How does Uniswap and other AMMs deal with the “poisonous flow” problem?


 135 total views

Traders In all markets, the goal of market makers is to make a profit while processing the incoming order flow-that is, market makers’ requests for trading on their liquidity. To the naive observer, this goal seems very simple. After all, customers are trading on the basis of the bid-ask spread, so they have to pay a premium to the market maker. In addition, market makers can often charge fees to market makers to further consolidate their profitability.

The view that traders are marginal paying customers and market makers are profitable companies is generally correct, but fundamentally wrong. My company Three Arrows Capital and some other non-bank participants have made huge profits in the foreign exchange market over the past few years using such assumptions about the market structure of the bank.

Where is the manufacturer’s yacht?

This table from the US Commodity Futures Trading Commission shows the short-term profit and loss of ES (CME’s SP500 futures) classified by different counterparty types (including HFT and non-HFT).

It should be noted that aggressive HFT (high-frequency trading) made $22 million, hybrid HFT (high-frequency trading) made $13 million, and passive HFT (high-frequency trading) made $500,000. The biggest loss came from opportunistic opponents, which was $31 million.

Obviously, the first question you will ask is-why is passive HFT so poor and why is radical HFT so rich? Isn’t the aggressive company paying the bid-ask spread, while the passive company is making the bid-ask spread? Isn’t it the business of customers and hands? In addition, why are opportunistic traders so unprofitable in a short period of time?

One reason is information. The average trader may not have relevant data about alpha, but the most sensitive trader fundamentally has more information than the average trader. I mentioned this in my previous article on whether the market maker rebate is a free lunch, but the basic idea is very simple, the market maker does not need to trade. They can wait in place for opportunities to appear, and only trade if these conditions are met. But on the contrary, market makers must continuously provide their liquidity to anyone who wants to trade.

Consider the card counting blackjack bettors against the casino. The casino welcomes everyone’s betting action, and most people will lose money because of it. However, only a few large card counters can bankrupt the casino, so the casino spends a lot of resources to ensure that such betting actions do not occur.

In the market, a shrewd market maker who can count cards cannot be driven out of the market. They can trade freely like manufacturers. Regardless of the cost or the perceived price difference, these numbers are irrelevant, because the dealers can put these numbers into their models accordingly. The higher the fee and the greater the spread, the lower their trading frequency. This will increase the cost for everyone.

Opportunistic traders, although the word “opportunism” makes people think they should be informed traders, they actually lose the most on a short time scale because they don’t even try to win on this. They don’t know anything about price movements in the next few seconds or minutes, but just want to fill their positions. Their holding period may be one thousand to one million times longer than the HFT time scale, and they trade on positions rather than tactics.

How to create “Toxic Flow”?

The money flowing out of keen traders is called “toxic” because the market maker will find that they are almost filled up and immediately run out of money. It is like radioactive material, once a participant has a risk, they will immediately pass the risk to another participant.

Traders mainly generate a “poison flow” in two ways: delay and coverage. The incubation period means that they have a faster connection with other markets with similar products, so they can actively purchase market makers’ pending orders when they know that the market in other places is already high or low. Coverage means that they are connected to more exchanges than market makers, so they know market activities that market makers cannot access. Examples of this might include an off-site buying plan to increase every offer on every venue.

In fact, if you ask the top HFT companies, they will tell you that their market-making strategy itself makes little profit, but the real profit comes from being able to use this information in the market-making strategy. For example, in some venues, the spreading time of replenishment information is earlier than the announcement of the transaction time and sales volume. This means that if you want to buy yourself a delay advantage, the only way is to first enter on a small scale as a market maker.

Give 2 specific examples of “toxic flow” (theoretical examples):

A trader combined with 5 venues (exchanges, over-the-counter trading counters, etc.) has a general view of the available liquidity of each venue/counterparty. They found that the quotations of the 3 major exchanges are being substantially cancelled, while the quotations of the 4th venue are being reduced/cancelled. But in venue 5, the price has not changed. They will immediately obtain liquidity from the bond and re-adjust it, thereby immediately avoiding risks and capturing a small spread.

A buyer buys/sells at multiple venues at the same time, and the scale is large enough to change the price. Since liquidity has a “multiplier effect” (liquidity can be obtained from one place and deployed in a second place), this means that many hedging measures of market makers have failed. Therefore, liquidity providers will need to actively change their prices to get their delta back, which in turn provides market makers with liquidity to exit.

What does this mean for AMM (Automated Market Maker)?

Just like in traditional markets, we should be able to predict that keen traders’ transactions with AMMs (automated market makers) will eventually make more money in AMMs (automated market makers) than LPs (liquidity providers). . As long as AMMs can figure out how to keep the cost of insensitive traders at a low level, and the cost of sensitive traders at a high level, this may still be acceptable. In foreign exchange, with the emergence of market makers’ flow characteristics and segmentation, it becomes feasible. What they did was to provide market makers with a way to mark the information of various market makers and turn off the ability of these market makers to trade on them when the traffic was not attractive to market makers. In this way, market makers can continue to tighten quotations for non-savvy traders while reducing the losses of sensitive traders. In terms of foreign exchange, the rise of “last look” liquidity also means that LPs (liquidity providers) can provide non-determined quotations, allowing liquidity providers to choose to accept, reject or requote before execution. Since being an LP (liquidity provider) can be compared to providing buyers with a steady stream of short-term options, the last look at intraday liquidity provides LPs with a choice to protect themselves. But needless to say, this does not exist in AMMs, because the pricing is determined according to the formula (such as x*y=k).

Therefore, in view of the lack of sybil attack defense in the encryption market, minimizing the LP loss caused by the “poison flow” will be a difficult challenge. Faster predictions, more robust and dynamic calibration of pricing parameters, and systematic recovery of profits from predators may be part of the solution.

Having said that, one difference between the “toxic flow” in the traditional limit order foreign exchange market and the “toxic flow” in AMMs is: in the traditional market, the reason for the “toxic flow” is that LPs don’t know they may Provide liquidity in case of loss. In AMMs, the risk of impermanent loss of LP is well known. From the analysis of the chain, we can even say that the vast majority of flows performed on AMMs are “toxic” (arbitrage driven). Fees are a (key) part of this equation, and this problem has now been resolved (if there are no fees, LPs are always in a worse position than buying and holding portfolios). However, there may be a range of fees that allows the LP to achieve a higher rate of return than a portfolio that is bought and held. When everything else remains the same, this is the most important thing for LP, which is equivalent to the bid-ask spread. In addition to the cost constant, other parameters include the shape of the joint curve, the (adjustable) midpoint of the joint curve, and the composition of the pool (away from the constant 50/50 fluidity).

Finally, regardless of whether the market maker is automated or non-automated, its liquidity cannot be prevented from being used to generate alpha for the keen market maker. The sustainable market ecosystem of AMMs should be that passive market makers can outperform users who buy and hold, and benign market makers can still execute at a reasonable price, but there is still basically no question of how to achieve this goal in DeFi. solve.

Disclaimer: does not endorse any content or product on this page. While we aim at providing you all important information that we could obtain, readers should do their own research before taking any actions related to the company and carry full responsibility for their decisions, nor can this article be considered as investment advice or recommendations. Every investment and trading move involves risk, you should conduct your own research when making a decision.