Philosophy
Motivation for Edgevana Stake Pool's algorithm
See Delegation Strategy for more details on each score.
Background
At a high level, Edgevana's Stake Pool is designed to enable the user to earn network rewards by contributing to the security of the Solana network while minimizing risk. To achieve this, Edgevana developed an algorithm that decides which validators to stake to and the proportion of their pool stake.There are many ways to rank validators, each with their pros and cons. For example:
-
You can rank validators by network reward earnings. While this approach in theory ensures stakers get the highest amount of network rewards, it's actually rather naive. If an algorithm just ranked validators by network rewards, the pool might stake to some validators who are:
-
Not sustainable: The validator could cease operations at any time which would leave the pool staking to a validator who is no longer online and earning network rewards.
-
Malicious: A validator could have a 0% commission rate today but set a 100% commission tomorrow. This is an attack vector, informally known as "commission rug-pulling." If this happened, the pool's network rewards would drop.
-
-
You can rank validators by reputability. In practice, this is a bit hard to do, however it is reasonably achievable. This approach is optimized for the least amount of risk, however it's also rather naïve because some more reputable validators extract higher commissions which results in lower network reward earnings.
-
You can rank validators by performance. However, with this approach, you run into the same issues you do by ranking validators by reputability because the pool might end up staking to validators with high fees.
Generally, it is our belief, that a pool should incorporate these metrics into an algorithm:
- Performance & Reliability
- Reputability
- Commission adjusted voting
- MEV Rewards (note: we do NOT mean MEV commission)
- Decentralization
Below, we will go into detail on each one of these bullet points.
Performance & Reliability
Performance is the lifeblood of any validator, especially on a high performance blockchain like Solana. If a validator is not online and voting, they are not participating in network consensus and earning network rewards. Because of this, we weight 50% of a validator's score on performance.Vote credits is a great measure of performance. Edgevana's algorithm uses a max commission adjusted, 10 epoch average of vote credits to judge performance & reliability.While it is true that some validators use software modifications to gain more vote credits than other validators, there are mechanisms in place to protect against "modders" from getting most of the stake in the pool (see Decentralization).
Reputability
This is arguably the hardest metric to quantify, however one of the most important ones. The pool needs to avoid two main scenarios:
- Validators who might cease operations or validators who might incur large periods of downtime.
- Validators who might "commission rug-pull" See Background)
Because of this we bake in reputation into our algorithm. We do this with two measures:
-
History: A validator that has been voting for 10 epochs (~20 days) is less likely to cease operations then a validator who has been online for just 1 epoch (~2 days). This is because running a Solana validator is costly. If a validator has been running for 10 epochs, it is likely they already have achieved enough stake to break even, or close to it. While Edgevana's algorithm does NOT require a validator to have been online for 10 epochs, we do include an operating history score with a formula of epochs_online / 10 with a maximum score of 1. We do not weight validators who have been online for more than 10 epochs higher than others. This is because we also want to decentralize the network by spreading out stake.
-
Published Information: On Solana, a validator can publish info to advertise their validator to stakers. A validator that is "undoxxed" is likely to be less reputable than one that is "doxxed." This is because if a validator does something malicious (ie: "commission rug-pull"), they have more to lose. Edgevana's algorithm checks for:
- Name: The name of the validator (ie: Coinbase, Edgevana, Kraken, Figment, etc). If a validator has published this, they get 25% of the info score.
- Description: A brief description of the validator. If a validator has this published, they get 25% of the info score.
- Icon URL: A logo that can be displayed to stakers If a validator has this published, they get 25% of the info score.
- Website: A website for the validator. Many validators publish websites, however some of them don't publish a real website. For example, some validators set their website to a Twitter / X link. If a validator has a real, unique website published, they get 25% of this score. If a validator simply does something like set their website to a Twittter / X link, they get 0% of this score.
Commission
SOL is an inflationary asset. Every epoch, any validator who has voted is eligible for receiving part of the inflation rewards emitted by the Solana network. This is known as commission. The commission a validator sets comes directly out of the staker's rewards. There are 2 sides to the commission argument:
- Stakers will always want a validator to set the lowest commission possible. This is because they want the highest returns possible.
- Validators will always want to set the highest commission they can. They are businesses, have real bills to pay, and are also the lifeblood of the network.
Edgevana's Stake Pool does not look at the commission level set by validators but rather their commission adjusted voting performance. Historically v1 algorithm used a commission score.
MEV Returns
There are many ways for validators to earn rewards and MEV is a smaller, but important variable to that equation. One way to judge how much MEV rewards a validator might collect is by looking at MEV commission rates. This is approach is rather naive however. For example, If you compare validator A who has an MEV commission rate of 8% and validator B who has an MEV commission rate of 7%, you might assume validator B will earn higher rewards due to a lower commision. However this is not the case. MEV is dependent on many things such as leader slots and network latency to a block engine / relayer, not just commission.Because of this, Edgevana's algorithm uses a stake weighted, max adjusted metrics on MEV distribution. This means Edgevana's algorithm actually allows for validators to take higher MEV commission rates if they are able to generate more MEV rewards and proportionally distribute more to stakers. This is a win-win for everyone, and we see no reason to penalize validators with higher MEV commission rates if they are able to generate higher rewards.
Decentralization
Like reputability, this metric is also hard to quantify. Some other stake pools use GeoIP data to quantify this. We think there are issues with this because:
-
Legitimate infrastructure providers get penalized if a pool uses a "provider score", or anything that weights stake by ASN. What ends up happening when weighting stake by provider is validators will slowly move to lesser-known, less reputable infrastructure providers, which we do not believe is a good thing.
-
GeoIP data is extremely unreliable. For one given IP, it is not rare for two different GeoIP providers to report different locations. This renders the data rather useless. In addition, GeoIP providers do not keep the data up to date enough to use this data reliably in a delegation algorithm.
Instead, Edgevana's algorithm uses two different inputs to maximize decentralization:
-
Edgevana Region: We calculate the amount of stake you are contributing to a region as a portion of the overall region of edgevana customers. This allows us to add value to the decentralization of Edgevana's offerings as well as the Solana Ecosystem.
-
Supporting smaller validators: Another way to decentralize the network is simply to have more validators. In a hypothetical scenario, if a network had 10,000 validators all running in unique cities and data centers but 51% of the stake was delegated to a single validator, the network is still not decentralized. To support smaller validators, what we do is calculate the average stake of all validators. The closer a validator is to the average stake, the higher score they get. This also means that validators with high stake levels get lower scores. With this method, Edgevana's algorithm is dsigned to optimize for distributing stake across more validators, increasing decentralization.
-
TPU Concentration (Deprecated): If a high percentage of stake uses the same TPU (Transaction Processing Unit), this becomes a huge area of centralization for the network. In that scenario, the TPU provider could decide to filter out certain transactions where several validators would not receive the transaction affecting it's confirmation time. While it's certainly the validator's choice to decide which transactions they want to include, it's objectively not good for decentralization if multiple validators share the same TPU. Therefore, we weight stake by TPU, subtracting out the validators own stake, and score a validator accordingly. If a validator is the only one that's using a given TPU, they get 100% of this score.
Attribution
We'd like to thank the following people and organizations for making our algorithm possible: