This repository is a collection of python scripts for and documentation on a research project preforming statistical analysis on Cryptocurrency Markets.

The working report can be found at Documentation/mircomoedlling.pdf

• Scripts to manage downloading individual trade data from Binance and Coinbase APIs (found in /Download/)
• Scripts to record web sockets (previous versions work, the current implementation in is yet to be implemented using asynchronous style python)
• /lib/ Simple Library for order management, stores trades in a SQLite database and can be used to get orders (collections of trades that we can assume to be part of the same order).

• See /PreliminaryAnalysis/ For quick scripts not ordered very well performing basic analysis
• Analysis of order rates - seem to follow a Weibull distribution, indicating a sort of short term reaction
• BTC models - modelling the order states as Markovian (see report) and changes to this

• Testing statistical significance of markov model (given parameters are the predicted price distributions seen?)
• Application of Recurrent Neural Networks to calculate the statistical distribution of parameters given previous market data. (Use neural network to predict parameters of a multivariate distribution rather than values themselves). This will be achieved by altering the error function to be a measure of the predictive accuracy.
• Developing statistical models that have less of a dependence on the predicted parameters as current markov model just gives a normal distribution with the expectation and variance given by (functions of) the parameters predicted through linear regression. The models should be more dependent on the initial market state.