Am I Missing Issues in My Trading Bot Process?

I’ve been working on a trading bot with a simple strategy that decides when to buy, sell, and close trades based on timings learned from historical data. Here’s how the process works: * **Training**: The bot trains on 1.5 years of data across 21 symbols, aiming to find strategies that perform well overall. * **Testing**: Strategies are tested on 6 months of unseen data, evaluating performance symbol by symbol to select the best ones. * **Daily Updates**: This process is repeated every night, and the selected strategies are only used for the next trading day. A key observation is that not all symbols from training will replicate results during testing, but those that perform well in testing consistently performed well in training too. My goal is to replicate the test performance in live trading. In my view, the simplicity of the strategy, combined with the use of out-of-sample testing across multiple symbols, and a reasonable 60% win rate, should ensure that test results directly translate to live performance without overfitting. Am I missing any structural issues with this approach? Could there be factors I haven’t considered that might prevent test results from matching live trading performance? Any advice or insights would be greatly appreciated! Thanks!