As proof of stake on Ethereum becomes closer, new PoS blockchains (Cosmos, Tezos and others) have been launched, there are and will be stakeable ERC20-compatible tokens (NuCypher, Livepeer, Ren Project, Skale, KEEP, The Graph), staking-as-a-service looks more and more as a lucrative business.
However, centralized exchanges (such as Binance) quickly jumped on the bandwagon of this business model and offered staking while the underlying token can be traded on the exchange. This created a situation of tough competition for cryptocurrency stakers while harming decentralization.
In order to fix this, the concept of Liquid Staking was created. There are a few groups working on this concept and collaborating in Liquid Staking working group, with the groups including but not limited to StakeDAO, StakerDAO, P2P and others. However, while it's certainly possible to tokenize stake, how easy it is to do automatic market making for it?
Staked assets behave very similarly to interest-bearing tokens (such as cTokens): they have a uniformly growing "target price". Real price can go slightly above and below the target, and that's where good market-making opportunities appear.
Let's consider for the sake of example staking NuCypher's NU tokens which cannot be quickly unbonded, and let's call the staking derivative which grows as more NU are created stakedNU. In NuCypher, you can restake - reinvest staking compensation into staking process, or not.
In the proposed arrangement, 5% of NU is not staked but placed in automatic market-maker created by Curve, to do MM between stakedNU and NU. The stakedNU token is created for all other (95%) of NU which are staked.
Let's consider two situations: price of stakedNU/NU being lower and higher than the target price.
Suppose that at some point the target price p*=1.10, and the actual price appeared to be lower: p=1.05. As soon as the price becomes significantly lower than p*, we stop restaking NU, so that staking process has free (unbonded) staking compensation. This compensation is used to buy off the discounted stakedNU from the market created by Curve, and the obtained stakedNU are immediately burned, so that all the underlying NU now correspond to less stakedNU. Since p<p*, the increase of amount of staked NU per stakedNU is more than the amount of NU not restaked, so this process makes a profit for "liquid stakers" equal to:
[amount of NU compensation] * (p* - p) / p.
If it appears that p>p*, it's instead more profitable to sell stakedNU to the pool. The staking DAO takes stakedNU from reserves and sells those to the pool for more NU that the actual underlying NU the sold pieces had. Those NU are immediately bonded and more stakedNU than was originally taken are created. However, we need to redistribute profit from this arbitrage, so we burn the excess of stakedNU to do that. The burnt amount is:
[amount of stakedNU sold] * (p - p*) / p*.
While NU was just chosen to be a good example, the mechanism above can be applied to many digital assets: ETH 2.0, Livepeer, Ren Project, Skale, KEEP, The Graph and others. Depending on restaking or unbonding dynamics, the source of funds for buying discounted staked tokens could be different: if unbonding time is fast, for example (several days for Livepeer), a larger amount can be unbonded, and the price can be quickly brought up to the "peg".
In conclusion, the process described above allows to use Curve to do automatic market-making for any staked asset, and use arbitrage with staking/unstaking to push the price of the liquid stake back to its fundamental value and make money in that process at the same time.