I underestimated impermanent loss

Summary

1. For some reason, I thought uniswap v3 and curve v2 addressed impermanent loss. That’s wrong.
2. Current AMM designs are inherently much more expensive for liquidity providers than order books. Most liquidity providers are losing money before rewards.
3. Protocols aiming to grow liquidity for their tokens should be thinking deeply about avoiding this impermanent loss issue. It is a significant cost above and beyond liquidity on centralised exchanges.

My prior thinking on Impermanent Loss

I knew about impermanent loss in liquidity pools from when I wrote a mini-course on DeFi that included exchanges. I realised then that impermanent loss is a cost to liquidity providers; a cost that can only be made up with fees and (usually) added token rewards from the protocol.

Then, Uniswap v3 came out, and I started to think they had solved the problem of impermanent loss because Uniswap v3 — rather than providing liquidity with an x*y = constant type approach — allows liquidity providers to concentrate their liquidity within a specific price range. Looking back, I don’t know if I actually read this anywhere, but I assumed that Uniswap v3 got rid of impermanent loss (and, to a lesser degree, I thought that Curve’s custom liquidity curves achieved the same effect):

Thanks for reading Hype or Hodl! Subscribe for free to receive new posts and support my work.

Hype or Hodl

Uniswap v3 is a Big Deal — and not just for crypto

Summary: Uniswap v3 is going live on May 5th. A summary of how non-defi exchanges work How Uniswap v1 was radically different Why Uniswap v3 should allow Uniswap to outperform all other decentralised exchanges Background Uniswap v3 is going live on May 5th…

Read more

a year ago · Pinotio.com

How I realised the mistake in understanding?

Only a few months ago, I came across Falkenblog’s post on Automated Market Makers and realised a big mistake in my thinking — Uniswap v3 (or Curve) doesn’t reduce impermanent loss! It only allows liquidity providers to take more concentrated positions with their capital!

Liquidity providers sell options to traders and the price of those options is related to the pool’s fees (and, I guess, to slippage, which relates to pool size). Turns out, the fees charged by most pools are insufficient to cover the cost of issuing those options. What’s more, it doesn’t matter if you provide liquidity across the entire curve or portions of the curve — one can run the same analysis by considering a whole curve as a sum of smaller liquidity ranges!

Most liquidity pools are loss making without rewards for liquidity providers. And, if fees are to be increased to compensate liquidity providers, then Automated Market Makers are significantly more expensive than centralised exchanges.

How this got worse

Many protocols tried (and are still trying) to solve the impermanent loss problem by compensating liquidity providers for impermanent loss. This was at the core of Bancor v3, which (at least partially) paid out Bancor tokens to compensate for impermanent loss. In late June, Bancor had to turn off this compensation due to price volatility in the market, and it caused a significant drop in BNT value. I’m surprised the token is still at $0.5 today. Here is a podcast from Unchained on the topic — skip to the second half (or you can also listen to the segment about Solend in the first half, also recommended listening).

The way I heard about this is because Nexus (an insurance platform) had protocol tokens stashed in Bancor v3 and there is currently a Nexus governance vote live to withdraw those tokens and realise a 40–50% haircut in making the withdrawal!!!

Where from here?

I see a few options:

1. We accept that higher fees on AMMs are the price for doing a decentralised trade
2. We move back to building order book type exchanges, like Loopring have done (an Ethereum Layer 2), and like dydx have done.
3. We find some other designs for AMMs. I’m not sure that’s possible, but I’m open to it being possible. These designs should just try to make the costs go away with protocol rewards.

Learning for protocols trying to grow liquidity

If a protocol is trying to grow liquidity, I think a key point to consider is how to overcome, and ideally avoid, this impermanent loss problem — otherwise tons of money is going to be burned compensating liquidity providers for higher fees via current AMM structures.

Thanks for reading Hype or Hodl! Subscribe for free to receive new posts and support my work.