[test] Extensive simulation scenarios covering extreme edge cases

[Hyung-bharvest]

• testing decimal truncations
• message combination attack vector with swap/deposit/withdraw
• attacks with significant price distortion
• messages with very small amounts
• various limit order combinations in a batch with very edge situations
• very large number of random simulation with enoughly random messages

for milestone 2

[Hyung-bharvest]

paper review : https://web.stanford.edu/~guillean/papers/uniswap_analysis.pdf

[Hyung-bharvest]

Pool Match Only Simulation

• to verify swap result when every order is fully executed by liquidity pool
• one order per one batch
• relatively very small order amount compared to liquidity pool reserve : 0.1% of reserve
• large max slippage : 1%

Theoretical result to verify

• case1) swap X for Y

  • SwapPrice = (Reserve_X + 2 * OrderAmount_X) / Reserve_Y
  • SwapFeeY = (OrderAmount_X / SwapPrice) * SwapFeeRate
  • Receive_Y = (OrderAmount_X / SwapPrice) * (1-SwapFeeRate)
  • Reserve_X_AfterSwap = Reserve_X + OrderAmount_X
  • Reserve_Y_AfterSwap = Reserve_Y - Receive_Y

• case2) swap Y for X

  • SwapPrice = Reserve_X / (Reserve_Y + 2 * OrderAmount_Y)
  • SwapFeeX = (OrderAmount_Y SwapPrice) SwapFeeRate
  • Receive_X = (OrderAmount_Y SwapPrice) (1-SwapFeeRate)
  • Reserve_Y_AfterSwap = Reserve_Y + OrderAmount_Y
  • Reserve_X_AfterSwap = Reserve_X - Receive_X

[Hyung-bharvest]

Other edge case tests

deposit

• deposit non-reserve tokens
• amount <= 0
• amount is very small
• deposit only one kind of token
• deposit more than 3 kinds of tokens
• not enough balance to deposit

withdraw

• request withdraw from wrong token
• request withdraw from pool tokens with amount larger than current pool token supply

pool price

• divided by zero
• too small denominator causing overflow

[Hyung-bharvest]

Actual trading simulation with arbitrageurs and traders

• reference : https://web.stanford.edu/~guillean/papers/uniswap_analysis.pdf
• using random walk global price with normal distribution (every 6 seconds)
• arbitrageur model
  • arbitrageur trades when enough profit opportunity occurs by observing historical price and pool price
  • strategic order price and order amount following given risk model from the paper
  • assuming risk upon transaction delay, price volatility during latency
• traders randomly trading tokens with random order prices and amounts
  • traders tend to buy underpriced tokens more than overpriced tokens (following model from the paper)
  • various assumptions on trading volumes
  • order price above/below current price
  • limit orders remaining in orderbook
• strategic deposit/withdraw to/from liquidity pool following Markowitz portfolio optimization from the paper

From this simulation we can observe the risk/return profile of pool investor from different market assumptions

[Hyung-bharvest]

@dongsam , @defihyung lets close this issue if we think edge cases are enoughly tested from simapp.

[dongsam]

@dongsam , @defihyung lets close this issue if we think edge cases are enoughly tested from simapp.

yes we've tested enough, but there may be some edge cases that we haven't found, so we are going to look them up more and close this issue this week.