The Micro-Price (Stoikov, 2017) is a high frequency estimator of future prices. (See: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=2970694)

This repository contains source code that calibrates Stoikovs Micro-Price model to BitMEX quote data and creates optimal limit order postings at the best bid and offer using stochastic optimal control and by solving HJB-QVI equation.

The weighted mid-price is a noisy and counter-intuitive measure of the "fair-value" of a traded asset. The Micro-Price is the expected mid-price in the future given a Markov-Chain that drives the future evolution of mid-price using Euler finite difference scheme.

The Micro-Price may be combined with stochastic optimal control to create a market making strategy that accounts for the micro-price as a short-term prediction of the asset price. For a mathematical treatment see: https://gist.github.com/sebjai/c6ff3850dea37d28d3fa7d3aef59722b

In the model, the short-term alpha component is assumed to be mean-reverting. Here, we assume that this process is the Ornstein-Uhlenbeck process which is easy to estimate from data using linear regression.

Trade size distribution for marketable buy orders. Mean of the distribution is marked with red vertical line.

Trade size distribution for marketable sell orders. Mean of the distribution is marked with red vertical line.