Strategies to "Bitcoin-Gold" Trading Decoupling the Qualitative Decision and the Quantitative Investment

Abstract

: As investment fever rises, investment strategy is a critical choice for investors. In this paper, based on the price data of gold and bitcoin from 9/11/2016 to 9/10/2021, the corresponding mathematical models are established by the LSTM, evaluation model, and single-objective optimization model in an attempt to obtain the optimal investment strategy. Firstly, the price is predicted by the LSTM model after data cleaning. The evaluation system is constructed from three perspectives: market trading intention, price stability and market environment trend. Based on the scores and thresholds of the evaluation system, the trading directions (buy, sell, hold) are obtained qualitatively. Then, based on the corrected predicted price, a single-objective optimization is established for maximizing total assets, obtaining the final investment strategy, where an initial $1,000 asset is able to reach a final total asset of $134,226 after five years. Secondly, we establish the evaluation system of investment strategies and establish three evaluation indexes of Sharpe Value, Max Drawdown, and Interest Rate in terms of both return and risk. We construct four new investment strategies with different types of investment product and model selection. Comparing the five investment strategies, we can find that our investment strategy has the highest Interest Rate, higher Sharpe Value, and smaller Max Drawdown compared with other investment strategies. Thirdly, we test the impact of transaction costs on the model and learn that the model is more sensitive to the change of bitcoin transaction costs and not sensitive to the change of gold transaction cost.