How do I start building a strategy? I have backtested many strategies, and some seem good enough for production deployment. But to reach the best strategy, is there a specific path or process that should be followed?
1 Like
There is no best strategy. There is only Win rate and Risk to Reward based SL and Target. The only problem that I see from the screenshot is too many strategies which is the cause of doubt. Just take 1 or 2, deploy, have a strict discipline for Target/SL and enhance the deployed strategies based on observations.
1 Like
yes
Maybe I didn’t phrase my question clearly.
At the moment, I am conducting systematic research on multiple trading strategies using different approaches, including technical indicators, pure price action models, and machine learning algorithms. For each strategy, I perform extensive backtesting and evaluate performance across various metrics such as monthly returns, CAGR, Calmar Ratio, Sharpe Ratio, Sortino Ratio, maximum drawdown, win rate, profit factor, and other risk-adjusted measures.
My question is more focused on the research methodology itself. Given the enormous search space of potential features, indicators, price-action patterns, market microstructure signals, and machine learning models, there are virtually endless permutations and combinations that can be tested.
As someone who is still developing expertise in quantitative strategy research, I would like to understand how professional quant researchers approach the strategy discovery and optimization process. Specifically:
-
How do they systematically explore the hypothesis space without falling into brute-force parameter optimization?
-
What framework is typically used for feature selection and signal generation?
-
How do they decide whether to use indicators, price action, market microstructure data, or machine learning techniques?
-
What methodologies help reduce overfitting and data mining bias during the research phase?
-
Are there any recommended books, research papers, blogs, or quantitative trading resources that focus on the scientific process of strategy research, alpha discovery, and strategy optimization rather than just individual trading strategies?
I am looking for resources that explain how to build a robust research framework for discovering and validating trading strategies, rather than resources that simply provide predefined indicators or trading systems.
Hi @Hemant_Kumar
Try this book
Finding Alphas: A Quantitative Approach to Building Trading Strategies
1 Like
I understood your question and my point to that was “Focus“ on a specific thing and enhance it. There is nothing called “Robust“ or “Perfect“. You need to understand that the basic difference between a retail researcher and a professionally backed researcher is how much capital he can deploy across the Stocks that has been analysed by the following
If you are building a product for others to use then yes what you have mentioned is the right thing and ML is what you have to do. A basic approach to identify a trend is KNN and ideally you can apply it for prediction for a scalp, intraday and swing based on time frames. I personally do AI assisted trading but its good to build conviction and confidence but will never guarantee anything. Hence we need to fallback to the basics which is proper Target & SL and specific focus on WinRate. In today’s AI age, its just getting the prompt right which is also the benefit with a professionally backed quant researcher.
1 Like