median range will be incentivized to spin up their node, instead of taking the easy path of delegation. Thus true centralization can be achieved.
Smaller stakers might finally benefit from pooling together to run a collective validator. Current pooling wouldn’t differ from regular staking and is not worth the effort. Overall I believe this would result in a vast increase of unique validators.
POSSIBLE DRAWBACKS
Big validators would just spin up several validator addresses
That’s a balancing act. Each validating address would require at least one node and thus vastly increasing server cost. Every additional BLS key on the node would get subsequently less profitable, as the fee is per address. Staking as a service businesses would have to juggle the additional rewards with the increased server-cost and the optimal number of BLS keys per node.
The network would be flooded with low-quality validators
The current network is dominated by a relatively low number of unique validators willing to do the work. The overwhelming majority of participants are taking the easy way out by staking. Further incentivization of validators might increase the competition for precious slots. Validators will still be interested in keeping an effective node running, so they don’t drop out in the next election.
Lower staking rewards would disincentivize stakers to participate in the network
I believe that the drop for regular stakers would be negligible (13,5% instead of 15% apr), compared to the price appreciation benefits of a healthier network. EPOS rules still apply and would compensate for a potential drop in staking participation. Big Stakers who would like to receive the extra bit of rewards, would be incentivized to compete for a validation slot themselves.
EXAMPLES
A) Jack has 12 million ONE. He’s been with Harmony from the start, is passionate about its decentralized values, and would love to contribute to the network, but he’s not interested in running a business and marketing to potential stakers. With his holding, he could be elected easily on his own, as the current median is 7 million ONE.
With the current situation:
Staking 12m ONE (-5% fees) → ~ 1,7m ONE/year in rewards
Validating 12m ONE (effective 7m ONE) → ~ 1m ONE/year in rewards
Validating (9m with overhead for fluctuations) + Staking (3m) → ~1,45m ONE/year in rewards
Unfortunately, validation would put him far outside of the effective range and he would lose the rewards of several million ONE. Staking the overhead with a different validator puts him at risk of dropping out of the elected slots, should the median rise unexpectedly. Despite the fees, staking would yield better rewards at lower risk than running a validator. Even when staking with an overhead for median fluctuations and staking the rest, pure staking wins by a huge margin
With Elected Validator Rewards:
at 50 elected validators:
Staking: 12m ONE → ~ 1,6m ONE/year in rewards
Validating 12m ONE (effective 7m ONE) → ~ 2m ONE/year in rewards
Validating (9m with overhead for fluctuations) + Staking (3m) → ~2,4m ONE/year in rewards
at 100 elected validators:
Staking 12m ONE → ~ 1,6m ONE/year in rewards
Validating 12m ONE (effective 7m ONE) → ~ 1,47m ONE /year in rewards
Validating (9m with overhead for fluctuations) + Staking (3m) → ~1,8m ONE/year in rewards
at 200 elected validators:
Staking 12m ONE → ~ 1,6m ONE/year in rewards
Validating 12m ONE (effective 7m ONE) → ~ 1,2m ONE/year in rewards
Validating (9m with overhead for fluctuations) + Staking (3m) → ~1,6m ONE/year in rewards
Even with a massive overhead for median stake fluctuations, Jack manages to beat, or at least stay competitive with pure staking.
B) Mike has 8 million ONE. He’s a numbers guy and always looking for the biggest reward. He could run his validator and fall within the effective range easily. There’s even a little overhead to keep running should the median rise.
With the current situation:
Staking 8m ONE (-5% fees) → ~ 1,14m ONE/year in rewards
Validating 8m ONE (effective 7m ONE) → ~ 1m ONE/year in rewards
Even if Mike would not have to pay for servers, staking would outperform validating by a sizeable margin. There’s no reason for Mike to validate.
With Elected Validator Rewards:
at 50 elected validators:
Staking: 8m ONE → ~ 1,08m ONE/year in rewards
Validating: 8m ONE (effective 7m ONE) → ~ 2m ONE/year in rewards
at 100 elected validators:
Staking: 8m ONE → ~ 1,08m ONE/year in rewards
Validating 8m ONE (effective 7m ONE) → ~ 1,47m ONE /year in rewards
at 200 elected validators:
Staking: 8m ONE → ~ 1,08m ONE/year in rewards
Validating 8m ONE (effective 7m ONE) → ~ 1,2m ONE/year in rewards
Even with high validator numbers, spinning up his node pays for Mike big time. That’s all the convincing our numbers guy needs, validation it is!