Screen Indian Stocks by ROE, ROCE & Revenue Growth With AI Agents

Use the Researcher agent to filter NSE equities by fundamentals (ROE, ROCE, revenue growth, margins) and hand the universe to the Quant agent for backtesting.

fundamentalsscreeningROEROCENSEai agentsstock screener india

A great quant strategy starts with the right universe. alphabench's Researcher agent can screen 60,000+ NSE/BSE instruments by 30+ fundamental fields — market cap, P/E, P/B, ROE, ROCE, revenue growth, margins, dividend yield, debt/equity — and pass the filtered universe straight to the Quant agent for backtesting.

This guide walks through a full fundamental-to-backtest workflow, end to end.

1. Describe the universe you want

In chat, describe the kind of companies you want to test on:

"Find Indian large-cap stocks with ROE > 18%, ROCE > 20%, revenue growth above 15% over the last year, and net profit margin above 10%."

The Researcher agent calls find_instruments with server-side SQL filters — typical response time is ~35ms even across the full universe.

2. Inspect the screen

The agent returns a ranked table. Each row includes the symbol, market cap, the metrics you filtered on, and a few you didn't (P/E, dividend yield, 1-year price return) so you can sanity-check the screen makes sense before committing to a backtest.

If the universe is too narrow or too broad, just say so:

"Tighten ROE to 22% and add P/E < 35."

The agent re-runs the filter in place.

3. Hand the universe to the Quant agent

When the screen looks right, ask for a strategy:

"Backtest a 50/200 SMA crossover on this universe over the last 5 years, equal-weighted across positions, max 10 concurrent positions."

The Planner agent hands off to the Quant agent, which calls backtest_basket with the screened symbols. The Researcher's filtered universe is shared via the message history — no copy-paste needed.

4. Compare against the broad market

Always benchmark a fundamental screen against a passive index:

"Compare the strategy's equity curve against NIFTY 50 buy-and-hold over the same period."

If the alpha is < 2% annualized after slippage, the screen probably isn't doing real work.

5. Common screens worth trying

  • Quality at a reasonable price: ROE > 18, ROCE > 20, P/E < 35, debt/equity < 0.5.
  • Compounders: revenue growth 1Y > 15%, net profit growth 1Y > 15%, market cap > ₹10,000 cr.
  • Margin expansion: net profit margin trending up, operating margin > 18%, gross margin > 35%.
  • Value with momentum: P/B < 3, 1Y price return > 15%, RSI(14) between 50 and 70.

6. Next step: validate

Run a walk_forward_test and a monte_carlo_test on any fundamental screen before deploying. Quality screens are particularly prone to survivorship bias.