[ARFC] Flashloans Update

Gauntlet Flashloan Analysis

Simple Summary

Gauntlet recommends keeping flashloans enabled across Aave v2. There is inherent tail-asset risk should flashloans be disabled on v2 caused by low flashloan liquidity sources outside of Aave, given the amount of current tail asset borrow.

  • In general for liquidations, users may bypass Aave’s fees by flashborrowing for free on other platforms and using that as collateral on Aave and then borrowing the debt to repay during liquidation.
  • However, with tail assets frozen on v2, this method may be less effective for liquidating tail asset debt. Flashloans on Aave v2 thus may be necessary for liquidating tail debt and ensuring arbitrage opportunities.

For v3, allow-listing flashloans could work since the liquidators are already concentrated and few. Should the liquidator behaviour remain the same, this likely does not add additional market risk. However, any migration to new addresses by flashloan liquidators or change in spec of flashloan contracts used by liquidators would introduce both opportunity costs to liquidators as well as inefficiencies in the market via the whitelisting process. We defer to the community on the choice to whitelist flashloan participants on v3 deployments.

Some key findings from the analysis[1]:

  • Flash loans on Aave v2 Ethereum account for about 7.8% of all liquidations and 2.9% of liquidation volume, these flashloans represent those that utilized the flashloan mechanism on Aave.

  • Top 10 flashloan liquidators on Ethereum Aave v2 liquidated 96% of all flashloan liquidations.

  • In 2023, there is an increase in flash loan liquidation volumes for some tail assets (BAL, REN, FEI) compared to previous years.

  • Flashloan liquidations on Aave v2 Polygon make up approximately 10.4% of liquidation frequency and 13.9% of liquidation volume. However, the top 10 flash loan liquidators represent over 93% of all flash loan liquidations volume, indicating higher concentration.

  • All v3 Aave protocols have a high density of flash loan liquidator profiles, with the top 10 liquidators contributing significantly to flash loan liquidations. All v3 protocols, show that the the top 10 flashloan liquidators contribute to >90% of all flashloan liquidation volume.

v2 Protocols

Ethereum Aave v2

As stated previously, disabling flashloans could impede liquidating tail asset borrow positions, which may add risk, given the amount of tail asset debt.

  • tail assets have been frozen on v2, preventing the flashloan fee get-around method described above during tail asset debt liquidations

An initial breakdown of flash loan liquidations with respect to all liquidations shows that flash loans that utilized the flashloan mechanism on Aave only make up about 7.8% of all liquidations and comprise of only 2.9% of all liquidation volume.

Liquidations Amount (USD) % Count % Volume
Flash loans 2228 36.8mm 7.8 2.9
All liquidations 28647 1.28bn 100 100

A further breakdown of liquidators by their liquidation frequency shows that the top 20 flash loan liquidators make up only 7% of all flash loan liquidators.

image

The number of flashloan liquidators per volume also show that just the top 10 liquidators contribute to 96% of all flash loan liquidation volume. This further corroborates the influence and concentration of flashloan liquidators.

image

A distribution of frequency of flash loan liquidations vs normal liquidations over the course of 4 year shows a down trend.

image

Gauging over the same time period for volume of flash loan liquidations vs normal liquidations, we see an a decrease in flash loan liquidation volumes post 2021.

image

However, analysis on a per asset basis reveals that some tail debt liquidations utilize flashloans more for 2023. FEI has an increase proportion of flashloan liquidations this year and, for assets such as BAL and REN, the only liquidations this year happen to be flashloans liquidations on Aave from Aave’s flashloan mechanism.

Asset Proportion of Flashloan Liquidation Volume in 2023 Proportion of Flashloan Liquidation Volume Historically Proportion Flashloan Liquidation Count in 2023 Proportion of Flashloan Liquidation Volume Historically
BAL 100.0 0.7 100.0 20.0
REN 100.0 0.0 100.0 4.2
FEI 9.8 0.2 41.7 11.5
BUSD 5.4 0.1 8.8 5.1
GUSD 4.7 6.4 16.7 12.5
ZRX 2.2 5.7 16.7 16.7
DPI 1.9 1.9 16.7 14.3
DAI 0.4 2.8 1.1 9.6
TUSD 0.2 1.6 5.6 13.5
USDC 0.1 2.6 0.4 6.1
WETH 0.0 16.0 1.6 10.4

Some users bypass Aave’s fees by flashborrowing for free on other platforms and using that as collateral on Aave. However, this method is no longer effective on Aave v2 due to the freezing of tail assets. Two recent transactions involved flashloans of ZRX and REN for liquidation purposes. Assets such as GUSD, REN, DAI, FEI, DPI, and BAL have all had flashloan liquidations this year as well. To add to this, some tail assets do not have substantial flashloan-able liquidity outside of Aave v2. As a result, disabling flashloans on v2 could impede otherwise healthy liquidations, which may add risk given the total amount of tail debt.

Last Liquidations by timestamp
last_liquidation debtName last_flashloan_liquidation
2023-11-15 04:30:11+00:00 ZRX 2023-11-15 04:30:11+00:00
2023-12-09 14:49:47+00:00 BUSD 2023-11-01 01:58:11+00:00
2023-12-11 02:08:23+00:00 TUSD 2023-10-12 22:48:35+00:00
2023-11-29 02:29:23+00:00 GUSD 2023-10-10 02:59:59+00:00
2023-10-07 20:10:35+00:00 REN 2023-10-07 20:10:35+00:00
2023-12-11 02:13:59+00:00 DAI 2023-09-11 20:02:11+00:00
2023-08-17 13:34:35+00:00 FEI 2023-08-04 01:18:59+00:00
2023-11-28 23:46:23+00:00 DPI 2023-07-26 12:36:11+00:00
2023-12-11 02:16:59+00:00 USDC 2023-03-22 12:00:59+00:00
2023-03-22 11:49:35+00:00 BAL 2023-03-22 11:49:35+00:00
2023-12-12 20:25:47+00:00 WETH 2023-03-20 02:11:11+00:00
2023-12-17 02:58:59+00:00 USDT 2022-12-06 13:26:23+00:00
2023-10-09 02:56:23+00:00 1INCH 2022-10-15 15:30:59+00:00
2023-12-14 00:47:47+00:00 WBTC 2022-08-28 01:04:38+00:00
2023-11-30 05:09:47+00:00 sUSD 2022-08-27 07:37:42+00:00
2023-12-09 13:17:47+00:00 LINK 2022-08-20 12:02:34+00:00
2023-09-11 08:44:47+00:00 FRAX 2022-08-20 07:02:03+00:00
2023-10-09 01:57:47+00:00 BAT 2022-08-07 05:42:48+00:00
2023-12-17 02:37:23+00:00 AMPL 2022-07-30 08:32:11+00:00
2023-09-27 00:18:11+00:00 MKR 2022-03-30 06:49:26+00:00
2023-11-14 16:18:11+00:00 SNX 2021-07-06 05:06:26+00:00
2023-11-02 08:34:47+00:00 xSUSHI 2021-06-22 20:48:24+00:00
2023-01-22 15:33:11+00:00 CRV 2021-06-19 10:08:47+00:00
2023-10-09 04:28:47+00:00 ENJ 2021-06-19 07:16:20+00:00
2023-03-03 04:19:11+00:00 YFI 2021-06-09 04:08:08+00:00
2023-02-19 15:14:35+00:00 UNI 2021-06-02 01:39:30+00:00
2023-04-23 15:52:11+00:00 MANA 2021-05-25 15:47:44+00:00
2023-02-03 06:51:35+00:00 UST Na
2023-02-13 22:41:11+00:00 renFIL Na
2022-08-21 00:57:16+00:00 KNC Na
2022-07-23 03:13:18+00:00 CVX Na
2023-01-27 12:06:23+00:00 ENS Na
2023-08-17 21:42:23+00:00 LUSD Na
2022-12-28 20:34:47+00:00 RAI Na
2023-12-07 03:09:47+00:00 USDP Na

Polygon Aave v2

A similar trend is observed across Polygon Aave v2, where flash loan liquidations (that utilized the flashloan mechanism on Aave) are not as significant with only 10% of liquidation frequency and ~14% of all liquidation volume being attributed to them.

index Liquidations Amount (USD) percentage count percentage volume
Flash loans 7183 71754525.7 10.4 13.9
All liquidations 69209 516770580.5 100.0 100.0

However, when observing flashloan liquidator concentration, the top 10 flash loan liquidating entities make up over 61% of all flash loan liquidations and over 93.5% of all flash loan liquidation volume. This suggests that the liquidators are more concentrated and have more influence on protocol liquidations than observed in Ethereum Aave v2.

image

Dissecting the flash loan liquidations compared to all liquidations over time shows increase in flash loan frequency but decline in flash loan liquidation volume in 2023 when compared to 2022.

image

image

v3 Protocols[2].

Most v3 Aave protocols share similar trends in terms of flash loan liquidator profiles, however the make-up of flash loan liquidations (utilizing the flashloan mechanism on Aave) with respect to all liquidations follow different trends (See table below). Should the liquidator behaviour remain the same, whitelisting flashloan-enabled accounts likely does not add additional market risk. However, any migration to new addresses by flashloan liquidators or change in spec of flashloan contracts used by liquidators would introduce both opportunity costs to liquidators as well as inefficiencies in the market via the whitelisting process. We defer to the community on the choice to whitelist flashloan participants on v3 deployments.

All v3 protocols have high density of flash loan liquidator profiles as follows:

  • On Ethereum, top 6 contribute to 100% of all flash loan liquidation volume
  • The top 10 flash loan liquidators contribute to 99% of all flash loan liquidation volume on Polygon, Arbitrum, Optimism and Avalanche
  • On Avalanche, the top 10 contribute to 98% of all flash loan liqudiation volume

The table below summarizes the flash loan liquidation metrics with respec to the overall liquidation metrics for each chain.

Chain Flash loan Liquidations All Liquidations Flash loan Amount (USD) All Liquidations Amount(USD) Flash Loan Percentage Count All Liquidations Percentage Count Flash Loan Percentage Volume All Volume
Ethereum 10 439 4166.5 16188271.4 2.3 100.0 0.0 100.0
Polygon 1459 11169 1739439.6 19850784.8 13.1 100.0 8.8 100.0
Arbitrum 2106 3564 3884916.2 9265787.8 59.1 100.0 41.9 100.0
Optimism 605 913 6312843.3 7724241.3 66.3 100.0 81.7 100.0
Avalanche 910 6419 3425733.8 35782500.2 14.2 100.0 9.6 100.0

To give an example of distribution of liquiditaros and their frequencies, below is an image of Polygon Aave v3 flash loan liquidator profile.

image

Here, we see high concentration of flashloan liquidators with one liquidator contributing to over 600 liquidations (~41% of all flash loan liquidations).


  1. flashloans in the context of this analysis and discussion mean flashloans that have utilized the flashloan mechanism exclusively on Aave. ↩︎

  2. We have only accounted for protocols with significant TVL (>$50mm) and hence have excluded Fantom and Base v3 instances of Aave. ↩︎

9 Likes