Choosing a stock screener: what to look for

I built PatternsRadar because of one question Chartink would not answer for me, so start by weighing the source: I am not a neutral party here. Chartink made NSE screening free and fast and shareable, and its library of public scans taught a generation of Indian traders what a screener is. I used it for years. The five things below are what I would check on any screener, including this one.

1. Does it tell you what happened after the match?

Every screener answers "who matches today". Several will also replay a scan across past sessions and list what matched, Chartink among them, free and nine years deep on daily bars since 2019.

The distinction that matters is what comes back. A trigger history hands you symbols and dates: which stocks matched, when, in what sector. An outcome measurement attaches prices to those matches and reports a hit rate and an average return over a stated holding period. Both get called backtesting and they are not the same product. Without the second, scan-building runs on folklore. You run "RSI below 30" because a book said so. With it, every scan is a claim you keep or discard on evidence, and most of mine got discarded. On each scan page here, the 52-week high breakout for instance, the hit-rate panel reports what matches did over the next 1, 5 and 20 days across 250 sessions, close to close, with no costs or slippage modelled. That panel is the feature I would not give up.

2. Can it express conditions about time?

Form-based screeners are good at snapshots. RSI below 30, price above the 200-DMA, done. The conditions that describe actual setups are about sequence and persistence:

rsi(14) crossed above 30 within 3 bars

close has been above ema(21) for 10 bars

volume rising for 3 bars

close is highest in 52w

Where a screener has no event operators, users fake them with towers of offset comparisons, which is what most complicated Chartink scans turn out to be under the hood. A readable query language is not decoration. It decides which questions you can ask, and whether you can still read your own scan six months later.

3. Is delivery percentage a first-class field?

Delivery data, the fraction of traded volume actually taken home, exists on Indian exchanges and almost nowhere else. Most screeners bury it in a column. It belongs in the scan logic, combinable and testable like any other field: delivery surge with a price move is a better signal than either half alone. The delivery scan family shows how far it goes.

4. How deep and how honest is the data?

Two questions expose most tools. How far back does the daily history go, far enough to cross regimes or just a recent window? And is it properly split-adjusted? One unadjusted bonus issue sitting in the lookback quietly poisons every moving average and every test that touches it. Here it is 30 years, split-adjusted, rebuilt after each close.

5. Can you get results out programmatically?

If a screener is part of your process, its output eventually has to reach the rest of your process: a script, a sheet, an alert pipeline, an agent. An API turns a screener into an ingredient. The MCP endpoint exists so agent tools can run scans without a browser.

Where Chartink is still ahead

It scans during market hours and this dataset updates once, after the close. Its community library is a decade deep and mine is not. Its trigger history reaches nine years back, against the 250 sessions replayed here. If your process fires on intraday triggers, Chartink is the right tool and I would keep using it.

The short version of the difference: Chartink shows you which stocks matched, and this shows you what they did next. Run any scan here, open the hit-rate panel, and decide whether you want to screen without one again. The feature-by-feature table is maintained separately.