Skip to content

DOCUMENTATION.md

Latest commit

 

History

History
20 lines (14 loc) · 2.43 KB

DOCUMENTATION.md

File metadata and controls

20 lines (14 loc) · 2.43 KB

Documentation

This documentation provides details on the various pieces of the MEV agent code.

  • mev_project_interface: to know more about how the main function to process the user request given the venues and the JSON file storing such information.
  • classes: to know more about the classes employed in the project and how they work, get a look at

Brief explanation of how the code works

  1. Read the User intent and store the information of the required transaction in the order object;
  2. Read the Venues and store them in a list of venue objects;
  3. Create a market oject representing the non-directed graph. The edges are the coins present in the liquidity pools, whereas the edges are the venues. Edges store information about the venue and the liquidity of each token. Moreover, we endow each edge with a price function. Before using such a price function, however, we will need to assess the direction of the exchange of coins in the pool;
  4. Create a MEV agent agent object. This object then reads the order intent and the market graph;
  5. From the order intent and the market graph, agent determines the connected token pairs that allow performing the user requested trade;
  6. The agent creates the directed multi-graph connecting the requested sell token A to the requested buy token B. This is stored in agent.strategy;
  7. The agent now creates a set of paths, each of which is a list of the edges (i.e. the venues) that need to be visited along a simple path connecting A with B. The price function of the edge now can be determined since the direction of token swap is known;
  8. Having the list of simple paths to walk, the agent runs optimize_strategy(). This procedure creates a vector x of dimension N which is the number of simple paths connecting A to B. Defines the surplus function and the constraint functions. The surplus function is built in such a way that agent propagates the i-th x component through the i-th path and obtaines the i-th component of the bought b vector. The surplus is then obtained with the standard formula Γ(x) = sum(b(x)) - sum(x)/π. The surplus is optimized thanks to scipy.minimize.
  9. The optimal x maximizing the surplus under the constrained defined by the order is then used to update venue information and to compute the coin conservation error.