kylefang/docs
1
0
Fork 0
mirror of https://github.com/uniwhale-io/docs.git synced 2026-01-12 22:54:45 +08:00
Code Issues Packages Projects Releases Wiki Activity

GitBook: [#84] No subject

Browse Source
This commit is contained in:
Good Sanai
2022-12-12 05:32:50 +00:00
committed by gitbook-bot
parent 7665d1be89
commit 02c2ca5580
1 changed files with 37 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

40
unique-oracle-design.md
View File

@@ -14,11 +14,45 @@ Such harmful wicks can be caused by many factors, but mainly we are concerned th
Oracle-based trade execution mitigates such harmful candlestick wicks because the oracle prices are published based on aggregate data contributed by multiple data sources. This makes it almost impossible to create such illiquid/wrong/manipulated prices/pricing sources-driven wicks.
Uniwhale is currently using [Pyth Network](https://pyth.network) for its oracle prices. Pyth Network is an oracle that publishes financial market data to multiple blockchains. Market data is contributed by over 70 [first-party publishers](https://pyth.network/publishers/), including some of the biggest exchanges and market making firms in the world. Each price feed publishes a [robust aggregate](https://docs.pyth.network/how-pyth-works/price-aggregation) of publisher prices that updates multiple times per second.
## Pyth Network
As part of its product roadmap, uniwhale will be migrating to an custom-built oracle aggregator based on multiple service providers (including Pyth) to determine trading prices.
Uniwhale uses [Pyth Network](https://pyth.network) for its oracle prices. Pyth Network is an oracle that publishes financial market data to multiple blockchains. Market data is contributed by over 70 [first-party publishers](https://pyth.network/publishers/), including some of the biggest exchanges and market making firms in the world. Each price feed publishes a [robust aggregate](https://docs.pyth.network/how-pyth-works/price-aggregation) of publisher prices that updates multiple times per second.
By aggregating and filtering oracle prices across multiple service providers, our oracle aggregator can provide an oracle price that reflects event better the true market activities.
### Latency
Any protocols dependent on oracles must account for the difference in latency between on-chain oracles and off-chain sources (e.g. centralized exchanges). No on-chain oracle can match the latency of an off-chain source due to the added overhead for consensus and security. Therefore, protocols must assume that there will be players who see price changes slightly before the protocol does. [Pyth Network](https://docs.pyth.network/consume-data/best-practices#latency) describes this situation as follows:
`This situation is analogous to market making in traditional finance. Market makers place resting orders on exchanges with the hope of earning the bid/ask spread. When the “true price” moves, these market makers get picked off by adverse “smart flow” that is faster than they are. The smart flow is balanced by two-way flow, that is, people wanting to trade for other reasons besides a price change.`\
The last sentence of the above quote is also important, because this relates to a key assumption of an oracle-based execution venue - that the objective oracle-based execution can create a better two-way flow, whose benefits then far outweigh its shortcoming. We then [tune the bid/ask spread and offered liquidity](execution.md#risk-management) to limit adverse selection from smart traders while still interacting with two-way flow.
Further, we give ourselves a "last look" to decide which trades to accept, based on the following conditions:
* Price is not stale: last slot update is not too far behind the current slot;
* Price is not negative;
* Price is not too volatile: EMA/spot ratio is less than 5, and;
* Price is not too uncertain: confidence internal is less than 10%
If Price do not meet the above conditions, then the trade is rejected.
### Price Availability
While Pyth Network provides a good protection against returning a stale price, the main public endpoints for its price service, operated by the Pyth Data Association, are centralized and therefore create risks for the protocols using its service.
The price service listens to the Wormhole Network (which has built in redundancy and is therefore inherently more reliable) for Pyth price updates and is a critical component used by Uniwhale Exchange app to fetch on-demand price updates.
To mitigate this risk, Uniwhale Exchange is built on its own custom network of 8 price services, whose weighted average (using the slot-weighted/inverse-confidence weighted scheme like [EMA price aggregation](https://docs.pyth.network/how-pyth-works/ema-price-aggregation)) is then submitted for on-chain price update and trading. 
Our custom price service network therefore provide for maximum resilience and decentralization.
As part of its product roadmap, we are also building a custom oracle aggregator based on multiple service providers (including Pyth) to determine trading prices.
By aggregating and filtering oracle prices across multiple service providers, our oracle aggregator can provide an oracle price that is even more reliable.
### Oracle network health
You can monitor the health of Pyth Network at [Dune](https://dune.com/cctdaniel/pyth-oracle).