Optimization Techniques to Improve Accuracy

AI stock screeners can give powerful results, but they’re not perfect out of the box.

Sometimes they include too many irrelevant stocks. Other times, they miss important ones.

That’s why optimization is important — it’s about tuning the way you use AI, so the results become sharper and more accurate.

Start Simple, Then Add Layers

Don’t overload your prompt with 10 filters at once. Start with 2–3 important ones and add layers step by step.

Example:

• First: “Show me auto stocks under ₹500.”
  
• Then: “Now filter those with ROE above 12%.”
  
• Finally: “Add RSI values below 30.”
  

This reduces confusion and ensures the screener narrows down in a logical way.

Use Realistic Ranges

If your filters are too strict, you’ll miss good stocks.

Instead of saying “ROE above 20%”, start with “ROE above 12%”.

Instead of “P/E below 10”, try “P/E below 20”.

Broader ranges make the screener more inclusive, and you can refine later.

Cross-Validate with Multiple Prompts

One prompt can never capture everything. Run the same idea in different ways. Example:

• Prompt 1: “Find banking stocks with NPAs under 2%.”
• Prompt 2: “Show banking stocks with high ROE and low debt.”

Compare the two outputs. If some names repeat, they’re stronger candidates.

Mix Fundamentals and Technicals

Accuracy improves when you combine both sides of analysis.

Example:

“Show IT companies with 3-year profit growth above 10% and trading above their 200-DMA.”

If a stock passes both tests, it’s less likely to be a random fluke.

Keep Refining Over Time

Markets change. What worked in 2021 may not work in 2025. So, keep adjusting your prompts based on outcomes.

Example: If your screener keeps throwing up high-debt companies, add a rule: “Exclude debt-to-equity above 1.”

Small refinements like these keep your results fresh and reliable.

Final Takeaway

Accuracy in AI screening doesn’t come from one perfect prompt.

It comes from iteration, realistic filters, and constant refinement.

Start simple, test different prompts, combine fundamentals with technicals, and adjust as markets change.

The goal is not to create flawless shortlists, but to get results that are practical, consistent, and closer to how real investing decisions are made.