Author: East Asia Venture; Source: X, @EastAsiaVenture

The market structure in DeFi determines the liquidity, pricing efficiency and game pattern between participants during the transaction process.Different trading models (such as AMM, CLOB, dark pool, DFBA, etc.) have their own advantages and disadvantages.Studying market structure helps traders understand the mechanisms behind trading, risks and role distribution of different participants, and thus develop better strategies.Overall,Market structure research revolves around the following three core elements: the balance between liquidity and price discovery, the “impossible triangle” of trading, and the distribution of power of market participants.

This article aims to study the framework of market structure and analyze several major market structures of the current market based on this framework.

Core points

The most important concept of the financial market is called liquidity. The book “Traders and Traders” explains it as follows: Liquidity refers to the possibility that market participants can quickly conduct large-scale trading transactions without causing significant fluctuations in the underlying price.Every market participant likes liquidity, and the first thing to consider for the design of each market structure is liquidity: how to obtain and how to use it more efficiently.

And an often overlooked fact is that the continuous increase in liquidity does not necessarily accelerate price discovery.In markets with extremely abundant liquidity and high order book depth, the price impact of a single transaction is diluted, and the pace of information reflecting in the price tends to slow down.Even if insiders have the advantage of information and have a clear judgment on the reasonable value of the target, the convergence of the price to this range may still be delayed.

The opposite also holds true: the reduction in price discovery difficulty will lead to a decrease in liquidity.Taking insider traders as an example, their profit paths are: ① Have extremely important information; ② Trading first when liquidity is limited and other participants have not yet obtained the information; ③ The market responds step by step; ④ The information is digested and the target price is discovered; ⑤ When insider traders make profits, their returns come from the liquidity provided by other participants; ⑥ The corresponding liquidity is consumed by them.It can be seen that the existence of insider traders reduces the difficulty of price discovery (because they have mastered complete information earlier and accelerated market digestion), but will lead to liquidity losses.

The above phenomenon is often called “there is an adversarial relationship between liquidity and price discovery containing information.”The first point of this constitutes the framework: any market structure must weigh liquidity and price discovery containing information.

Around liquidity, the second fact is the “impossible triangle of trading”: under the condition that liquidity is not infinite, it is impossible for a single transaction to meet the best price, execution time and transaction quantity at the same time.Take common AMMs as an example, usually sacrificing the best price in exchange for execution time and transaction quantity.This is the second point of the framework.

The third key point is the power allocation of market participants.There are many types of participants in mature markets: natural flow traders, market makers, high-frequency traders, arbitrage workers, etc.; different market structures will give different weights to various entities and form a differentiated game environment.In the long run, a healthy and mature market should be the ultimate service target for natural flow traders; an ideal market structure should be able to reasonably adjust the distribution of interests and forces of all parties and create a fair competitive environment for the group – for example, compete more in price and scale rather than in speed (the latter is the advantage of high-frequency trading).

In summary, the research on market structure ultimately comes down to the following three points:

1)Confrontation between liquidity and price discovery containing information;

2)The impossible triangle of transactions;

3)The power allocation of market participants.

Concept interpretation

Natural flow trader: Trader who enters the market for trading due to external economic reasons.This includes hedgers, investors, speculators, etc. Basically, anyone who enters the market for trading with external reasons.

Toxic order flow: has advantages such as information and speed, quickly obtain insufficient market pricing, compete with market makers and natural flow traders, and steal the latter’s profits.

Analysis of the main market structure

AMM

As one of the largest innovative primitives of crypto,For the first time, ammm achieved complete openness of asset launch and transactions(permissionless), because of its atomicity, has become the basic module of Defi “Lego”.

Taking Uniswap v2 as an example, its classic pricing formula is X * Y = k.LP deposits two types of assets x and y, and the formula ensures that each transaction will cause price changes; but at the same time, it will naturally bring slippage, and enable LP to passively accept the trader’s order flow, accompanied by problems such as MEV.

Since the redemption is completed in a single chain transaction, it is almost instant from the perspective of blockchain execution; but the existence of slippage means: if you pursue a better price, the number of transactions that can be completed in a single transaction will be limited; if you pursue a large amount of transactions, you will inevitably bear higher costs in terms of price.

This mechanism cannot suppress the consumption of liquidity by the price discovery process, and the result is often reflected in the impermanent loss of LP.Moreover, LPs can only make “reverse selection” in essence, and cannot make active operations based on information except for withdrawing liquidity; based on this, it can even be considered that transactions on AMM constitute a “toxic order flow” for LPs, which significantly weakens LP’s bargaining and profitability.

After uniswap v2, some new amm class designs appeared, and most of them were optimizing some of the above-mentioned issues;

For example, the AM made by Curve specifically for the stablecoin swap. Behind the smoother curve, it essentially increases liquidity by reducing price discovery. The increase in liquidity will greatly solve it, or alleviate the impossible triangle problem of trading. Therefore, Curve has become the first choice for the stablecoin swap. However, since the price discovery ability has been greatly weakened, it can only be used for stablecoins such as assets with very small fluctuations;

Let’s talk about Uniswap v3 again. The beginning of the white paper reads: “…provides increased capital efficiency and fine-tuned control to liquidity providers, …”; It can support LP custom range to make markets, and improve capital efficiency by concentrating liquidity. This is to enhance the power of LP and reduce the loss of price discovery for LP liquidity. The problem is that this does not change the fact that “Amm’s LP can only be reversed”. It cannot actively adjust prices, but can only be accepted passively. This is a relatively inefficient way;

So overall, AMM transactions are actually not friendly to lp and are even “toxic”.

Regarding MEV, the frequent arbitrage of AMMs does not only originate from AMM itself, but also is related to the underlying public chain mechanism.In short, on-chain transaction applications lack control over their own transaction sorting, resulting in a large number of opportunities to be exploited.Further discussion will be made in the future.

CLOB

Order thin is the most classic market structure of tradfiCompared with amm, its biggest function is to truly give market makers the ability to participate in the active choice and adjust prices, rather than being able to passively accept pricing from traders. In other words, the asset efficiency of market makers has been greatly improved;

At the same time, the actively selected design makes clob more inclined to price discovery in terms of “the trade-off between liquidity and price discovery” than amm, and has stronger pricing power;

In the impossible triangle of trading, traders can choose to ensure the best price and transaction quantity, but then the order will often become pending orders and enter the order thin, which will take longer to complete;

If you want to ensure the best price and execution time, or in other words, traders want to make an order immediately as a ate order, then you cannot guarantee the transaction quantity. After all, the number of pending orders that match the ate order in the corresponding order is limited;

Clob has two very important features: continuous matching, time priority;

Continuous matchmaking is a trading period, which means matchmaking is “as you can match”.Whenever a new order arrives at the matching engine, the system immediately tries to trade it with the counterpart price in the book; if the remaining orders are not completed, the unsold part will be placed in the order book and wait for the next transaction to arrive.There is no fixed “call auction moment” (this is another common trading period in the market structure), and orders may trigger transactions at any time;

Time priority, often combined with price priority, is called price-time priority; matchmaking is better or worse than price, and in the same price range, then queue according to the order of arrival: first come first transaction;These two characteristics will produce a concept called “queue position is valuable”. Many traders will invest a lot of resources to improve speed and grab the most valuable queue position at the forefront, which is very likely to trigger an arms race in speed;

In the blockchain environment, these effects are further amplified: First, on-chain data is disclosed, resulting in the exposure of orders/intents, and the potential price impact is visible to all participants; second, transactions are reorderable because the underlying block producer/block builder has the right to reorder.The resulting MEV attacks have caused the natural flow traders of CLOBs on-chain to bear high implicit costs.

Dark pool

Dark pool is also a classic trading place in tradfi,Features are: Hide order book, anchor external reference price, and not disclose the depth to the outside world;Many of the service targets are institutional investors, allowing them to execute some large-scale orders without disturbing the market (if the order flow is made public, it may cause some unfavorable price fluctuations), so as to conceal liquidity and trading intentions;

Although often criticized for fragmented liquidity and lack of transparency, it is undeniable that dark pools are an indispensable market structure for institutions that need to deal with large transactions;

So, why does DeFi also need a dark pool?

Based on previous discussions: Although AMM provides the most open liquidity, LP passively endures toxic flows and has low capital efficiency for a long time; although CLOB gives market makers the right to choose actively, the orders on the chain are transparent and re-arranged, resulting in natural flow traders being exposed to the costs of MEV and rushing; large-scale traders have almost nowhere to hide in such markets: they will be harvested by slippage in AMM, and in CLOB they will be pre-empted due to order visibility;

The dark pool provides a solution, using private liquidity, non-public quotations driven by professional market makers, providing traders with more favorable quotations, and varying degrees of privacy protection;

In fact, there are many implementation paths for crypto’s dark pool:

1. Dark Amm on Solana (Private Liquidity Pool)

   Through professional off-chain market makers to continuously quote (the mechanism is opaque) and on-chain settlement, the transaction is hidden before submitting to the verifier. Most users reach the contact between the aggregators; privacy level: orders are hidden before settlement:

2. private tx relays of ethereum

   Typical examples are Flashbots.Users or searchers bypass mempool and send transactions directly to validators. The effect is similar to Solana’s dark AMM, both of which can hide orders before transactions are completed; the difference is that they are not usually via an aggregator.

3. Privacy Agreement on ethereum

   For example, Railgun uses on-chain zero-knowledge proof and anonymous pool to allow orders to hide capital flows before and after settlement.

4. Native dark pool on the chain

   For example, Renegade deployed in Arbitrum follows the traditional dark pool logic: match according to the midpoint price of Binance to avoid spreads and slippages; transaction details are matched in the zero-knowledge circuit, and the results are not disclosed until settlement.

Among them (3) and (4) are closest to traditional dark pools in terms of privacy.Given the natural transparency of blockchain, the implementation difficulty is not low.

Regardless of which implementation path, crypto’s dark pool is more inclined to use the weakening of price discovery to strengthen liquidity protection;

The privacy protection of the dark pool, no matter what degree, can protect traders from price risks such as mev. At the same time, the opaque quotation mechanism fundamentally creates an information blocking (which also means that price discovery is difficult). In the dark pool, the information advantages are greatly reduced, and market makers do not have to worry too much about the risk of reverse selection brought by toxic order flow, so it further gives traders a price advantage;

The biggest problem with dark pools is that traders do not know how likely the execution is to be successful, because no matter which implementation path is, since pricing is opaque, the market depth is also opaque, which may bring traders an intuitive feeling of execution time, etc.;

There is another problem with the dark amm type on solana: liquidity is dispersed and is mainly aggregated through aggregators. From the perspective of the overall market, it is a reduction in capital efficiency.

DFBA

DFBA is a new market structure proposed by jump crypto,Full name is: Dual Flow Batch Auction, dual-stream batch auction mechanism;

From the design concept, it refers to the previous market structures, “take its essence and remove its dross”, aiming to create a market structure that is truly born for natural flow traders;

Based on previous discussions, the natural efficiency of the on-chain market structure such as amm is not high enough, so we need to introduce on-chain clobbers. Due to the blockchain environment, the characteristics required for on-chain clobbers: continuous matching and time priority open the door for delayed arbitrage, MEV exploitation and reverse flow, thereby pushing up the costs of market makers and natural flow traders;

DFBA was born to solve these problems:For market makers, it can reduce the risk of toxic flows and create a level playing field through unified pricing;For organic flow traders, it provides smaller spreads and deeper liquidity by addressing the structural inefficiencies that CLOB brings to the diversified market;

In CLOB, taker and maker are defined by the order arrival order book sequence: the first one is a pledge order, and the next one is a taker, and the transaction is matched with the pledge order that arrived previously.

Although the term taker/maker is used in DFBA, it is no longer distinguished by arrival time, but is based on the role and intention of the order sender: if it is a market maker and the goal is to provide liquidity, then its order is a pending order; if it is a natural trader (or may be a market maker) and the goal is to obtain liquidity and seek immediate execution, then its order is a pending order.

DFBA will first collect all limited-price trading orders within a fixed period (about 100 milliseconds) (including those that were not sold in the previous round of auction), and divide them into pending order groups and eating order groups.Then there will be two separate and simultaneous auctions:

• “Bidding” between buying orders and selling orders

• “Inquiry Auction” between placing orders and buying orders

In each auction, find a liquidation price that maximizes the matching amount in the auction, all matching orders are sold at the same price, and all unmatched orders are left at a more favorable price;Then repeat the above process;

There are some points worth paying attention to in this mechanism:

1. DFBA uses batch auctions and sets prices, rather than continuous matching

   Continuous matchmaking of clobs will create time priority, which will trigger an arms race among participants in speed, but in the end, the biggest winner is HFT. Market makers often cannot match HFT in speed, and can only choose in reverse and bear the losses caused by toxic order flow. In the end, this loss will allow market makers to transfer part to natural flow traders by expanding the bid and offer spreads and other means; (on the chain, due to the mev problem, the losses suffered by natural traders will be more serious).The batch auctions used by DFBA are essentially within a fixed period. No matter whether they come first or later, there is no difference between all orders, thus eliminating the concept of time priority, greatly reducing the risk of toxic order flows suffered by market makers, and also a means of protection for liquidity. In the end, natural traders can also benefit from better bid-ask spreads provided by market makers, and these measures greatly reduce the hidden costs of user transactions that cannot be eliminated by traditional clobs such as delayed arbitrage, and mev;

2. Two-way auction

   In the clob, market makers A and B are both participants of “maker identity”, but A can switch to takers at any time to use time priority to pick out the outdated/distorted orders of opponent B. This is equivalent to a maker and maker transaction, or a “market makers fight each other”: market makers use temporary acts as takers to hit the other party’s maker price, and whoever is faster and has a better queue position will be able to make this “speed money”, which will increase the risk of the other party’s reverse selection, forcing everyone to widen the spread and narrow the depth;Under the DFBA mechanism, the attributes of the orders are confirmed when submitted, and are divided into two groups, and two groups of auctions are conducted. Taker can only trade with Maker; Maker orders focus on providing liquidity, and taker orders do not have time priority to earn speed money from other market makers due to unified pricing and other methods, thus fundamentally eliminating the transaction between Maker and Maker, weakening the game between market makers, and allowing market makers to focus more on providing better services (better quotations);Strengthening the role of market makers as liquidity providers, and the recipients benefit from clear and competitive execution; the distinct roles can establish direct and efficient interactions between liquidity providers (market makers) and liquidity demanders (recipients) to enhance price discovery, thereby improving the pricing efficiency of the entire market;

3. The liquidation price is between the best pending order price and the worst eating order price (included)

   Both makers and takers have a price improvement effect, which is equivalent to enhancing the power of market makers, while traditional clobs will only have a price improvement effect on takers;

4. The time period is 100 milliseconds

   According to the original description, jump concluded through research that the impact of 100ms on price volatility is almost negligible, and for natural flow traders, this time difference has no effect on this time difference

From a higher level perspective, DFBA’s core innovation lies in a more reasonable allocation of market participants’ power (more focused on serving natural flow traders): its mechanism significantly weakens the speed advantage of HFT, reduces the impact of toxic order flows that market makers suffer, and structurally lowers the reverse selection.As a result, market makers tend to report tighter spreads and provide thicker depth; natural flow traders therefore benefit from better spreads and depths, while MEV risks are significantly alleviated.

Liquidity is better protected, at the cost of price discovery being discrete in 100ms one by one, and jump research also shows that this period does not affect the volatility of prices, and there is no perception of natural flow users.

Market structure and public chain

Defi’s market structure is closely related to the development of public chains;

Ethereum’s early throughput/latency and public memory pools made it difficult to scale the on-chain order book with “high-frequency hanging + queue game”, but it can give birth to applications such as amm with relatively low performance requirements; and now high-performance chains like solana regard low latency + high concurrency as first-class citizens, so the active quotation and other designs of order book become feasible and natural;

The MEV in the blockchain system is essentially derived from the control of the transaction sequence in the public chain and nodes rather than the application side; this arrangement is not reasonable in terms of mature and healthy financial markets.Solana is obviously aware of this problem and introduced a new block construction method BAM (Block Assembly Marketplace) into the recently proposed “Internet Capital Market” plan. One of the effects is to return a considerable part of the sorting sovereignty to the application itself; for transaction-oriented scenarios, this model is closer to the execution layer “born for the market structure”.

Therefore, innovative designs such as DFBA may be the first to be implemented in Solana, using BAM’s plug-in capabilities to build a new Internet capital market for natural flow traders.