We have over 180K backtests in our database and decided to analyse which features using XGBOOST and SHAP most lead to profit. This extensive report is extremely insightful, yet should also be taken with a grain of salt - given this is just backtest data.

Supporting excel book for above video. Copulas enable us to formulate a deep understanding of the true dependency between financial assets - in normal, booming and plummeting markets. We can gain insight into how the specific dependency between assets change in such times and thus also gauge their conditional probability in reaching their respective price or returns deviations from the mean.

Supporting excel book for above video. Cointegration exists when there is a linear relationship between 2 or more assets of some kind that is stationary (integrated of order 0). We will use the Engle Granger method for Cointegration testing, with the stationarity test being conducted using the Augmented Dickey Fuller test.

Supporting excel book for above video. For an asset to be weakly stationary, means that its mean, variance and auto-covariance must by constant.

Supporting excel book for above video. A GARCH model allows us to model current, long term and historical volatility for a given asset. We use apply MLE to modelling volatility for BTC and ETH along with a quick and convenient way to deciding how much of our capital to allocate to each asset.

Supporting excel book for above video. Maximum Likelihood Estimation is a very useful estimation technique for all sorts of models and processes, not just ARMA. So using it well know could help you with understanding and implementation of many other modelling processes in financial time series.

Supporting excel book for above video. Ordinary Least Squares (OLS) is a powerful and fast linear regression estimation technique which allows us to assign values to relationships between financial assets, sentiment or just about any data you can think of.

Supporting excel book for above video. Autocorrelation does seem to appear to be present with certain Cryptocurrencies. Not one to miss. Autocorrelation can be a powerful tool and will lead on nicely to understanding other time series analysis.

Supporting excel book for above video. Pearsons Correlation coefficient is a useful concept to understand when seeking to determine whether assets behave jointly. Pearsons correlation coefficient assumes that the joint probability distribution of assets returns follows a bivariate normal distribution (when comparing two assets). Even though this has its pitfalls when it comes to financial time series analysis, it serves as a great way to learn about math notation for mean, variance, covariance and correlation.

Position sizing can make or break your trading results. Use the Kelly Criterion to identify the mathematical optimum amount of capital which you should allocate to any given trade (or even a fraction of this) to ensure the maximum gain for your trading strategy.