Socean launched with an initial delegation strategy, and is transitioning into a more sophisticated optimal delegation strategy.
Socean started off delegating in equal proportions to the top-performing N (~40) validators that are not in either the minimal security group or the minimal data center security group.
We don’t delegate to either minimal group in order to increase decentralisation. By definition, validators in either minimum group can halt the network. For the health of the network, we need to move stake away from validators in minimum groups.
We derived an optimal delegation strategy based on Bayesian mean-variance optimisation and utility theory. We give a simplified explanation below. The full explanation can be found in our technical whitepaper.
We first define our utility function, which expresses how much we care about APY, risk, and decentralisation:
$$ U = -e^{-\beta \cdot \textrm{TotalSOL} - \alpha \cdot \textrm{NetworkHealth(\textbf{w})}} $$
Here, $\beta$ is the degree of risk aversion, and $\mathbf{w}$ is a weighted allocation of stake.
The $\textrm{NetworkHealth(w)}$ function is a function of stake concentration in individual nodes as well as datacenters and jurisdictions.
We use recent work on Bayesian statistics (Bauder et al. 2020) to express the expected utility as a function of historical data, and use gradient descent to find a weighted allocation corresponding to the approximate maximum utility.
Socean will transition over time into becoming fully autonomous. Holders of our governance tokens will be able to vote on pool parameters, including parameters such as $\alpha$ and $\beta$ above.
Check out our whitepaper for more details.