Initially, the goal of the project was to create an algorithm using machine learning to compile an optimal portfolio of digital assets. An algorithm was proposed for predicting the value of a cryptocurrency for the next day and correcting it for the next day. However, the task of predicting the value of cryptocurrencies the next day turned out to be extremely difficult. As a result, the developed algorithms showed a stable minus. As a consequence, the task of the project was simplified to testing existing strategies. We looked at 3 main strategies and 2 additional ones. The first strategy was to buy and hold bitcoin. This strategy was chosen as a benchmark because it was the easiest investment to buy and hold a coin. Therefore, we compared other strategies with it. The second strategy was to buy bitcoin and ethereum in equal proportions and keep balancing every day. The third strategy was to buy the top 10 most capitalized currencies and keep rebalancing every day. As 2 additional strategies, there is a similar third algorithm, but for the top 5 and top 10 by cryptocurrency capitalization. Rebalancing is a strategy that involves buying and selling assets to maintain a given proportion. All strategies take into account that a commission for the purchase of 0.025% is collected.
Sharpe ratio, Sortino ratio, maximum drawdown percentage and profitability were used as measures for comparing investment instruments. As a result of evaluating portfolio formation strategies, the best was to buy bitcoin and ethereum in equal shares with daily rebalancing to maintain a given ratio.
React
Typescript
Django
Laravel
Facebook
Microsoft
Google
Alibaba
D3
Tencent