README.md

Overview

This repo contains a tool for downloading and analyzing arbitrage opportunities by solving the longest path problem for a network of assets.

MarketTrees

A MarketTree is a bi-directed graph where each node represents an asset, and bi-directed links between nodes represent the bid and ask price at the time of the latest data retrieval. Thus, the product of links along a path between nodes represents the quantity of the end asset obtained after following a path connecting the beginning and end asset .

Longest Path Example

For example, take A, B and C separate assets, each represented as a node in the graph, which are traded among each other at independent bid and ask prices (each represented as a directional link connecting two nodes). Trading 1 unit of A for B at an ask price of 0.25 would yield 0.25 units of B. Trading 0.25 units of B for C at the ask price of 2 would yield 0.5 units of C. Thus the quantity of the end asset obtained is the distance (i.e. the product of the links) along the path from asset A to asset C. This example models a realistic trading scenario as the bid/ask prices between all possible pairs of assets are competitive - executing a trade between a single pair of assets, and then executing the opposite of that trade yields a distance of less than 1 (similar to how selling at the bid price and then immediately buying at the ask price in a market would yield less than the initial starting asset).

Executing one last trade to exchange 0.5 units of C for A at an ask price of 2 yields 1.5 units of A (which is more than we started with, even with the competitive bid/ask prices!). Thus, any cyclical path (i.e. starting and ending at the same node) with a distance greater than 1 represents an arbitrage opportunity - a series of trades which would yield a greater quantity of the starting asset after execution.

Repo Structure

Market_Tree.py contains the implementation of this data structure, and a function called get_max_path which finds the longest path that starts and ends at the root of a specified target node. MarketTrees need not be limited to triangular graphs as shown in the example - the longest path can be found for any arbitrarily complex bidirectional network in O(n*p) time, where n is the number of nodes and p is the number of connections in the network. To avoid infinite cycling, each link between nodes is restricted to be used only once in a path .

Arbitrage_tester.py contains a basic set of unittests to test the Market_Tree implmenetaiton.

Kucoin_Arbitrage_ETL.py is a script that loads the latest Kucoin trading data into the Market_Tree structure. The script then finds and executes the greatest arbitrage opportunity, if one exists, using functions from KuCoin_Trade_Functions.py.