Validator scoring and rewards
Validators and nominators both receive revenue for securing the network. The amount, rewarded as VEGA and infrastructure fees, depends on factors including how much stake is nominated to the validator.
To be considered for staking rewards, a tokenholder must associate VEGA to a Vega key and nominate one or more validators.
Use governance.fairground.wtf to try out associating testnet tokens and nominating validators. Staking rewards are paid into your Vega wallet after each epoch ends.
VEGA received as staking rewards must be withdrawn to an Ethereum wallet, and then associated to a Vega wallet, before they can be staked.
In each epoch, rewards are distributed among validators in proportion to the number of tokens they represent (i.e., their total stake). The total stake includes a validator's own stake and the tokens nominated to that validator. Of this reward, a fixed amount is distributed among the tokenholders the validator represents. The proportion of staking rewards distributed to nominators is 👀0.883.
The reward received by each validator, and therefore passed onto their nominators at the end of each epoch, is based on their validator scores. The score accounts for any penalties. However, there is no token slashing, i.e., a validator or a tokenholder will not lose tokens through any validator actions or poor performance.
A validating node’s score is calculated based on three factors:
- Total stake
- Penalties for too much stake
These factors drive a number of individual scores which combine to form the overall score for each validator in the epoch.
See each validator's rewards scores by querying the API.
Raw validator score
The raw validator score represents the overall share of total stake the validator represents, and determines if a node is overstaked, and thus penalised.
Overstaking is a risk to network security since it can lead to a concentration of voting power with a small number of consensus validator nodes, allowing those nodes to potentially halt the network.
See below for how the validator score is calculated.
min_validators= value of the network parameter that defines the minimum viable number of consensus validators to run the network
num_validators= actual number of validators running nodes on Vega
comp_level= value of the network parameter that defines the competition level
total_stake= sum of all stake across all validators and their nominations
optimal_stake= total nomination divided by the greater of
min_validators, OR (
comp_level): Optimal stake is how much stake each validator is expected to have, at most
optimal_stake_multiplier= value defined by 👀reward.staking.delegation.optimalStakeMultiplier), which indicates how many times the optimal stake a validator is penalised for, if they are further than the optimal stake
validator_stake_i= stake of the given validator whose score is being calculated
flat_penalty= the greater of 0, OR (
higher_penalty= the greater of 0, OR (
The raw validator score is calculated as follows:
raw_validator_score = (
In other words, the network calculates an optimal stake that represents an even distribution of stake for the current number of consensus validators and the desired competition level. It then penalises any validators that have stake that exceeds that amount. The raw validator score is then the resulting amount divided by the total stake on the network.
Proof of stake rewards spec ↗: Read the full details on how scores are calculated.
For the network to run effectively, it requires nodes that are highly available, highly performant, and process the transactions expected of them. Therefore a performance score is calculated for all validator nodes. This calculation of performance score differs slightly between consensus and standby validators. Candidate validators do not receive a performance score.
In tendermint consensus, the number of times a consensus validator can be expected to propose a block is roughly proportional to their voting power from the previous epoch. Therefore, in order to assess the performance of a node, the protocol compares the number of times that node was expected to propose a block in the previous epoch against the number of blocks it actually proposed.
p be the number of times the validator proposed blocks in the previous epoch
b be the number of blocks in the previous epoch
v be the voting power of the validator in the previous epoch
t be the total voting power in the previous epoch
v/t x b
The number of blocks the validator is expected to propose is equal to their share of the voting power in the previous epoch multiplied by the total blocks proposed in that epoch.
performance_score = max(0.05, min((p/expected, 1))
The performance score is equal to the ratio of number of blocks actually proposed vs number of expected blocks, with a score of 1 representing a node that has proposed at least as many blocks as were expected.
Standby validators do not propose blocks, therefore the method above is not suitable for assessing their performance. Instead, the performance score of a standby validator is calculated based on them successfully submitting heartbeat transactions to the network. The performance score for new standby validators is set to 0 for the first epoch.
Validator candidates that have submitted a transaction to become standby validator nodes will need to send a hash of block number
b, separately signed by the three required keys and submitted, during each epoch and every set number of blocks (
numBlocks = the higher of (the lower of (50 and the epoch duration in seconds) and (epoch duration in seconds x 0.01)
The message with the signed block hash must be in blocks
b+numBlocks to count as successfully delivered.
The network will verify this to confirm that the validator owns the keys.
b is defined as:
- The first time, it is the block number in which the joining transaction was included
- Subsequent times, it is incremented by
The network will keep track of the last 10 times a standby validator was meant to submit the transactions, and the performance score is the number of times this has been verified, divided by 10.
Validator score and voting power
The validator score for the epoch is calculated as:
multisig_bonus is used to ensure consensus validators are registered to the multisig contract. If a consensus validator is not registered to the multisig they will get a score of zero, otherwise they get 1. Non-consensus validators are always awarded a
multisig_bonus of 1 since they are not required to be registered to the multisig.
Finally, the validator score is normalised to sum to 1. This
normalised score is then used to define a node's
voting power in the next epoch. Note that the
voting power is multiplied by 10,000 and rounded to integer values, to meet Tendermint requirements.
The Vega network does not prevent tokenholders from nominating validators with an amount that would cause a node to be overstaked. Tokenholders must actively manage their stake and keep track of the nodes they support.
At the end of each epoch, staking rewards are distributed to validators, and then their nominators. The full pool of staking rewards is allocated to validators, with each validator receiving
total rewards x
Each validator keeps a share of these rewards, where that share is:
1 minus the nominator (delegator) share of 👀0.883.
If the validator does not have sufficient self-stake as defined by 👀3000 tokens then it will not receive these rewards and they are returned to the reward pool.
The remaining rewards are then distributed to the tokenholders nominating the validator node in proportion to their share of nomination on that node.