[RFC] Proposal to active 2, 3, 4 bps fee-tiers on Base

[aadams]

TL;DR

• This proposal analyzes decreasing Uniswap v3 market share on Base to Aerodrome, and conjectures that a core input into that change in market share is Aerodrome’s 4bp fee tier ETH/USDC pool. The pool undercuts Uniswap v3’s 5bp fee tier pool.
• We propose the creation of 2, 3, and 4 bps fee tiers on Base in order to win back DEX market share from Aerodrome.

Primer

The Uniswap Protocol has long since been the dominant AMM trading venue in EVM markets. The crown jewel of this trading is the ETH/USDC pool on Ethereum mainnet. The pool between ETH/USD trading pairs is the most important pool within a chain’s ecosystem because it represents the transition between flight-to-safety and risk-taking.

Most other pools within the AMM ecosystem utilize ETH as their pair asset. Pools generally do this, because ETH is correlated with most assets trading on blockchain-based markets. However, ETH is by definition not correlated with USDC, thus making it the most difficult pair to provide liquidity for in AMM markets.

On July 27th, Aerodrome ETH/USDC overtook both the Uniswap ETH/USDC both on Base and Ethereum - marking the first time for this newcomer DEX.

https://x.com/jessepollak/status/1817225900737654863

One could say that Aerodrome has now entered the decentralized exchange thunderdome. Many have asked why has Aerodrome seen such a surge in volumes and what new technologies has it used to get here?

General DEX Marketshare

Source: https://dune.com/queries/3946918/6639606

Across EVM chains, the Uniswap Protocol is by far the dominant trading protocol. While the volume share of the protocol shifts day to day and in different market conditions, the share generally sits around 50%.

On Base specifically, the share of the Uniswap Protocol by volume is around the same, but has generally been trending downward from highs above 80% to most recently below 40% since April 28th.

Source: https://dune.com/sealaunch/dex-metrics-on-base

As we can see from the chart above, there is a stark difference between the Uniswap Protocol + Aerodrome vs. the other DEXs. We will focus only on those two protocols for the rest of the piece.

Share by Volume

Source: https://dune.com/queries/3946723/6639218

The volume of Uniswap Protocol on Base has not seen a huge decrease. The majority of the change in volume share is actually from Aerodrome’s increase in volume.

The increase in volume from Aerodrome can mainly be attributed to their introduction of their “Slipstream Protocol”, which is a self-described Uniswap v3 fork, which made “only strictly necessary changes to the existing implementation due to the complexity and risk of introducing any regressions or potential security issues.”

Aerodrome Volume by Pool

Source: https://dune.com/queries/3946737/6639241

The USD volume from Aerodrome is concentrated around one singular pool, which is their ETH/USDC 4 bps pool, which is shown in black. Over 70% of the entire volume share of the protocol comes from this pool. The next highest pool has 6.4%, which is the 4 bps ETH/USD+ pool.

Uniswap Protocol Volume by Pool

Source: https://dune.com/queries/3946743/6639254

This is in great contrast to the USD volume share of the Uniswap Protocol on Base, which has the largest pool at 28.4%. This is the ETH/USDC 5 bps pool, which is also shown in black. For reference, the large orange share (labeled 52.4%) above is all pools outside of the top 10 combined.

Observant readers will notice the difference between these two pools right away - the Aerodrome ETH/USDC pool has a 4 bps fee while the Uniswap v3 pool has a 5 bps fee. Because most retail users have nearly 0 price impact on their trades, liquidity differences between the two pools will not bring back flow. This is because the fee from the trade dominates the cost of trading on that pool. Liquidity mining alone will not bring back the flow. For the rest of this piece, we will focus on the ETH/USDC pools on both of the protocols.

Most retail users’ trades are semi-sophisticated because they use interfaces to find the best routes for their trades. Because of this, the difference in 1 bp will cause sophisticated interfaces to choose Aerodrome over Uniswap.

To test this, we labeled the top 400 interface contracts by volume as either retail or non-retail (ie MEV). We checked where interfaces were routing their volume to and how much volume is from arbitrage volume (volume routed from MEV bots). These lists can be seen in the queries below.

Our sample is over the last 14 days as of the time of writing. On Uniswap Protocol, $293m of retail flow went into Uniswap v3. On Aerodrome, $425m of retail flow went into their Protocol. This is about a 45% difference in retail flow. This is significant, but the difference is not without a cost.

The higher fee tier of the pool protects LPs more from arbitrage losses. For reference, the arbitrage volumes on Aerodrome are $2.18b while the Uniswap Protocol was $588m, almost a 4 times increase. While there is a retail volume difference, the volume generated from arbitrage is massively different as well.

Notice that while Aerodrome has higher retail volumes, the arbitrage volumes are also significantly higher. More importantly, the volume of non-retail trades is proportionately much more than on Uniswap Protocol. This means much more of the flow that the LPs of Aerodrome are taking is worse. Because fee-tiers are inversely proportional to arbitrage losses, while the difference between 4 bps and 5 bps may seem small, the inverse relationship combined means the impact of these fee differences is super-linear.

ETH/USDC volume categorized by originator on Aerodrome

Source: https://dune.com/queries/3946817/6639387

Note: The data is Jul 13th - Jul 27th and the selected ETH/USDC pool

ETH/USDC volume categorized by originator on Uniswap v3

Source: https://dune.com/queries/3946818/6639388
Note: The data is Jul 13th - Jul 27th and the selected ETH/USDC pool

Not all flow is equal, and more volumes may not necessarily be better for LPs. However, there is also evidence that Layer 2 Protocols significantly lower losses and increase LP capital efficiency - this is due to lower blocktimes and lower fees. Because of this, the market clearing fee for LPs on L2s may actually be lower than 5 bps. Because Uniswap v3 has fee-tiers chosen by governance + the initial deployers, it is likely that these fee tiers may not totally be accurate.

Proposal

In the analysis above, we suppose that the reason why Uniswap has lost market share to Aerodrome over the last 4 months is Aerodrome’s lower fee tier on its ETH-USDC pool. This pool has attracted more retail flow as Uniswap’s flows have remained similar. The increase in retail flow and lower fees have in turn attracted significant non-retail flow.

Since Aerodrome is a Uniswap v3 fork, we see this analysis as good evidence of the viability of a lower fee pool for ETH-USDC on Base and suggest it as evidence to governance that this proposal has merit.

Cons:

One benefit of Uniswap v3 is the incentive to collect liquidity into one pool, creating network effects. We posted a tweet about it here

.

Source: https://x.com/whetstonedotcc/status/1803821112020676694

However, we believe that liquidity providers will respond quickly to updates of the fee. This will overcome the effects from liquidity fragmentation.

Proposal

Just like the 1 bp fee tier originally proposed 3 years ago by Getty Hill, we propose to additionally add a 2, 3, 4 bps fee tier to Uniswap v3 on Base.

You may question why we are proposing the creation of three new fee tiers when our analysis only focused on 4 bps. There are two reasons for this. For one, Aerodrome is able to adjust fees automatically on Base. We anticipate they may respond by lowering fees below 4 bps and want Uniswap Protocol to allow LPs to move without governance intervention if this happens.

Second, we believe this analysis may be a single case study into a larger finding that the required fee for LPs on L2s may be lower than 5 bps. While Aerodrome highlights the success of a 4 bp fee tier, it’s unclear whether that’s the right market rate. It may be lower.

Lastly, we note that this change should not be blindly repeated on other chains. The community should do further research on the success or failures of this program before executing it on another chain.

[rfritsch]

Thanks for the great, detailed analysis!

You nicely demonstrate the motivation with the example of the 4bp Aerodrome pool. More generally, I understand the motivation as:

Add more fee tiers to give LPs more fine-grained choices to compete with other AMM pools. (Please correct me if I’m wrong.)

I’d like to add two points on this:

1. If we believe adding more fine-grained fee-tiers is good, should this be done across the full range of fee tiers?
   The current fee tiers of 1, 5, 30, 100 are relatively evenly spaced out with ~4-6x increases between them. Just as you’re suggesting to add new tiers between 1 and 5 bps, maybe the same should be done, for instance, between 5 and 30 bps with new 10 and 20 bp tiers.
   A possible set of fee tiers in the current fee tier range could be: 1, 2, 3, 4, 5, 10, 20, 30, 50, 100 bps. What are your thoughts on this?

2. Regarding the possible liquidity fragmentation, I wonder how much empirical evidence there is on this statement:

However, we believe that liquidity providers will respond quickly to updates of the fee. This will overcome the effects from liquidity fragmentation.

For instance, I believe liquidity has often consistently been split between 5bp and 30bp pools, even though they are a lot farther apart than 2,3,4 bp pools.

[kfx]

This is a clearly written and motivated proposal. Still, it presents some fairly significant risks. I would support it if it was clearly marked as an experiment, especially considering that v3 is expected to be obsoleted by v4 in the mid-term future, which reduces the impact of any potential negative outcomes.

A successful outcome of this experiment, as I understand it, would mean that all of the following expectations are met:

1. Uniswap regains ETH/USDC market share on Base.
2. ETH/USDC LPs on Uniswap@Base remain profitable according to standard metrics such as realized PnL, markouts, and/or LVR.
3. Most ETH/USDC LPs converge to a single fee tier.

Conversely, bad outcomes would include any of the following:

1. The market share remains reduced.
2. The change starts a race to the bottom, where liquidity is mispriced due to choosing a fee tier that is too low.
3. ETH/USDC liquidity becomes spread out and fragmented. (This is problematic not so much for Uniswap itself but for the many DeFi protocols that integrate with Uniswap, including those that use Uniswap pools as price oracles.)

As rfritsch mentioned, split liquidity is a common situation, so I think the third positive outcome is particularly unlikely if all of the 2, 3, and 4 bps fee tiers are enabled.

Additionally, even if the outcomes on Base are negative, there is a risk of blindly copying and deploying the proposal on other chains, where the environment is less favorable for AMM LPs than Base, or LPs are less experienced and more likely to misallocate liquidity.

Overall, my point is that if this is deployed as an experiment (as it should be), the DAO should be prepared to monitor the outcomes and acknowledge if the experiment fails.

[TimeRows]

I am in favor of this proposal, as it would be interesting to see the resulting data.

Risks:

I also have concerns about point #2 that @kfx raised: “2. ETH/USDC LPs on Uniswap@Base remain profitable according to metrics such as realized PnL, etc.” One of the main narratives I’ve encountered around Uniswap is the impermanent loss that LPs face, which may not always be fully realized.

This analysis focuses mainly on the 4% fee tier as the reason to introduce more tiers. Are other factors at play?

Are LPs profitable on Aerodrome, and is the Total Value Locked (TVL) attracted mainly through token distributions and voting via veAero?

Is there an incentive to engage in wash trading for token distribution or voting power?

I’m not familiar with how their model works and how it might skew the presented data. Is it a hybrid of Curve and Uniswap?

Final Thought:

It would be unfortunate to see Uniswap LPs presented with lowering fees and potentially incurring losses to compete with metrics influenced by token dilution. However, lowering the fee tiers may address many of the questions raised above. I believe the data will be useful, and LPs will eventually find an equilibrium around profitability in the long term.

[BlockworksResearch]

Based on our own findings about Aerodrome’s recent success, we also would like to support this proposal and recommend an experimental period with a new fee rate for the ETH/USDC pool. In case the results from this experiment are unsavory, would there be some additional attempts we can take to try to regain volume in ETH/USDC? If liquidity mining and fee rate changes do not work, what’s next? There also might be some incentives at play with the AERO token that may be worth exploring. When volume/pool fees increase, the incentive to unlock and sell the AERO rewards decreases, and vice versa. Could this also have an effect in some way?

[alicecorsini]

This is a highly relevant topic, thank you for bringing it up.

I think there are two aspects we should consider: volume and fee generation. Base is Uniswap’s most profitable chain (excluding Ethereum), despite not being the one with the highest volumes. We need to carefully consider whether our primary goal is to maximise profitability through higher fees or to expand our market share by attracting more volume, potentially sacrificing profitability.

In terms of volumes, Aerodrome is continuously expanding its market share at the expense of Uniswap. However, what is interesting is that Uniswap has an order-of-magnitude larger amount of daily active users compared to Aerodrome, despite the latter having larger volumes. To me, this means Uniswap attracts a broad and active user base engaged in smaller-scale trading where Aerodrome has a concentration of high-value trades by fewer users.

The question is: would adding multiple fee tiers do the trick and attract some of these high-value trades? Or, would it just reduce profitability on the existing smaller trades?

[aadams]

Happy to mark this as an experiment, which I believe is a good framing of this approach. It should not be repeated blindly on more chains, unless there is further research and discussion on the output of the change.

[aadams]

There could be further attempts or ideas to bring liquidity over if the current approaches fail, but I believe that they should not be expected and would require more discussion. I personally do not believe that token emissions are a sustainable market structure, so would prefer to go with no token incentives and then utilize the generated data to evaluate if token incentives are worth the value.

[aadams]

This is an important point to make. I will say that there is a distinct market structure difference between short-tail ETH pairs and long-tail tokens.

For the ETH/USDC pair specifically, over the last 14 days, the Uniswap v3 pool does have significantly more unique traders - around 3x more (48k for v3 and 16k for Aerodrome). However, the median trade size is around 4x larger for Aerodrome ($4.7k vs 1.1k). Source: https://dune.com/queries/3969057

Non-toxic large trades are very impactful, because they pay marginally more price impact than smaller trades. This ends up with LPs profiting marginally more money from these trade. They are also likely the most price sensitive/ use aggregatorsfrom previous analysis. All together it is notable that larger trades are seeing a switch over to Aerodrome, likely because of fee differences

[SEEDGov]

Hi. We would like to ask a question, do you have any metrics, data or projections on the potential impact of liquidity overfragmentation that would result from the introduction of new pools/fee tiers for the ETH/USDC pair? Thanks.

[kfx]

Can we go more into the details here, and agree if not on the expected results, then at least on the metrics to use? Also who will be doing the monitoring of the results, do Labs take on this task, or you see that more as a community effort?

[Gauntlet]

We have voted in favor of the Snapshot proposal to activate 2, 3, and 4 bps fee-tiers on Base. That said, this is an area for greater research and discussion around the potential risks associated with this experiment, especially since the ETH/USDC Pool on Base represents one of the most strategic pairs on one of the most strategic L2s.

Considerations we think are worth exploring from the existing conversation:

• The fractionalization of WETH/USDC liquidity across multiple fee tiers, potentially leading to even more outstanding market share for the Aerodrome 4bps pool.
• Exploring deeper the characteristics of Aero pool trading volume (fewer trades, higher volume).
• An analysis of Aerodrome’s ability to react aggressively to a new fee tier and the second-order effects of a fee race.
• What is the optimal rate between 2, 3, and 4 if the DAO were to make a single decision?

We look forward to greater community research and discussion around the topic prior to an onchain vote. It’s an interesting idea and certainly worth exploring more.

[Doo_StableLab]

Thanks for the proposal. However, for now, we will vote against it as proposal for research should come before activation.

[aadams]

Happy to chat about some of these valid concerns.

First, we believe that fractionalization of liquidity on L2s is inherently less impactful due to lower gas costs on the chain. Splitting a trade between two pools on an L2 has less fixed costs from reading contract into state, which means that a price sensitive transaction is more likely to split. We have seen this empirically with routers more likely to split trades on L2s. This likely means that there will not be a large market impact from liquidity fragmentation.

We believe that the below graphic is the best explanation of the current situation. Price sensitive retail flow is defined as flow that originated from an interface that is not associated with a DEX project. Some examples are Matcha, 1inch, Odos, etc.

We choose this metric as these users are purely looking for the best price for execution on the chain. As we can see, over the last 28 days, the majority of volume from these interfaces was directed to Aerodrome. This flow is key to capture.

Source: https://dune.com/queries/3991567/6717557

This tracks with our previous analysis on the size of trades on the two pools, which shows that larger (generally more price sensitive) but fewer traders are trading on Aerodrome.

Source: https://dune.com/queries/3969057

Retail flow and what impacts it is one of the current largest unsolved problems in AMM/finance literature. We cannot provide definitive impacts nor optimal rates, but see Aerodrome (which is a v3 fork with minimal changes) as a good natural experiment. Aerodrome can likely react quickly to any changes from Uniswap v3 LPs, which is why we suggested not liquidity mining at this time. We just believe that LPs should be given more choice.

[404DAO]

Hi @aadams, thank you for the proposal. We would like to voice concern around the theorized motivation for why aerodrome has overtaken Uni on Base. While the fee tier could be the reason, it very well could be the token incentives driving deeper liquidity which subsequently would lead to Aerodrome being the preferred outlet for larger traders.

If the current assumptions presented around the fee tier do not end up being the real cause for the decline in market share, then we risk collateral damage by hurting lp margins who then may take their capital to aerodrome where they are already offering a competitive, albeit subsidized, rate.

Undercutting fees because we are losing market share is a knee-jerk reaction and does not take into account the full situation.

To be clear, we are not categorically against this proposal but there needs to be more research before any decisions are made.

[aadams]

We generally believe that the fee-tier is one of the main reasons. There is certainly an impact from liquidity differences, but it should generally be negligible for the vast majority of retail traders.

Because retail flow generally has such small price impact, liquidity differences do not matter more than the difference in the fee. You can see this in Table 1 from our previous written papers on the costs of swapping on AMMs. While the data is from Ethereum Mainnet, it should be a decent data point.

According to the paper, the average medium-sized ($1k - 100k) ETH/USDC retail trade has a price impact of .7 bps. Since the fee difference is greater than this amount, no amount of liquidity differences for the median trade will bring back flow.

To bring back flow without changing the fee, the difference in price impact would have to be larger than the differences in fee. The fee costs is fixed, while the price impact is based on the size of the trade, meaning that only larger trades would be eligible for routing back to the Uniswap Protocol with increased liquidity.

[Userisky]

Thorough research! It makes sense to lower the LP fees considering the volume being aggregated through alternative frontends:

• $295,000,000 to Aerodrome

• $65,000,000 to Uniswap

Additionally, Aerodrome has high spending incentives via token emissions. Each Aerodrome epoch, which occurs every 7 days, disburses around 12,300,000 Aero. At the current Aero price of $0.64, this amounts to approximately $7,872,000 per week.

Scroll down to the Aero supply distribution section, this analyics page also includes data on bribes vs. fees, including Slipstream:

https://dune.com/0xkhmer/aerodrome

Potential Risk: Aerodrome has a highly aligned token with its users. Its evangelized users, who are also token holders, are likely to be sticky—remaining loyal to the protocol regardless of a competitor’s decreased LP fees.

[kfx]

Just to confirm the price impact with quick back-on-the-envelop math for Base, where liquidity in Uniswap’s pools remains lower than on the mainnet.

The USDC/WETH pool on Base has approximately $200M of virtual USD liquidity active at the moment, as computed from basescan’s liquidity and sqrtPriceX96 fields on basescan.

Using the approximation I ~= swap_size / (L * sqrt(P)) where L is the liquidity (L = value_usd / 2 / sqrt(P)) and P is the price, we get:

• price impact is 0.98 bps for a $10000 swap
• price impact is 9.8 bps for a $100000 swap

In short, each additional $10k of swap size creates additional ~1 bps of price impact. If currently a user pays x bps for a swap of size s (in total, for impact + swap fee) , then after reducing the pool’s fee tier by 1 bps the user would have the same total cost for a s + $10k swap. (This is from a single sample at the current state of the pool, so take with a grain of salt!)

[SEEDGov]

Having analyzed the proposal and despite some concerns we have, we understand that it is worth trying to add new fee tiers in the ETH/USDC pair in Base, in order to make Uniswap more competitive, hopefully capturing part of Aerodrome’s market cap and adapting to new market conditions that may occur in the future, waiting for an expected reaction from that DEX to the move that Uniswap makes.

However, we do not want to overlook the potential disadvantages that the new implementation may bring:

1. Increased access to arbitrage: by reducing fees, previously unprofitable extractions of value will become profitable
2. Fragmentation of liquidity; starting new pools will keep liquidity more fragmented, which will increase LVR and IL, negative for liquidity providers; as well as potential cost overruns in more complex routing for the trader.

If the proposal is approved, we expect the proponent to conduct a subsequent analysis and present conclusions on the impact of the changes introduced, to corroborate that the proposed hypothesis has been validated. We encourage the proponent will consider a commitment to this sense for the onchain vote.

[blockchainedu]

We are in support of running more experiments to find the optimal fee tier.

It’s good to acknowledge that the optimal tier may be different due to the L2’s lower blocktimes and gas fees.

But the higher volumes on Aerodrome is most likely coming from the token emissions being given to LP’ers.

“Acknowledge if the experiment fails”

Not clear, is this a genie out the bottle situation where the fee tiers can’t be turned off if the experiment fails?

Also think @aadams should take charge on monitoring the results with the support of the community in providing research, graphs, and feedback.