This is the first in a series of posts on the dynamic fee policy for AMM liquidity pools.
It is well known that Uniswap V3 liquidity providers are inconsistently profitable at best. Furthermore, analysis we conducted a few months ago showed that even when conducting detailed analysis per wallet, it is extremely difficult to obtain systematic predictors of profitability of LP positions minus divergence losses (aka “immanent loss”) identify.
This is not a good situation! Abundant on-chain liquidity is a cornerstone of decentralized finance, which allows for easy compatibility between different protocols. Additionally, liquid token exchange is one of the most popular activities on the chain. If liquidity providers are systematically losing money, it would be wise for them to stop providing on-chain DEX liquidity, effectively shutting down a multitude of cryptoeconomic use cases.
These considerations motivate the following question: Is it possible to build a DEX that will help liquidity providers make money? Over the long term, a DEX that is optimized for LP profitability will attract the highest and most sustained liquidity compared to DEXs that are not, resulting in not only LP gain but also a superior user experience.
In this series of articles, we will encourage building a basic, provably dynamic fee policy that is superior to staying in a static Uniswap fee tier, allowing LPs to earn higher and more consistent profits minus divergence losses. Due to the inherent complexity of this topic, we’ll attempt to step through the details of a dynamic fee policy over the course of several posts. In the first part, we just want to briefly present our considerations as to why a dynamic fee policy would be helpful and, most importantly, to present our proposed empirical framework for setting up a dynamic fee policy in the context of the already high level of liquidity at other decentralized trading venues.
To understand the need for a dynamic fee rate, it is instructive to examine the simple case of ETH/USDC liquidity. Notably, the two largest non-stablecoin/stablecoin pools on Uniswap V3 are the USDC/ETH 0.3% and USDC/ETH 0.05% pools:
Broadly speaking, swaps between ETH and USDC on the Ethereum mainnet flow through either the 0.3% or the 0.05% pool – generally the latter due to its lower fee rate, but the former still manages to make a sizeable one to record volume. There is also a 1% fee pool, albeit with much lower TVL and swap volume.
We immediately see a clear motivation for a dynamic fee rate: there is significant overhead associated with increasing the “price” of the liquidity you provide! Suppose a liquidity provider anticipates higher market volatility over the next week. Accordingly, they may wish to shift their liquidity position in the 0.05% pool to a higher fee tier (roughly equivalent to a larger bid-ask spread in a traditional order book). This requires burning their liquidity position in the 0.05% pool and establishing a new liquidity position in the 0.3% pool, resulting in fixed gas costs and significant logistical complexity. As such, liquidity providers could miss out on highly profitable fee generation opportunities if they are not well equipped to handle the complexities of active position management.
It is important to note that the idea of a dynamically adjusting liquidity fee is not new. For example, a recent article by Milionis, Moallemi, Roughgarden, and Zhang suggests that AMMs may want to adjust their trading fees based on market conditions:
Concrete suggestions have also been made for specific models to calculate an ‘optimal’ dynamic fee, such as the work of Guillaume Lambert, which proposes a fee that is dynamically adjusted to target a constant implied volatility. In particular, however, the discussion of dynamic fees has typically focused on a theoretical environment, assuming that the DEX operates in a relatively stylized environment, with no other competing sources of liquidity apart from perhaps a reference exchange where price discovery takes place.
Although the theoretical question of which is the “best model” of dynamic fees remains important, we propose to first analyze a simpler and possibly more empirically understandable question. Suppose a new DEX is created with a dynamic fee policy. For simplicity, consider only ambient liquidity. What are the considerations for the marginal unit of additional ETH/USDC liquidity? We see immediately that we need to consider the overall state of the liquidity market – namely, the swap fee should probably not be too much lower than 0.05% and not too much higher than 1%. At the 0.05% and 1% fee rates, Uniswap V3’s liquidity is so swamped that venturing too far out of this range for a marginal unit of liquidity is likely to be inefficient. If the fee is well below 0.05%, liquidity will likely capture the same swap volume at 0.05% or slightly below; On the other hand, if the fee is much higher than 1%, the liquidity is likely to be crowded out by the large amount of surrounding liquidity available in the 0.3% and 1% Uniswap pools.
We are now beginning to agree on a simpler operationalization of the problem of setting a dynamic fee rate, at least (1) in the case of ETH/USDC and (2) assuming that the amount of liquidity added is only a small fraction of the amount available accounts for Uniswap. The observation that a dynamic fee policy in this case must empirically consider the dynamics of the already existing liquidity on other trading venues allows us to turn a challenging theoretical question into an easier to understand empirical question, and all the conclusions we draw here allow us ultimately help in relaxing these two assumptions.
To be specific, we can ask the following question: For a marginal unit of additional liquidity, and assuming Uniswap liquidity pools remain as they are, what dynamic fee rate should then be set at? We suggest this fee be well estimated by the following procedure. At any point in time, we might be asking the following question: which of the three ETH/USDC Uniswap pools (0.05%, 0.3%, 1%) will incur the highest fees for one unit of ambient liquidity in the next few minutes? ? We should be able to estimate a probability distribution across the three pools: for example, we can have 80% confidence that the 0.3% pool will be the best performer over the next 10 minutes, 15% confidence in the 0.05 % pool and 5% confidence in the 1% pool. There are various options for choosing the dynamic fee, which of course result from the concept of such a probability distribution; For example, we could choose a weighted sum of the three Uniswap pool fee rates: 0.8*0.003 + 0.15*0.0005 + 0.05*0.01 = 0.002975 or a swap fee of 0.2975% ; Alternatively, we could simply choose the fee rate with the highest probability estimate, in this case 0.3%.
Note that we have reduced a difficult theoretical question to something that can even be formulated as a simple regression problem. One could imagine collecting a number of historical variables at each point in time: statistics about the three Uniswap pools at past points in time, or even external metrics such as on-chain activity or CEX funding rates. With this data, one can then predict the fee growth for each of the three Uniswap pools over the next time interval, the likelihood that a particular pool will have higher fee growth than the other two, etc. The resulting predictions can then inform us to choose a dynamic fee policy. In the special case where the dynamic fee is chosen to correspond to one of the three Uniswap fee tiers, the performance of the dynamic fee policy can easily be compared directly to the static fee tiers by simulating the growth of surrounding liquidity that is changing moved between the three static pools according to our dynamic fee policy.
Furthermore, the scope of these results is not necessarily limited to just ETH/USDC pools! Rather than analyzing each liquidity pool as its own isolated environment, we can capitalize on the well-known fact that further out on the risk curve, altcoins often trade to some extent like “leveraged ETH”, suggesting that the same statistical signals are in place as the Predicting fee tier performance in ETH/USDC static liquidity pools can also retain their value in liquidity pools for, for example, ETH/LDO or ETH/FXS. In general, since market volatility is often downstream of factors that affect all markets, such as macroeconomic developments or crypto-wide regulatory considerations, we should assume that the same signals that predict fee tier performance in static ETH/USDC pools can be generalized corresponding conversions to liquidity pools for other types of assets. Therefore, in-depth empirical analysis of ETH/USDC fee dynamics — in a market that can plausibly be claimed to represent the purest “crypto market beta” — can in principle lay the groundwork for a dynamic fee policy that can be applied at will Pairs of assets ranging from pairs of custodial stablecoins to highly speculative liquidity for new protocol tokens.
In the following articles in this series, we will follow the course of the empirical analysis outlined above. In particular, we will examine the dynamics of fee growth between the three ETH/USDC static tiers, construct a dynamic fee policy that plausibly outperforms each of the static fee tiers, and characterize its generalizability over other dimensions (to different chains, different token pairs, etc.). .). Stay tuned for the next episode – and in the meantime, any thoughts are of course welcome!
-0xfbifemboy
Learn Crypto Trading, Yield Farms, Income strategies and more at CrytoAnswers
https://nov.link/cryptoanswers
Comments are closed.