In a multi-chain world, interoperability protocols are a key piece of the infrastructure. The problem is that this segment is still very experimental and often unsafe. Indeed, bridge exploits are by far the largest category with around $2.5bn in stolen funds since 2020. Despite the challenges, the interoperability landscape continues to innovate at a rapid pace. Users and builders are struggling to chose the right bridge, i.e. one that is secure and convenient. There are interoperability trilemma’s to consider, think about trust assumptions, ...

Enter LI.FI. It’s a bridge aggregation protocol that enables cross-chain asset swaps by aggregating bridges and connecting them to DEXes & aggregators. In other words, it can route any asset on any chain to any desired asset on any chain.

They are primarily a B2B business by providing an SDK and widget that other projects can use. LI.FI is thus the middle layer between DeFi infrastructure and the application (Figure 2). For example, LI.FI is integrated in MetaMask for cross-chain swaps (80-90% of MM’s cross-chain swaps is routed via LI.FI). While LI.FI’s focus is on B2B, the team recently recently launched Jumper (formerly transferto.xyz), an exchange to swap assets across chain.

Dune Dashboard

I created a Dune dashboard that tracks the usage of LI.FI. Note that the dashboard does not give a 100% complete overview of LI.FI usage: 1) Dune doesn't cover all chains and 2) Dune doesn't have all pricing info of each wrapped asset (which in turn relies on Coinpaprika data). I semi-fixed the latter issue by manually relabeling some of the addresses of wrapped tokens to the underlying asset.

LI.FI currently covers about 20 chains and over a dozen bridges. Since its launch, over 400k transactions or about $400m in value has been transacted via LI.FI. Over 150k unique users have transacted with LI.FI.

Let’s take a look at some of the Dune charts. In recent months, several bridges including Multichain, Hyphen, cBridge and Hop (Figure 3). In terms of volume, most of it has been routed via Multichain and cBridge (Figure 4). Which applications are routing most via LI.FI? Figure 5 shows that the bulk of the daily unique users (+85%) are utilizing either Jumper (transgferto.xyz) or MetaMask. Most of the $ volume also comes from these two applications. Finally it’s interesting to look at the origin and destination chains of the swaps that LI.FI routes. This is shown in Figure 6. Most of the volume is routed between L2s (Arbitrum and Optimism) and Ethereum.

Concluding thoughts

It’s still early days for interoperability protocols, so it remains to be seen how it will evolve. But it’s an exciting area that I will keep a close eye on. Especially use cases that interoperability protocols can enable. Think about lending and borrowing in a chain agnostic way, where the loan and collateral would sit on different blockchains. Or decentralized exchanges that pool liquidity from different chains. Other examples that would rely on interoperability infrastructure are NFTs that would allow the user to access its utility on another chain. Or governance that allows users to vote cross-chain.

More and more protocols need interoperability, but teams often don’t have the time or expertise to navigate that labyrinth. LI.FI’s solution takes away that burden so others can focus on their own value proposition. I think LI.FI is well-positioned (see also this recent piece on how LI.FI fits in the aggregation theory). Finally, worth to note is that LI.FI has an excellent research team, and I always learn a lot from their writings.

The first part explains how liquidations work in Aave. The subsequent sections walks through some of the charts from the Dune dashboard. I cover on-chain data on liquidations, the liquidated users and the liquidators. The piece also looks into May 19th 2021, the day Aave saw it’s highest amount of liquidations. Note that the analysis is restricted to Aave V2 on Ethereum.

1. How do liquidations work in Aave?

Aave is a collateralised lending protocol. Specifically, Aave applies the concept of over-collateralisation, which means that users can borrow less than the value of the collateral they provide. This design choice is not surprising given the price volatility of crypto assets and Aave only considers the quality of the collateral when issuing loans. No information about the borrower such as income, past defaults, … is used.

The maximum amount that can be borrowed depends on the quality of the collateral. This is determined by factors such as smart contract risk, counter-party risk and market risk. For more information, Aave has some good documentation. For example, USDC has a loan to value of 86%, meaning that for each $100 worth of USDC that is deposited as collateral a user can borrow up to $86 worth of another asset. SNX is deemed more risky and has a loan to value of 49%. Only $49 worth of assets can be borrowed when pledging $100 worth of SNX. This table gives a full overview of the assets that can be pledged as collateral and how much can be borrowed against it.

The loan to value ratio is of course not constant due price fluctuations. The borrower has to ensure that the value of the collateral sufficiently covers the value of the loan at all times. If not, the loan can become under-collateralised. The liquidation threshold specifies at what loan to value ratio a loan is under-collateralised. This threshold also depends on the type of collateral. For example, USDC has a liquidation threshold of 88%, which means if the loan to value rises to 88%, it becomes available for liquidation. The liquidation threshold for SNX collateral is 65%.

It’s important to note that Aave looks at the aggregate health of each user (address). Users can have multiple loans and deposited different collateral assets and it’s this aggregate picture of loans and collateral that matters when it comes to liquidations. Aave calls this aggregate view the health factor, which is calculated as follows:

The higher the health factor, the better. When the health factor falls below 1, positions can be liquidated. There isn’t a centralised party that is responsible for liquidating loans. Instead, anyone can be a liquidator. This is done by making a liquidation call, pay back a part of the loan and receive the collateral valued at the repaid loan plus a liquidation bonus (or a liquidation penalty from the perspective of the borrower).

The histogram below shows the distribution of the health factor of all users of Aave V2

Let’s look at a simplified example that explains the above:

1. Stewart deposits 1 WETH when it has a price of $1,500 and borrows 1,000 USDC worth $1,000 (if he wanted he could borrow up to $1,245 worth of USDC). His loan to value ratio is 66%

2. Due to a sharp decline in prices, the deposited WETH is suddenly worth only $1,175. Since the loan of 1,000 USDC is still $1,000, the loan to value rise above the liquidation threshold of 85%. Stewart has only one loan outstanding (and only 1 WETH as collateral), so his health factor drops below 1. The position can now be liquidated

3. Liquidators (bots) are constantly monitoring the health factor of each Aave borrower. Andy’s liquidation bot notices that Stewart’s health factor has fallen below 1. Immediately, Andy’s bot makes a liquidation call and repays 50% of the loan, i.e. 500 USDC. In return he receives 500 USDC worth of WETH + a liquidation bonus of 5%. Note that with Aave V2 up to 50% of a loan can be liquidated (with V3 it is possible to liquidate 100% of a loan)

4. Stewart is left with a debt of 500 USDC and $650 worth of WETH as collateral ($1,175 - $500 - $25). His health factor has increased to 1.1. Andy has $525 worth of WETH. His profit is $25 minus the gas fees he paid. Note that I have ignored interest payments throughout the example for simplicity

With the basic concepts of liquidations in Aave out of the way, it’s time to dig in the data!

2. Liquidations

Since its launch on Ethereum Mainnet in December 2020, the Aave V2 protocol has seen over 27,000 liquidations, amounting to more than $1.1bn. The chart below shows the liquidations on a daily basis. Unsurprisingly, the amount of liquidations spiked during days of high price volatility. For example, liquidations reached +$170m on May 19th 2021, +$60m (Nov 22nd 2022) and +$50m (June 14th 2022). We will look closer into the events on May 19th at the end of this post. Interestingly, almost all liquidations from Nov 22nd can be traced back to one single user (more on that below)!

So which tokens have been liquidated the most (in $ value)? The piechart shows that the liquidations are very concentrated in just a few tokens. Around $610m or 55% of all liquidations was WETH. Moreover, the top three tokens (WETH, WBTC and LINK) account for almost 80% of all liquidations.

What about the borrowed assets? The next chart shows the total dollar amount that has been liquidated (y-axis) for each type of collateral (x-axis). Per collateral asset you can see which debt asset was repaid. Stablecoins were the most popular borrowed assets that were repaid. For example, from the $620m WETH collateral that has been liquidated, 80% of the repaid loans were USDC ($280m), USDT ($180m) and DAI ($100m).

3. Liquidated users

A total of 5,650 addresses have seen their Aave loan(s) liquidated. The piechart shows there is quite some variability of the total dollar amount that is liquidated per user. Seven users account for about 25% of all liquidations!

Let’s take a closer look at the users that have seen the largest liquidations. The table below ranks all users by total liquidated amount in USD. The address 0x57e04786e231af3343562c062e0d058f25dace9e has the ‘honor’ to lead the list, having been liquidated 385 times for a total of almost $64m. Who is behind the 0x57e account? It turns out it is Avi Eisenberg. Interestingly, all liquidations happened on the same day, namely November 22nd. He is thus responsible for the day that saw the second largest amount of liquidations! For more background, he borrowed CRV tokens to short, but ultimately failed.

4. Liquidators

Since the launch of Aave V2 300 unique addresses have acted as liquidators, by repaying a part of the loan and receiving some collateral as a bonus. The piechart and table below shows that 0xb7990f251451a89728eb2aa7b0a529f51d127478 has liquidated the most in terms of dollar value ($210m), and more than half of those liquidations took place on May 19th 2021. Looking at the other liquidators, it is worth noting that the top-5 liquidators account for almost 50% of all liquidations.

5. May 19th 2021

May 19th 2021: What a day - I was actually listening to a presentation about crypto at that time. I don’t think the presenter and the rest of the audience realized the crypto market was burning down. I remember searching for exchanges or websites that could still give accurate prices!

It was also a memorable day for Aave, seeing a record $172m in liquidations! A total of 670 loans were liquidated by 17 different liquidators.

The chart below shows that WETH was the collateral used most to liquidated loans (60m), followed by WBTC (45m) and LINK (20m). The most popular loans that were liquidated were stablecoins (USDT, USDC and DAI).

Who got liquidated? A total of 414 users saw their loan(s) liquidated. One user lost more than $25m - 0xe47d5cc33517d5b8433ff249efe094b989989acc is third on the leaderboard

6. Future analysis

This is just a very basic analysis, but there is more interesting data to look at. Ideas for future analysis:

• Liquidations in Aave V3 (just launched on Ethereum Mainnet!)

• Gas fees paid by liquidators

• Profitability of liquidators

• …

Thanks for reading! Link to Dune dashboard: https://dune.com/dcooper/aave-liquidations

The purpose of the competition is to write a Dune query that analyzes the quality of token pairs on Uniswap. The definition of ‘quality’ can be defined in many ways, but one possibility is to consider a pair of high quality if it has high usage (e.g. trading volume) and has a good stability. I decided to keep it simple and look at the size of the liquidity pools and trading volume to assess the usage of each pair. The stability of a pair is measured by the variability of the liquidity pool. Ideally you want to have a liquidity pool that is stable over time.

Rather than looking at all token pairs on Uniswap, I restrict the analysis to DAI token pairs (DAI-WETH, DAI-USDC, …). It turns out there are over 150 trading pairs on Uniswap V2 that have DAI.

The first metric to consider is the amount of liquidity that each pool has. Figure 1 shows the development for all DAI pools over the past six months. What stands out is the dominance of the DAI-USDC pair. Currently the DAI-USDC pool has over $30m in liquidity, accounting for almost 60% of all DAI pools on Uniswap V2. The second largest pool is DAI-WETH, with almost $15m in liquidity, followed by DAI-USDT ($7m). These three pools alone cover 90% of all DAI pools. Since November 2022, liquidity has been relatively stable across the largest pools. The total liquidity of all DAI pools is around $60m, which is down from $200m six months ago. The significant decline is largely due to the DAI-USDC pool that had a liquidity of $190m in August 2022.

A second metric is the trading volume of each DAI pair. Figure 2 shows the daily trading volume of these pairs in USD. The DAI-WETH, of accounting for well over 50% of the daily trading volume of all DAI pairs. Most of the pairs see very little volume on a daily basis.

Figure 3 puts these metrics together in a scatterplot. The X-axis shows the liquidity of each pool, expressed in USD. The Y-axis shows the variability of the liquidity pools. This is measured by the standard deviation of the liquidity pools over the past six months. The size of the bubbles represent the total trading volume seen over the past six months.

As one would expect from the prior charts, DAI-USDC and DAI-WETH stand out. Comparing these two pairs, DAI-WETH has seen a notably higher trading volume as can be seen from the size of the bubble. At the same time the pool of the token pair DAI-USDC is the most volatile pair (Figure 1 already showed the dramatic decline in the DAI-USDC pool).

It’s worth noting that only five pools have a balance of over $1m and only ten pools saw a trading volume over $1m over the past six months. That doesn’t look very positive for the other pairs...

Based on these simple metrics, one could argue that only DAI-USDC, DAI-WETH, and to some extent DAI-USDT, are the higher quality pairs. The rest has seen very little usage (low trading volume, low pool balances). It’s not a real surprise that the WETH and USDC pairs come out as the ones with higher quality, given their market cap and high usage in general. Arguably the current bear market doesn’t play in the favor of the other pairs either. Another caveat is that I only looked at Uniswap V2.

The Dune dashboard can be found here: https://dune.com/dcooper/dai-pools-uniswap