Proposal: Run controlled experiments to assess effect of swap fee changes on medium/large pools

TL;DR

  • There is a lack of data to inform decisions about appropriate swap fee levels
  • We can utilize Bancor as a test-bed to run experiments to collect data
  • Run a 6 day experiment on lowering the ETH-BNT-MKR and ETH-BNT-LINK at 0.3% overall swap fee [one day on (LINK, MKR 0.15%), one day off (LINK, MKR 0.2%)] to see the effect on swap fee apy
  • Run a 6 day experiment lowering the swap fee on USDT stable coin (0.15%) for 6 days relative to other stable coins (USDC, DAI, 0.2%) to see the effect on swap fee apy
  • At the end of the experimental period (6 days) all swap fees will return to the status quo

Introduction

There has been much discussion on the potential for changing swap fees on large Bancor pools to drive increased liquidity, but quantitative, real-world data is lacking. Data that has been collected previously is not directly comparable to the current competitive DEX environment and liquidity pools.

In essence, the challenge in setting an appropriate fixed swap fee centers on the swap fee paradox: higher swap fees are attractive to LPs at the expense of traders, while low swap fees are attractive to traders but at the expense of liquidity and therefore slippage.

This proposal seeks to run two carefully controlled experiments to collect real world data on large/medium pools. The experimental design seeks to use controls to allow assessment of the effects of the changes. At the end of the experiment, pool fees will return to the status quo, and the data will be analyzed and used to support future proposals around changes to large pool swap fees.

Experiment 1

Change the swap fee on two important trade paths for ‘one day on, one day off’ repeated three times over six days. The two paths proposed are ETH-BNT-LINK and ETH-BNT-MKR, which currently sit at 0.15+0.2% = 0.35%). As the ETH/BNT pool is current at 0.15% while the LINK and MKR pools are at 0.2%, it is proposed to change LINK and MKR pools to 0.15% (for a total of 0.3%). In summary, the schedule would be:
Days 1, 3, 5: MKR and LINK, 0.15%
Days 2, 4, 6: MKR and LINK, 0.2%
Day 7. Return to MKR and LINK, 0.2%
This will allow the effect of changing the swap fee upon volume and apy to be examined in triplicate and basic statistics to be performed. By running the experiment over a short period we avoid the potential for major external events to impact the experiment, and with juicy LM rewards in play on these pools, the risk of capital flight is minimal.

A rationale for this experiment is provided here:
1d

The fee options for pools in Uni v3 are 0.05%, 0.3% and 1%. Industry standard is generally .3%.

As you all know, swapping on Bancor from TKN to TKN usually takes 2 swaps. So swapping from ETH to LINK currently costs .35%. With the proposed changes in mbr’s chart above, the cost will be .4%.

The ETH and LINK pools are currently Bancor’s 2 largest pools but trying out different size trades on 1inch, it’s pretty hard for Bancor to win these swaps right now.

A $10k swap from ETH to LINK goes through Bancor for 100% of the swap.
A $100k swap from ETH to LINK drops Bancor to 40%-80% of the swap
A $1M swap from ETH to LINK drops Bancor to 20% of the swap(!!!)
A $10M swap from ETH to LINK keeps Bancor at 20% of the swap
A $100M swap from ETH to LINK increases Bancor’s share to 40% of the swap
A $1B swap from ETH to LINK stays at 40% of the swap
A $3B swap from ETH to LINK increases and maxes out the Bancor share to 50% of the swap

This suggests that for ETH to LINK swaps, Bancor has industry leading liquidity to support large trades. However, for trades greater than $100k and less than $100 Million, the fee impacts the total cost of the swap on Bancor more than price impact of pools on other exchanges.

I believe that Bancor would be in a better position if we match the costs of our competitors instead of charging a premium that is probably costing Bancor quite a bit of volume.

My proposal is that we approach fee changes from the trader point of view and not the pool’s. I propose that we set our fee structure so that our larger token pools have their fees set to .15% and our smaller token pools to set their fee to .85%.

This would result in swaps like ETH <-> LINK to cost a total of .3%.

A swap from a large tkn to a smaller tkn like ETH <-> WOO will have a total cost of 1%.

Finally, small tkn to small tkn like WOO <-> MTA will have a total cost of 1.7%.

Experiment 2

Change the swap fee on the medium pool USDT to 0.15% for 6 days, and collect data on volumes and swap fee apy. At the end of the 6 days, the swap fee is returned to the status quo (0.2%).

This experiment will allow the effect of changing the swap fee on one stable coin to be compared to other similar stable coins (USDC and DAI) that will act as controls. By running the experiment over a short period with LM rewards activated we avoid the potential for major external events to impact the experiment and minimize the risk of capital flight.

Conclusion

Swap fees on large and medium pools is a common topic of discussion but is mostly based on opinion rather than fact. This proposal seeks to design some controlled, short term experiments to assess the effect of lowering swap fees on performance. This could form the basis of future experiments to help optimize the operation of Bancor.

Voting Instructions

  • Vote FOR to run two experiments: Experiment 1: Days 1, 3, 5: MKR and LINK, 0.15%; Days 2, 4, 6: MKR and LINK, 0.2%; Day 7. Return to MKR and LINK, 0.2%. Experiment 2: Change the swap fee on USDT to 0.15% for 6 days. Day 7, return to 0.2%.
  • Vote NO. Do not make any changes.
4 Likes

We all agree that we don’t have enough data to make informed decisions on how we should be updating the fee structure on Bancor pools.

I think that this proposal is a great start for trying to reach some of that information that we need. In my opinion, 6 days is too short. I think that extending both experiments to 14 days would be beneficial to also allow week over week comparisons, especially for Experiment 1.

Overall though, I fully support this proposal to experiment on fee changes to try to figure out if Bancor is pricing itself out of significant volume with the current premiums.

Interesting, Looks good I like this first of it’s kind proposal. Would be cool to see how we could quantify the results in a non-biased ways. this has been said before but we need to know how and what we are going to measure.

1 Like

More than anything, I’m interested to see how the USDT pool compares to the USDC and DAI pools during this time. Pool fee will be reduced but how much more volume will it bring? Does Bancor’s volumes increase overall as there is a new gateway token with a low fee? Does it increase volume enough to earn more fees than the bigger USDC pool?

mbr
Can you deprecate this topic? It has been replaced with a simpler proposal:

Proposal. Experiment to assess effect of 0.1% swap fee on USDT stable coin pool