This series will focus on the process necessary to build automated trading systems. Many individuals manually trade their strategies, often stressing out bar by bar what is going in the markets, yet want a way out of the daily grind. This is possible, but you need to be an expert in software development, machine learning, and broker APIs/Platforms. With years of study, you can implement these for yourself and take away the stress of looking at charts for hours on end.
1. Refine vagaries in manual strategy so that they can be configured as optimizable parameters in automated strategies.
2. Build out API integrations and databases to store information necessary to backtest and forward test strategies.
3. Implement, backtest, and forward test basic versions of the strategies, testing for 100s to 1000s of parameter combinations (if applicable) and validating against null models + benchmark models (like the SP500).
4. Compute advanced performance metrics to analyze results, including probability of overfitting on backtest results, alongside standard measures such as sharpe ratios, QQ-plots, intra-correlation of returns, and hedging factors that can reduce risk.
5. Analyze these results conclusively to find strengths and weaknesses of the algorithms, as well as any biases in the automated algorithms not present in the manual algorithms.
6. Implement improvements for the automated strategy, including, but not limited to:
A. Machine Learning integration for price action prediction, order-book dynamics models, and options bid/ask spreads.
B. Portfolio variance reduction strategies.
C. Tail hedging against defined macro economic events, such as Federal Reserve interest rate changes and inflation reports.
D. Reinforcement Learning, Deep Learning, and ML on alternative datasets (such as Twitter sentiment data).
7. Repeat 3–6 as necessary until results are satisfactory.
8. Implement and integrate with a broker (IBKR, MT4/MT5, TradingView, QuantConnect, and/or Crypto Platforms). This is one of the most challenging aspects as scalability of algorithms and upto 100s of settings come into play. Build databases to store state and historical trading data as well.
9. Paper trade the automated algorithms in (8).
10. Once results are satisfactory, implement the live version.
The goal is to build structured scalable software and algorithms that best match your trading hypotheses. Or, start with prebuilt automated systems and adjust them to mimic what you already do. In either case, take your hands off the computer and relax!