A comprehensive understanding of the evolution and development trends of the DEX paradigm

Author: Kylo@Foresight Ventures

Tips:

• AMM and RFQ actually represent the difference in thinking between DeFi and TradFi.

• AMM improves capital utilization efficiency through LP leverage.

• The RFQ model has a natural advantage for cross-chain trading.

• The introduction of Core Pools will significantly change the yield structure of Balancer.

• Liquidity incentives related to price ranges can reshape the form of Liquidity Position.

This article mainly discusses the development history and possible future directions of DEX. The research on Perpetual Trading DEX has been described in detail in “Perpetual DEX: The Path of LP Productization”, while this article focuses on the spot DEX field, covering directions such as aggregators, veToken models, and Uni V3 models of DEX.

I. Spot Trading Aggregators

1) AMM vs. RFQ

Before diving into the main content, it is necessary to review the history of AMM vs. RFQ. Today, with various foundational frameworks of DeFi already relatively complete, we seem to have become accustomed to AMM DEX. However, the initial form of DEX was not what we see now; the order book system and RFQ were the earliest methods for on-chain asset trading. In the iterative evolution of DEX, RFQ has been surpassed by AMM in competition. The event of AMM vs. RFQ actually reflects the difference in thinking logic between DeFi and TradFi. From the perspective of TradFi, the essence of finance is the process of improving capital utilization efficiency. However, compared to the RFQ model, the capital utilization efficiency of the AMM mechanism is extremely low, and pricing assets can easily expose them to price fluctuations, which means AMM should be inferior to the RFQ model. But the actual result is just the opposite; the AMM model has progressed further down the path of DeFi, ultimately even influencing the model of Perp Trading.

This actually indicates that thinking about the development model of DeFi from the logic of TradFi is biased. As a financial system native to blockchain, DeFi must have its own underlying logic regarding finance, which determines its development. We can roughly summarize it as follows:

• Democratization of LP market making

• Transparency of pricing power

• LP leverage

The first two characteristics are features of DeFi endowed by the Web3 ideology, while LP leverage can actually be integrated with the logic of TradFi. In the logic of TradFi, improving capital utilization is the essence of finance. Since the TradFi system has now developed to a relatively mature state, the combinability between different TradFi applications is not that strong, so for TradFi, improving capital turnover means improving capital utilization efficiency. The RFQ model is a typical trading model that improves capital utilization through high capital turnover. But for DeFi, there are two ways to improve capital utilization: increasing capital turnover within a single project and increasing asset combinability between multiple projects. This means that DeFi can improve capital utilization efficiency not only through capital turnover but also by leveraging combinability to stack assets and achieve capital reuse through LP leverage.

Therefore, if one understands the above logic, it can be easily deduced that P2Pool in Perp Trading will become a trading model on par with order books—GMX and GNS are merely extensions of Uniswap in the development logic of DeFi.

In simple terms, AMM and RFQ achieve efficient use of capital from two different angles: the former through combinability and the latter through high turnover. Due to the fact that the entire endogenous DeFi architecture had not yet been established in the early stages of DeFi development, improving capital utilization through DeFi's combinable model was more meaningful compared to merely increasing capital turnover. This is why AMM's development in the early days of DeFi outperformed RFQ. However, the logic of development is often indirect; AMM's dominance in trading is not absolute. Currently, the method of achieving capital reuse through combinability seems to have reached the limits of leverage, and AMM DEX has begun to focus on improving capital turnover, at which point the RFQ model, representing high capital utilization, is once again brought to the forefront, with the narrative still centered around the issue of capital utilization.

2) The Complement of RFQ Model to AMM

As mentioned above, the advantage of the RFQ model lies in high capital turnover and lower slippage, and we can see the RFQ model in various trading aggregators like 1inch and Paraswap. The RFQ model is introduced as a complement to AMM trading in aggregators, and it also contributes to increasing the complexity of the aggregator's token model. The economic model of trading aggregators is difficult to design; if they implement trading commissions, they can easily be undercut by other aggregators in price wars, so trading aggregators generally do not charge commissions. The introduction of the RFQ model provides some utility for the protocol tokens. Taking 1inch as an example, since the RFQ model requires the introduction of market makers, market makers need to stake a certain amount of Para to be distributed among different market makers before executing the RFQ function on 1inch, with the distribution ratio determined by the $Para voting ratio.

3) RFQ as a Complement for Cross-Chain Trading

Currently, the common cross-chain trading model is still AMM, but the RFQ model has a natural advantage for asset cross-chain transactions. The AMM cross-chain liquidity solution faces issues of liquidity fragmentation and efficiency: each additional chain requires adding duplicate liquidity pools on that chain, and exchange rates are also heavily influenced by pool depth. However, using the RFQ model with market makers as intermediaries avoids the aforementioned liquidity fragmentation issues. The specific solution is to introduce Brokers on different chains, managed by a public ledger that oversees the accounts of Brokers across chains. When users have asset cross-chain or cross-chain swap needs, they only need to connect with the corresponding Broker on the relevant chain.

Of course, the description above regarding Brokers participating in cross-chain trading is a simplified narrative; the actual process is more complex. Current projects implementing cross-chain trading or asset cross-chain using similar methods include WooFi DEX and UxUy. WooFi DEX has connected multiple networks and allows cross-chain swaps between various currencies, while UxUy is still in the testing phase.

II. Changes in the Yield Structure of the veToken Model — Introduction of Core Pools

The veToken model is a term for a type of DEX, representing a series of DEXs that determine the release ratio of tokens in different pools through Gauge Voting, specifically including Curve, Balancer, and various Ve(3,3) DEXs. Taking Curve as an example, the yield closed-loop logic of the veToken model lies in the fact that holders of veCRV can obtain discounts on future bribery from all B-side protocols for veCRV, and the continuous issuance of $CRV essentially only overdraws future bribery revenues. However, as DEXs, Curve and Balancer themselves do not seem to perform the trading functions of DEXs; they are more liquidity protocols that accumulate liquidity in the market through continuous token subsidies, with tokens empowered by B-side bribery. This model actually leads to a waste of liquidity, as most of the liquidity "lying" in Curve does not circulate in the market.

This issue is actually one of the controversial points of the veToken model: it attracts a large amount of liquidity through high subsidies, but this liquidity does not utilize its high liquidity function, thus failing to generate external revenue. Therefore, for the entire Curve system, the external revenue that the system can obtain, apart from a very small amount of swap fees, is only the bribery from B-side protocols for veCRV.

Token subsidies can be understood in economic terms as the system overdrawing future external revenues to distribute to users as debt, which can only be repaid by external revenues obtained by the protocol. In the case of Curve, the debt issued by the system is the $CRV rewards issued each period, while external revenues come from B-side bribery for veCRV and some swap fees. If external revenue sources for Curve are added to the entire economic design of Curve, the DEX model of Curve will significantly improve the system's debt level.

Thus, the current state of Curve's DEX can be summarized as follows:

• The liquidity within Curve has not been fully utilized.

• Adding external revenue sources for Curve can significantly improve the system's debt level.

The solution is therefore obvious: Utilize the underutilized liquidity within Curve's DEX to add external revenue sources for Curve.

1) Introduction of Boosted Pools

The first DEX to propose the above solution is Balancer, which introduced the Boosted Pool mechanism aimed at increasing LP yields. Boosted Pools are a more advanced form of liquidity created by combining Composable Stable Pools, Weighted Pools, and Linear Pools. Weighted Pools are AMM pools adjusted for weight; a common AMM weight is 50/50, while Weighted Pools can freely adjust weights when the pool is established. For example, Radiant Capital's dLP is set with an 80/20 weight for the RDNT/ETH trading pair; Linear Pools are liquidity pools specifically set for a certain asset and its derivatives, such as DAI and aDAI. Typically, Weighted Pools are paired with Linear Pools to form Nested Linear Pools. Taking DAI and aDAI as an example, the Nested Linear DAI Pool is a liquidity pool composed of DAI - aDAI with a 20/80 weight, and the issued LP token is represented as bb-a-DAI. Similarly, there are corresponding bb-a-USDC and bb-a-USDT for USDC and USDT. If bb-a-DAI, bb-a-USDT, and bb-a-USDC are combined into an LP pool, this pool is referred to as a Composable Stable Pool, and its issued LP token is called bb-a-USD.

Although bb-a-USD is an LP token, it still follows the ERC-20 token standard, allowing it to be paired with various assets in AMM pools. A typical example is the bb-a-USD / ETH trading pool. The characteristics of this trading pool include:

• The pool allows for two-way asset swaps between ETH, DAI, USDT, USDC, aDAI, aUSDT, and aUSDC.

• Due to the existence of aTokens, bb-a-USD generates interest at all times.

This type of AMM trading pool that generates interest continuously is referred to as Boosted Pools in Balancer's definition.

The high capital utilization efficiency of Boosted Pools mainly stems from their ability to achieve asset leverage through the issuance of LP tokens. The initial AMM pool needs to maintain sufficient liquidity depth to ensure price stability during asset swaps. This means that most of the assets in the AMM pool do not perform the function of asset swaps but rather stabilize the price of stablecoins. The innovation of Boosted Pools lies in splitting these two functions of assets; part of the assets continues to perform the function of asset swaps, while another part used for stabilizing stablecoin prices is taken to external protocols for yield, and LP tokens are issued through external protocols to perform the original AMM pool's stablecoin price stabilization function.

However, the high capital utilization efficiency of Weighted Pools can only benefit LPs and cannot improve the overall system's debt level from the protocol level. Therefore, after the BIP-19 proposal, Balancer introduced Core Pools, primarily aimed at correcting the defect that Boosted Pools could only benefit LPs, allowing the Balancer protocol itself to also benefit from Boosted Pools.

2) Introduction of Core Pools

The core proposal of BIP-19 includes the following points:

• Select some Weighted Pools to form Core Pools.

• The yield generated within Core Pools is distributed as follows: 50% to LPs, 17.5% to BalancerDAO, and 32.5% as Bribes for bribing veBAL holders to vote for Core Pools, thereby increasing the TVL of Core Pools.

• 25% of the Bribe funds are directly allocated to the bribery market of Aura Finance.

• The remaining bribery funds are allocated to Hidden Hands under Redacted Cartel.

The introduction of the Core Pools mechanism is actually guiding the future development trend of Balancer, encouraging other liquidity pools to transition towards Boosted Pools as much as possible. Since the BAL released for liquidity incentives each week is shifting towards Core Pools, the amount of $BAL available for liquidity incentives for other liquidity pools will significantly decrease. Therefore, for these liquidity pools, joining Core Pools as much as possible is the optimal choice for maximizing benefits.

Before BIP-19, the distribution of protocol revenue among stakeholders was as follows:

• 50% of trading fees to LPs, 17.5% to BalancerDAO, and 32.5% to veBAL holders.

• 100% of Bribes allocated to veBAL holders.

After BIP-19, the distribution of protocol revenue is as follows:

• 50% of trading fees to LPs, 17.5% to BalancerDAO, and 32.5% to veBAL holders.

• 100% of Bribes allocated to veBAL holders.

• The yield generated within Core Pools is distributed as follows: 50% to LPs, 17.5% to BalancerDAO, and 32.5% as Bribes for bribing veBAL holders to vote for Core Pools.

The yield of Core Pools is diverse, coming from lending protocols like AAVE, and can also come from LSD assets like stETH. From the perspective of asset swaps, Core Pools' LPs use 50% of their yield to obtain token incentives bribed with 32.5% of yield; as long as the value of this portion of BAL is higher than the original 50% yield, LPs are overall profitable. For veBAL holders, this is due to the real value of BAL.

Overall, Core Pools have changed Balancer's yield structure, increasing external yield asset revenues as a source of income for the Balancer system. Compared to other veToken model DEXs, it is more beneficial in reducing the overall system's debt.

III. Uni V3 with Liquidity Incentive Model — Reshaping the Form of Liquidity Position

During the Arbitrum airdrop, TraderJoe captured a large volume of trading related to $Arb due to the uniqueness of its V2 algorithm, making the narrative of Uni V3 a hot topic in short-term DeFi, while also bringing Timeless Finance, which has Uni V3 liquidity incentive attributes, to the forefront of speculation. Thus, voices have emerged stating that "the Uni V3 model with liquidity incentive properties will lead the next narrative of DEX development." Currently, there are two forms of liquidity incentives on-chain regarding liquidity incentives and bribery: one type is related to LP liquidity depth, and the other is price-related, concerning liquidity incentives for specific price ranges. These two liquidity incentive models have significant differences. The former primarily aims to increase LP liquidity depth, regardless of price; while the latter requires price guidance, mainly achieved by changing market-making profits at different price ranges.

From the perspective of market scale, deepening liquidity is a demand that all protocols need to address, while protocols requiring price guidance are limited to certain pegged assets, such as ETHLSD and stablecoins. Therefore, from the breadth of demand perspective, the demand for deep liquidity is broader, while the demand for price guidance is relatively limited. According to the classification of the above models, the liquidity incentive models of Curve and Timeless Finance targeting Uni V3 can be categorized as the former, focusing on providing deeper liquidity; while Maverick Protocol should be classified as the latter, providing more efficient and accurate liquidity.

Currently, there are two pain points in liquidity incentives for the AMM model:

• For the Uni V3 AMM model, only the liquidity near the price range is active, while liquidity outside the price range is almost unused; however, in liquidity depth incentives, these two types of liquidity with different activity levels receive the same rewards.

• For the Curve model, providing liquidity depth incentives for pegged assets may exacerbate the decoupling of that pegged asset.

The second point may be difficult to understand. Taking stETH as an example, before the Shanghai upgrade, since ETH deposited in the beacon chain could not be withdrawn, stETH had a liquidity discount, with a price of about 0.98 ETH. At this time, if one wants to obtain rewards from the stETH - ETH Curve LP, they must add liquidity to the Curve pool at the rate of 0.98 ETH = stETH. The added liquidity does not improve the decoupling of stETH; instead, it adds deeper liquidity at the original price. This could hinder the price of stETH from returning to its pegged state.

Maverick Protocol has provided its own solutions to the above two pain points, referred to as Boosted Position. The official description uses "surgical" to vividly illustrate the guiding effect of Boosted Position on liquidity. Again, taking stETH as an example, the "surgical" function of Boosted Position for stETH mainly manifests in two aspects:

• Providing additional incentives for LPs that automatically adjust their positions to follow the price of stETH to the right.

• Providing additional incentives for LPs making markets in the 0.98 - 1 range.

Therefore, it can be seen that Maverick's liquidity guidance mainly focuses on two aspects: guiding the direction of price changes and guiding the liquidity market-making range. Through additional yield incentives, it reshapes the form of Liquidity Position while ensuring that liquidity incentives are concentrated around the spot price, improving the capital utilization efficiency of the funds used for incentives.

The next DEX likely to introduce specific price range liquidity incentives is TraderJoe, mainly because its unique V2 mechanism can seamlessly integrate range liquidity incentives. TraderJoe's special mechanism includes the following points:

• The price range scales are pre-planned, and LPs can only add LP within the planned scale range.

• The interval formed between two adjacent scales is called 1 bin, with each bin corresponding to point liquidity; the trading price will not change until the liquidity within that bin is consumed.

Due to TraderJoe's clear division of price ranges and the vertical integration of LP liquidity within a single bin, the distribution of liquidity incentives among LPs will be greatly simplified. A simple distribution method for liquidity incentives means that introducing bribery functions in TraderJoe may only require adding a third-party component.

Summary

The examples of DEX listed above provide a relatively macro perspective on the changes and potential developments of DEX. In fact, DEX also has some interesting small innovations, such as Crocswap's mechanism for automatically adjusting swap fees based on trading volume, Cowswap's MEV protection through Batch Auctions, the Coincidence of Wants protocol launched by Cow Protocol, and the MEV rabbit hole function jointly launched by 1inch and Metamask. These small innovations in DEX are tools that users can choose to use when executing swaps. Additionally, researching the profit models of some no-fee protocols, such as 1inch and WOOFi, is also an interesting direction. "We all thought that the development of DeFi had reached its peak, but in fact, it is continuously exploring new mechanisms and business models." Whether it is RFQ, the new Balancer yield model, or Uni V3-style bribery, only time will tell whether they are gimmicks or the future.