CryptoMonkeys

This is the first iteration of simple strategies for building a cryptocoin portfolio. The initial idea was to verify how a simple buy-hold portfolio of 5 random smaller coins, starting from the 25th placed coin in Binance, would behave. This first strategy was labeled the 'Old Monkey', due to the joke about portfolio managers and monkeys being indistinguishable in the long run should the market be efficient. The idea was then expanded to have 4 additional strategies/monkey types, which would be executed at a given number of days and decide which coin to sell: 'Crazy Good',selling the coin which would generate the biggest profit/smaller loss; 'Crazy Bad', selling the worst performing coin; 'Crazy Alternating', which would alternate between selling the best coin and the worst coin, just like the previous 2 monkeys, performing as the Crazy Good Monkey in even intervals and as the Crazy Bad on even intervals; 'Crazy Random', which just sold a random coin. In order to evaluate the different strategies, every monkey started out with the same randomly generated wallet of coins avaliable on Binance at the defined starting date, which means that every monkey's initial wallet is identical to the Old Monkey's. It was also defined that every monkey must hold 5 coins at all times, so the strategies only defined, for now, which coin to be sold. The replacement coin would be randomly selected from the avaliable coins, but could not be the same as the sold coin. The test used data from November 1st, 2017, up until the date of the final version of the algorithm, the 12th of June, 2019, and 2000 seeds, generated from a fixed seed, were used to generate 2000 monkeys of each type. We also tested multiple time intervals for the execution of the strategies (5,10, 15,....,45 days), which means that we had 37 combinations of monkeys, each tested 2000 times. After some evaluation, it was decided that, since the portfolios had a huge variance, the correct thing to do would be take the vector of the returns for every combination and calculate their sharpe ratios, using the return on a buy-hold strategy for Bitcoin as the benchmark. It's important to note that the funds utilized for the monkey's strategy is initially in USD, and is then converted to BTC for the purchase of the wallet. After the end date, the wallet is then sold for BTC, and the BTC is then exchanged back to USD. Therefore, the returns are calculated in USD, in order to provide a direct and fair comparison to the benchmark strategy. The algorithm is still under development, and will be updated. Feel free to add forks and suggestions!