Er I agree with you on that -- it's much better to do something like what FTX does (take an index and bound all data points at 30bps away from the median).
The "it all comes down to an API feed" really meant "some function of some set of API feeds"
Conversation
Taking a simple median across exchanges doesn't really provide an adequate market coverage either though because it doesn't take into account volume consolidation or new exchanges popping up
It's more nuanced than just you need to trust the APIs, it's about data quality
2
16
It just scares me when you say things like this (linked tweet) because if projects are really going to be using serum as their only source of price data...
1
11
Replying to
To be clear:
1) I think they should use a sanitized version of it
2) Serum is on-chain. That makes it *way* different from the cases you cite; a lot of the problems you might worry about go away (at least for other projects on Solana).
3
6
to be clear there are still a lot of nuances here but many of your points don't really apply to on-chain oracles that you can trade against composably in real time with small fees.
1
3
Having the data being generated on-chain doesn't necessarily mean it's trustworthy
There have been numerous incidents involving DEX-based price oracles that led to millions in user funds getting lost
Promoting their usage is just downright dangerous in my opinion
1
8
Replying to
1) I think this is very different on Ethereum vs Solana. When it takes minutes to trade on a DEX and lots of tx's fail, you can't reliably hedge; that's way easier on Serum.
2) A lot of this is solved using EWMAs. E.g. mid(last, bid, offer), bounded by 5% away from 5m EWMA.
1
4
This Tweet was deleted by the Tweet author. Learn more
As are oracles on Ethereum layer 2, the issue here isn't the speed of the underlying blockchain, it's the market coverage and tamper resistant of the oracle price feed
1
6
L2 only really works for other things on the same L2
Yup! Serum oracles also won't work trustlessly --> Ethereum, or vice versa.
1
4
Show replies

