Skip to content

Latest commit

 

History

History
74 lines (47 loc) · 4.14 KB

exchange_oracle.md

File metadata and controls

74 lines (47 loc) · 4.14 KB
Raw

oracles

oracle networks create a common good, readable on-chain data. however, its hard to imagine publishing this data for free as being self sufficient in the long run.

presumably, the best oracle is the mid-point of a deeply liquid spot market. deeply liquid spot is expensive to create, especially if its a token from another chain (think BNB on solana). also, if the financial incentive to manipulate this spot market is higher than the derivative that it relies upon, it can break down. additionally, a spot market also doesn't really work for any abstract non tokenized information, like a prediction market.

another route is to make things not NEED oracles and be oracle-less for derivative markets. a futures markets doesn't need an oracle price if its physically settled at expiry. but an oracle can help reduce the incentive for wicks to hunt out a user's liquidation/stop price. perps markets and/or cash settled need some reference price still.

derivative exchange based oracle

TLDR: maker orders stake to earn a larger fraction of taker fees if filled or lose the stake if cancelled. they can reclaim the amount lost in a stake by successfully filling later on

the exchange relies on the oracle, so why not take advantage of the work by market makers being done on an exchange to build an oracle network? to avoid it from being too circular, a few distinctions need to be made.

active market makers DO a lot of work on exchanges. volume over a lookback window

terms:

order = {
  side: BUY | SELL,
  size: float > 0,
  price: float > 0, 
  stake: float >=0, // amount willing to lose if they cancel
}

X = total SOL staked on sitting orders
OP = oracle pool, lost staked amount
LS = a maker's total loss in staking

REP = a maker's reputation (some function of 24h volume maker volume, percentage of orders filled)

oracle bid = weighted sum (weight = size*stake) of price divided by total of bids
oracle ask = weighted sum (weight = size*stake) of price divided by total of asks
=>
oracle price = (oracle ask + oracle bid) / 2
oracle confidence proxy = (oracle ask - oracle bid)

walkthrough:

  1. market maker places an order with above parameters (lets say: BUY 100 SOL-PERP @ $30, stake $1.1)
  2. if the maker cancels this order, they lose $1.1 and its added to the OP, LS +=1.1
  3. if other trader takes it and pays a taker_fee=$1, they'll earn a fraction (rebate = sqrt(stake)/(sqrt(X+1)) = sqrt(1.1)/sqrt(2.1) = .72) of the taker_fee and have their stake returned, also if this maker has lost by canceling staked orders previously, LS, they can earn min(taker_fee-rebate, LS/2) = min(1-.72, 0/2) back with this fill
  • if orders are partially filled and cancelled. the stake loss only applies to the fraction of the order cancelled.
  • if a maker's bad reputation (LS) falls below a certain level, perhaps they will not be allowed to earn back

observations:

  • this system encourages makers who are confident in their orders being filled to stake and earn a maker rebate.
  • even if a maker wash trades themselves to collect their stake, the exchange took portion of their funds to discourage it
  • if a maker wants to manipulate and spend a lot of stake to be made back by wash trading later, they will pay the equivalent amount they earn back (difference between the taker_fee and rebate (exchange revenue))
  • a sufficiently large X is still encouraged with the sqrt in the reward function
  • the oracle pool can be used to both punish and return the amount staked in orders that were cancelled back to makers for successfully filling later
  • the oracle pool reward can never give a user a profit or even ALL of the losses back (can only approach zero)
  • the maker is incentivized to "earn it back" if they cancel and lose, so as to not be overly cautious with staked orders if prices move naturally
  • if no orders are staked, oracle can be stale at last staked order price

information leak:

  • if orders staked are only/more on one side (bids), makers could be less confident that asks will get filled. this leads to prices expected to go lower in the future
  • normally makers quoting like this leads to negative funding rates (to encourage more buying from takers)