Interpret the advantages, defects and improvement paths of Uniswap and emerging automated market makers from the perspective of professional institutions.
Written by: Parsec Research, research department under parsec.finance Compiler: Perry Wang
The automatic market maker AMM is a relatively new concept, and Uniswap has greatly promoted the prevalence of this concept. Discussions around the pros and cons of AMM usually take two forms: the first comes from experienced traders who cannot tolerate the inefficiency of AMM fund pool transactions; the second comes from people in the cryptocurrency circle who brag about the future of smart contract financial applications It will replace transactions between people or institutions, and advocate lazy liquidity will naturally occupy the world.
As a former algorithmic trader, I am of course empathetic with the first argument, and there have been endless arguments on this topic. But the answer should naturally lie between the two, and it will be a trade-off about the scalability of the core strategy.
AMM curve and order book transactions
What exactly does a market maker want?
Suppose you have some trading expertise and decide to launch a market making robot. The goal of the market maker is simple, buy low and sell high, with the spread in the middle. This sounds simple. Place a buy order at the lowest bid price, and a sell order at the highest selling price. Your buy order is filled first, and the price fluctuates. If someone takes your sell order, you Just reap the spread.
The ideal is full, but the reality is miserable.
First of all, you need to consider the cost of market making. You will definitely have a market making moment with extremely low transaction volume. In addition, for various reasons, a lot of the assets you buy and sell may remain in your hands. Now the value of the asset has fallen. By 10%, the profit you made from the spread is wiped out.
Asset inventory is a key point. The core of any market making is asset inventory management: how long can you keep assets on your books? Doing this well requires both powerful modeling and management of excellent technology. Market-making robots don’t just need to be opened, they can let go without monitoring. Your funds are in danger at all times, and any loopholes may cause high costs.
The stringent requirements of market making are reflected in the huge differences in the size and scale of market makers, and their power and influence are seriously uneven. There are relatively few first-rate market makers in centralized encrypted trading and traditional finance, but they occupy a dominant position in market share.
Successful market makers have two notable characteristics. One is that they have a very high sharpe rate. The crypto field usually does not emphasize indicators such as the Sharpe rate or the sortino ratio. The main reason is that long-term holding of cryptocurrencies, no matter how many currencies you hold, which can grow by more than 10 times, will completely erase Sharp. rate. But the basic model assumptions are valid, and high Sharpe rate strategies are still very good, because they can be leveraged and can reach almost any risk threshold. In traditional finance, the risk infrastructure that we described in an article provides high-quality market makers with huge amounts of credit funds and leverage. These funds are aimed at the return forecast of the strategies they run.
The second key feature is that they are limited in size. Most of these market makers manage their own funds, because raising funds does not actually bring any benefits, in fact it will only dilute their investment returns. Many people believe that a basic law of quantitative finance is that the Sharpe rate and scale of an investment strategy are closely and inversely related, and only a few curves move firmly. It is worth noting that Numerai is a radical attempt to break this law, but so far, this law has been correct in the capital market to a considerable extent. In short, Uniswap’s opportunity for success lies in copying the Sharpe rate of the king of hedge funds, Citadel.
In this way, the AMM trade-off becomes a question of policy scalability. AMM’s desire to expand has stalled and will never keep up with the kinetic energy and data margins of market-making robots. But the benefits of scalability are self-evident. AMM provides a more level playing field for LPs by locking liquidity providers (LPs) in exactly the same trading strategy.
Going back to the temporary market-making robot we talked about before, the reason why it was overturned was because other robots exposed our market-making robots to a disproportionately “toxic” transaction flow. The biggest risk that market makers face is the so-called adverse selection. For example, when a buyer understands information that you don’t know, whether it is a blockbuster order by Coinbase or the intrusion of an encryption protocol, etc…. A good AMM can be proportional Effectively dilute adverse selection to all LPs. Uniswap has done a good job in this regard, but LP has several ways to avoid toxicity. After all, Uniswap LP, with its superb algorithm, actively models price fluctuations and is also an arbitrageur.
Assuming that LPs have the infrastructure, they can predict and discover transaction flows that may lead to substantial price changes (meaning that prices will not “recover” soon), which will cause LPs to suffer some losses. In this case, the algorithmic LP can simply remove the liquidity it provides, execute arbitrage transactions, and re-add liquidity (all are automatically executed!). An audit report published by dapp.org briefly discussed this possibility, but this does not mean that Uniswap is particularly weak in this dimension—in fact, the opposite is true. The key is that LPs need to be constrained, because if some LPs have substantial advantages over other LPs, then the returns for the majority will be suppressed and the scalability of the strategy is limited.
Uniswap should not provide market makers with selectivity and make AMM more and more like an order book transaction, but must rely on its own strength to strictly restrict market makers. An additional method of restricting market makers is to impose a time penalty on LP withdrawals as suggested by Monet Supply.eth and Tarun Chitra. This will filter out LPs with high-frequency funds in and out, and charge 100 for LPs with a block deposit of 0 % Withdrawal fee to completely eliminate LP sandwich attacks. As a result of the restriction of market makers, the strategy is homogeneous in terms of identification.
Consider a new AMM, which allows arbitrage curves to map to each other, and even separate limit orders in mixed transaction flows. In this environment, all market makers use different conditions to operate, so they cannot tokenize their unique strategies, and identification and restrictions on market makers are almost duplicated. LPs play the AMM game of capital flight Will be considered unsafe.
Early research on LP identification has achieved good results (see the research done by Alex Evans of Placeholder), and it may become a truly unique DeFi native element by democratizing Goldman Sachs’ structured products and services. In practice, its income is difficult to determine, and the income is always compared with the benchmark (the impermanent loss) rather than with cash, thus incurring a bad image.
The ideal LP yield curve is a curve with a high transaction volume/liquidity ratio and a high Sharpe rate. Under this circumstance, AMM cannot cap the asset management scale AUM and leverage it, so it has to accept the new LP, flatten the income to a reasonable point, and achieve a level that matches the inflow and outflow. High liquidity.
YFI/ETH LP return rate in ETH (source: parsec)
The industry is usually convinced of this trade-off during discussions, but in reality the results will be probabilistic and depend on the path.
There is no doubt that Ethereum’s gas cost and delay definition give Uniswap a healthy advantage over order book exchanges, which may be the mechanism to clarify the work of AMM. Emerging, highly scalable order book trading solutions such as Serum have traditional market structures and may divert the flow of transactions from Uniswap. There is no doubt that AMM can allow LPs to generally obtain (risk-bearing) transaction fee income. This is very special and can only be achieved when the principles are in place and under strict strategic restrictions.
Parsec Research is the research department of parsec.finance. Thanks to Sherwin Dowlat for his feedback on this article.
