Get the relevant historical data, load it into proper classes in Python/C++ and try to find setups and strategies that have worked over time. Find the best places for entry/exit, and the potential upside/downside. Look for ways to identify the big profit or big loss days at the start. When you are satisfied, deploy it. Take manual trades at the start and automate it when you have a enough confidence and a big enough capital.
For example, take the infamous 9:20 short straddle, try to find how often it works? What is the optimum profit to wait for without risking a big loss? Is holding overnight a good idea? Has that changed over the years? How to identify the days when it doesn't work so you can avoid them or just do a long straddle on those days instead?
For example, take the infamous 9:20 short straddle, try to find how often it works? What is the optimum profit to wait for without risking a big loss? Is holding overnight a good idea? Has that changed over the years? How to identify the days when it doesn't work so you can avoid them or just do a long straddle on those days instead?