Proposal to reduce 3pool fee from 3 bps to 1 bps (based on simulation results)
Optimizing 3pool fees has become increasingly important as competition for stable-coin trades has increased. Previously, using arbitrage simulations, we found that lowering fees was unlikely to produce sufficient volume to make fee decreases profitable. However, these simulations only considered arbitrage trades against 3pool in isolation.
We developed a new simulation approach which simultaneously simulates arbitrage trades against 3pool and the major meta-pools that use it as a basepool (listed below). This approach allows us to estimate how changes in 3pool fees affect returns across the whole 3pool meta-pool ecosystem.
Simulating the last 2 months, we found that reducing 3pool fees to 1bps would increase overall returns across the system
Figure 1: Cumulative fees for 3pool + metapools
Figure 2: Accumulated fees over 2 months
As can be appreciated from the figures, the simulations suggest that lowering 3pool fees would improve overall returns by increasing meta-pool returns.
- Simulations suggest increased returns
- Lower fees are also likely to attract more “random”/non-strategic volume that is not captured by the simulations
- LPing in 3pool itself will be less profitable. However, only a small subset of 3pool deposits are not from metapools.
Pools used in sim:
3pool, usdn, wormholeUST, paxos, mim, frax, lusd, ust, alusd, busdv2, gusd, dola, ousd, fei
Excellent! Definitely something we should try.
Couple o Qs:
- When does it become less profitable to reduce fees?
- What about zero fee 3pool? Or 0.005% fee spread?
So by lowering the transaction hurdle (fee), more 3pool<->[XYZ]/3pool arbitrage trades are profitable, increasing trading volume to the point that more fees are earned than before. Probably safe to assume that if you’re looking into the question, this sort of arbitrage trading is already happening and will actually expand. (Bots trading a million through stables to roundtrip for $1,000 profit or something?)
On top of that, the CT crowd can hype lower fees.
Returns LPing in 3pool already so low that people are in there for other reasons (parking lot for stable arb trading [ha], dust, gas hurdle, whatever)… It’s really just a float (between Curve revenue and claims on it) that happens to be a pool, no? So the argument against seems very weak.
Sounds good, thank you. If I got anything wrong someone please correct me; thanks in advance.
Excellent work! Help me interpret the data a little bit. Am I correct to assume that if the same conditions that were used to generate the simulation data is to hold into the future, the reduction for 3bps to 1bps should increase fee revenue for the entire platform by roughly 30%?
Here are simulation results including lower fees (.1 to 1.1 bps in .2 bps intervals). So, it looks like going lower to even .3 to .5 bps would be feasible. My intuition is to try 1 bps first, but happy to be overridden. It looks like someone has already started a vote for 1 bps, though.
I completely agree. The advantage is indeed lowering the spread for very many pools (~95% of stablecoin volume is through meta-pools). The main risk is that the simulation is not right, but, as you’ve laid out, there are a lot of reasons the results makes sense intuitively.
That is correct, except it would be increased fee revenue for just stablecoins (not crypto pools, etc.). Another caveat is that the simulation simulates arbitrage trades only, so the magnitude of fee changes may be different once you factor in changes in random/non-strategic trading activity.