Stock screening in India: a complete guide

A screener is a filter. You describe a condition and it returns every listed stock that meets it today. That is the whole product, and it is worth saying plainly because most of the confusion around screeners comes from expecting more. A screener does not know which stocks will rise. It knows which ones currently look like something you described. How much that is worth depends on how well you described it, and on whether you ever checked that the description meant anything.

A screener versus advice

The distinction is not a legal footnote. It changes how you use the thing. Advice says buy this. A screener says these 23 stocks closed above their 200-day average on twice their normal volume today. The second statement is a fact about the tape. Whether it is worth acting on is a separate question and it belongs to you.

That has a practical consequence. Every screener output, including every list on this site, is a shortlist. Nothing about a stock appearing in a scan says it is cheap or sound or about to move. It says the conditions matched. The work of screening sits in choosing conditions that have historically been followed by something, and the last part of this guide is about how to check that.

End-of-day or realtime

Every screener runs on one of two kinds of data.

Realtime screeners refresh during the session, on tick or minute data. They exist for intraday traders. Somebody who enters at 10:15 and is flat by 3:15 needs to know at 10:16 that a stock has crossed its opening range. A list at 6 pm is no use to them.

End-of-day screeners run once the exchange publishes the session's final figures: closing prices, full-day volume and, on the NSE, the delivery data that only exists after settlement reporting. They serve anyone whose holding period is measured in days or longer. Swing traders, positional traders, investors adding on pullbacks.

The honest test is your holding period. If you hold overnight, the close is the number that matters, and a screen run on the close is the correct instrument for the job. Intraday data adds cost, noise and the temptation to act on half-formed bars. This site is end-of-day by design, and the delivery scans further down only exist because of it.

The setup families

Almost every scan anyone writes belongs to one of eight families. Each has a category hub here listing its presets, with a hit-rate replay on every scan.

Breakouts are price clearing a level: a 52-week high, a Donchian channel, a Bollinger band, yesterday's high. The canonical version is a close at a fresh yearly high on above-average volume.

where close is highest in 52w and rel_volume > 1.5

Run it as 52-week high breakout. The volume condition is what separates a breakout from a drift, and the Volume breakout scan makes that the whole test: twice normal volume with price up more than 3%.

Reversals are a stretched move giving way. RSI turning up from oversold, MACD crossing, price bouncing off the 200-day. The family's chronic problem is that most oversold readings belong to stocks in downtrends, which is why nearly every reversal preset here carries a trend condition beside the oscillator. The RSI guide goes through it properly.

Momentum is strength that persists: return over a quarter or a year, stacked moving averages, ADX. Its practical problem mirrors the reversal family's, in that a pure strength ranking hands you stocks at their most extended. The fix is buying strength on its pauses, which is what Pullback in an uptrend does with RSI under 40 in a stock still above its 200-day.

Delivery percentage is the NSE-specific family and gets its own section below.

Volatility is range compressing or expanding: NR7 days, Bollinger squeezes, gaps, the biggest single-day moves. Compression scans find stocks about to move without saying which way. Expansion scans find the ones that just did.

Candlestick patterns: hammers, engulfing bars, inside bars, morning stars. A pattern on its own is close to noise. A pattern at a meaningful location, a hammer at the 200-day or an inside bar inside an uptrend, is a setup, and the presets here attach the location for you.

Derivatives covers open-interest and futures data for the F&O segment: long build-up, short covering, futures at a premium or discount to cash. These describe how leveraged participants are positioned, which is a different kind of information from price.

Fundamentals puts shareholding, valuation and earnings fields next to price: promoter buying, low P/E in an uptrend, institutional ownership. I find them most useful layered under a technical scan rather than run on their own.

The families compose. A screening language lets you take one condition from each and state them in a single query, which is the main reason to prefer a language over a form full of dropdowns.

Why delivery percentage matters on the NSE

Every session, for every stock, the NSE publishes what fraction of the day's traded volume was actually delivered, meaning transferred between demat accounts rather than bought and sold back inside the session. Most exchanges do not publish this at all. Most Indian screeners treat it as a table column you can look at rather than a field you can scan.

It matters because volume on its own is ambiguous. A stock trading three times its normal volume could be institutions accumulating or day traders churning, and the price bar looks identical. Delivery splits the two. High delivery on a heavy day means the buyers kept what they bought. High delivery on an ordinary day, while the stock grinds higher, is the quieter case and the one I find more interesting:

where delivery_pct > 60
  and rel_volume < 1.2
  and close > sma(50)
  and return_1m > 0

That is Quiet accumulation, position-building without a headline. The delivery percentage explainer covers how to read the figure against a stock's own baseline, and why F&O stocks distort it.

How to test a scan before trusting it

This is the step that separates screening from folklore, and it is the one that got me building in the first place.

Chartink has had a backtest since 2019, free, nine years deep on daily bars. It will tell you which stocks matched your scan on a Tuesday four years ago, and the sectors they sat in. What it reports is the trigger: symbols, dates, sectors. It does not report what those matches went on to do, and that was the number I could never get.

Every scan is a claim: stocks matching these conditions tend to do X afterward. Checking the claim means a replay. Run the scan as of each past session, record what matched, measure what those matches did over the next 1, 5 and 20 sessions. Every scan page on this site carries that replay across the past year of NSE sessions.

Then read the result honestly, because it is a sketch. Close-to-close returns. No brokerage and no slippage. The universe as it stands today, which leans survivorship-positive by an amount I cannot put a number on. It is not a backtest of a trading system and it is nowhere near a forecast. What it tells you is whether the pattern has any historical tendency at your holding horizon, and roughly what a failed match costs. A scan that reads beautifully and replays at coin-flip odds is a story, and ten minutes of replay sorts stories from setups before money is involved.

Three things fake a good replay: testing only on stocks that exist today, using information that was not available on the signal date, and tuning thresholds until last year looks perfect. Survivorship, lookahead, overfitting. The backtesting guide works through each one, and the methodology page states exactly how the replay here is computed and where it stops.

Choosing a tool

Four questions decide most of it.

1. Does it run on the data you need? End-of-day for overnight holds, realtime only if you are flat by the close. If delivery percentage matters to you, check it is a field you can scan and not a column you can only read.
2. Can you state a condition precisely? Dropdown builders cap out fast. "RSI crossed above 30 within the last 3 bars, and the 200-day is rising" is one line in a language and often impossible in a form.
3. Can you check whether a scan works? If the tool cannot show you what past matches did, you are trusting every scan on faith.
4. Can you get results out? Alerts on your saved scans, an export of a run, an API if you build things.

The compare pages put this site against the tools most Indian traders start with, feature by feature, including the places where those tools are the better fit. Trendlyne's GuruQ tier undercuts my Pro plan on price and the compare page says so.

The vocabulary

Screening has a lot of jargon and most of it names a simple calculation. The glossary defines every field, every indicator and every pattern the scanner exposes: what it measures, the Sift name, working thresholds, a runnable example query. Reading the entries for the two or three indicators in a scan you are about to rely on is the cheapest due diligence available.

If you do one thing after this guide, do that and then open Quiet accumulation and run its replay at 20 sessions. The number will either be roughly what you expected or it won't, and both answers are worth having.

Not advice

A screen returns stocks that matched criteria you chose. Nothing on this site says which of them to buy, and the hit-rate replay describes what past matches did rather than what future ones will do. Treat every list as the start of your own work.