From an academic perspective, regardless of whether you use machine learning, tree models, or neural networks, it is very difficult to achieve an 80% win rate in a statistical sense every year. The essence of backtesting is to use known historical data to test whether a strategy remains robust under various conditions. However, robustness does not mean that the returns of a strategy can be replicated every year. A strategy that performs well in 2024 does not guarantee that it will also be profitable in 2025. Retail investors in the cryptocurrency market generally do not perform quantitative analysis.