Conversation

11:05 PM · Oct 15, 2020Twitter Web App

AMMs aren't broken Professional MM's don't need them, but they are great for the rest of us and improve liquidity for the space as whole Good thing that blockchain data is public so we can see what's really going on twitter.com/SBF_Alameda/st

Show this thread

Replying to

2) NOT INVESTMENT ADVICE

1

2

3) twitter.com/Rewkang/status yup, DeFi without incentivization has very little volume right now

Quote Tweet

Low latency, high throughput orderbook based hybrid DEXs have actually existed for years now e.g. dydx, IDEX, etc. They consistently are lower than 10% market share twitter.com/SBF_Alameda/st

Show this thread

1

2

4) twitter.com/Rewkang/status COMP listed in June. Prior to June, monthly volumes for all DEXes combined were less than daily volumes for the larger CEXes. The volumes were *tiny*. The largest (Uniswap) traded about $300m in May -- $10m/day, which is a rounding error vs CEXes.

Quote Tweet

I would refute the the idea that AMMs had very little usage before yield farming summer. Uniswap had already started to surpass volumes of some well known CEXs and had been experiencing hockey stick growth pre-yield farming

Show this thread

1

5

5) twitter.com/Rewkang/status The point isn't that it's just yield tokens trading on Uniswap -- it's that it's *pools with yield*. ETH-USDT, the $1 pool, is heavily incentivized by airdrops of multiple farming tokens. The TVL is being purchased.

Quote Tweet

Out of the top 10 pairs on Uniswap by weekly volume, 4 contain "yield farming tokens". The volume from them make up 22% of total top 10 volume. I'm not sure about calling incentivization "not natural", because liquidity incentivization is here to stay twitter.com/SBF_Alameda/st

Show this thread

1

4

6) twitter.com/Rewkang/status Yup! But it's poor compared to most of crypto. And it means you can't trust the data. Remember FCoin/Coinbene/etc. in 2019 -- do you feel similarly about those? They had billions of volume per day!

Quote Tweet

It's true - much of the growth of AMMs came from incentivization & subsidies, but so did a lot of the growth of most startups and even mature companies e.g. Uber, Amazon, Spotify, etc. It's a central strategy for every startup to bootstrap growth. twitter.com/SBF_Alameda/st

Show this thread

1

2

7) twitter.com/Rewkang/status The flip side of 30bp maker is 30bp taker -- it's extremely cost inefficient for natural buyers/sellers! 6x the price of most other types of exchanges. And that's without the token or protocol making any revenue yet!

Quote Tweet

Yes AMMs "make bad trades" on a trade by trade basis But the fees shouldn't be understated as they are game changers. Imagine if the Charlotte Hornets got 2x points per bucket. Yea they might be awful shooters but the bonus is enough to turn a L to a W twitter.com/SBF_Alameda/st

Show this thread

1

2

8) twitter.com/Rewkang/status There are 2 equivalent phrasings of IL. a) IL is independent of fees, but you make the maker rebate. b) IL is 30bps less bad on each trade if there are 30bps maker fees

Quote Tweet

IL is independent of fees. IL only changes based on start and end price. 30 bps is huge for HFT, but 90%+ of crypto volume comes from apes, not HFT. For Apes, 30 bps on 200+ vol assets is completely fine. twitter.com/SBF_Alameda/st

Show this thread

1

2

9) Also the claim that 90% of crypto volume comes from people who don't care about fees is not close to correct. If FTX charged 30bp fees on all its trades, volume wouldn't go down by 10%; it'd go down by 90%. Same for binance/okex/huobi/bitmex; with FTX that's >50% of volume

2

9

10) twitter.com/Rewkang/status This is a tricky one. But I don't think the obvious conclusion is correct. What I think is going on here: a) there's tons of incentives in defi b) e.g. incentives to provide in uniswap c) to do that you need USDT + ETH

Quote Tweet

LPs are not losing without incentives. You can look at pool yields net of IL, EXCLUDING incentives to see yourself. Here's ETH-USDT returns: 30D - 13% APY 90D - 60% APY APY dropped recently because of massive pool growth (large dilution) and IL twitter.com/SBF_Alameda/st

Show this thread

1

2

11) d) say Bob starts with just USDT e) Bob needs to buy ETH, so they buy from Uniswap f) UNI/ETH pool gets free 30bps fees g) Bob loses 30bps h) Bob stakes and makes back more than 30bps So the incentives can indirectly still be causing this. And I'm pretty sure they are.

1

4

12) twitter.com/Rewkang/status I don't think you can honestly call crypto mean reverting. Sure, BTC went from 4k - 10k - 4k - 10k - 4k - 10k but before that it went from 1 - 10 - 100 - 1k - 10k This is true of every coin. You can't only count the retrospectively reverty times.

Quote Tweet

AMMs are not bad, imo they will coexist with orderbooks and serve a pretty large population. Yes, they lose $ in trending environments, but actually perform the best in high volatility, mean reverting markets - exactly what crypto has been twitter.com/SBF_Alameda/st

Show this thread

1

5

13) twitter.com/Rewkang/status Most trades aren't part of a strategy of the form "A <> B will mean revert starting now", where A and B happen to be a demanded trading pair. And probably the strategy is more specific than that anyway.

Quote Tweet

LPs don't need to blindly enter and exit. Some will be smart enough to identify a time horizon and entry/exit targets to minimize IL and maximize profits. Single formula does NOT mean single strategy

Show this thread

1

3

14) There are definitely good uses of AMMs! And they should be used as such: for new coins to get initial liquidity, and for stablecoin pools, and with more flexibility as on-chain programmable bots. But I still don't think they come close to replacing orderbooks.

2

8