Skip to content

Experimental variations for a Median Price Oracle with Uni v4

License

Notifications You must be signed in to change notification settings

saucepoint/median-oracles

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

35 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

median-oracles

An experimental suite of Median Price Oracles

from ETHGlobal Paris 2023, built with Uniswap v4 Hooks 🦄

Why not use TWAP? lower-liquidity pools are prone to manipulation!


Algorithm Gas (Read)
Quickselect comically a lot (hundreds of thousands)
Frugal-2U comically a lot (but less than QS)
Running Frugal-2U 4812
Quickselect Time-weighted TBD
Frugal-2U Time-weighted TBD

Methodology: obtain 50 unique tick observations by running swaps in both directions. Each swap is spaced 12 seconds apart. Use gasleft() before and after reading the median

Median Price Oracle (Quickselect)

The classic: given a sequence of unordered ticks, fetch the median with the quickselect algorithm

  • Uses an O(logn) algorithm on-read
  • Depends on tick observations written to storage

Frugal Median Price Oracle

Approximates the median using Frugal-2U from a sequence of numbers (naive implementation)

Frugal median algorithm compares new numbers against the current approximation and updates the approximation according to a dynamic step

  • Uses the frugal median-approximation algorithm
  • Depends on tick observations written to storage

Running Frugal Median Price Oracle

The gas-optimized implementation of the frugal median approximation: calculating an on-going approximation of the median

  • Uses the frugal median-approxiation algorithm
  • Stores the running median (approximated) in direct storage
  • additional research™️ required for windowed support

Future work: step-optimization

In the frugal median algorithm, the dynamic step can be modified to favor stabilization or responsiveness/accuracy. The repo uses 1-tick as a step, but the implementation is set up for additional experimentation

Future work: time-weighted medians

The repo is in its early stages and did not have sufficient time to implement time-weighted medians. Time-weighted medians are likely to better represent the price since they account for the duration of a tick (price observation). The repo currently treats unique price observations as having equivalent durations.


src/
├── RunningFrugalMedianHook.sol   Running median approximation
├── TickObserver.sol              store tick observations for windowed median reads
├── lens
│   ├── FrugalMedianLens.sol      read TickObserver and approximate median
│   └── MedianLens.sol            read TickObserver and calculate true median
└── lib
    ├── FrugalMedianLibrary.sol   median approximation library
    ├── MedianLibrary.sol         median calculation library (quickselect)
    └── RingBufferLibrary.sol     optimized ring buffer for price observations

test/
├── FrugalMedianLens.t.sol        test the naive frugal median
├── MedianLens.t.sol              test the true median (quickselect)
├── RunningMedian.t.sol           test the running median approximation
├── TickObserver.t.sol            test the tick observer
├── implementation
│   ├── ...                       Uniswap overrides for testing
├── median
│   ├── FrugalMedianTest.t.sol    test frugal median algo
│   └── MedianLibrary.t.sol       test quickselect algo
└── utils
    ├── HookTest.sol
    └── RingBuffer.t.sol          test ring buffer

Additional resources:

v4-periphery contains advanced hook implementations that serve as a great reference

v4-core


requires foundry

forge install
forge test

Credits

About

Experimental variations for a Median Price Oracle with Uni v4

Resources

License

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published