RSI screener guide for NSE stocks

"RSI below 30" is probably the most-run scan in retail screening. Run bare, it is one of the worst. The list it returns is dominated by stocks that are oversold because they are dying: downtrends produce low RSI readings constantly, and buying them is catching knives with a number attached.

One condition fixes most of that. It helps to know what RSI is measuring first.

What RSI is

The Relative Strength Index compares the average size of recent up-days to recent down-days over a window, almost always 14 sessions, and squashes the ratio into 0–100. Above 70, the recent tape has been one-sidedly bullish. Below 30, one-sidedly sellers. What it measures is one-sidedness. Nothing in the formula says a stock is cheap, only that its recent sessions have leaned hard one way.

One-sidedness tends to resolve. That reversion is the entire edge, and it only pays when the larger context says the resolution should be upward.

The one-condition fix

where rsi(14) < 30 and close > sma(200)

The 200-day filter splits the oversold list into its two populations. Oversold above the 200-day is a healthy stock having a bad fortnight, with an owner base in profit and defending it. Oversold below the 200-day is a downtrend doing what downtrends do, and it can sit under RSI 30 for months. Same reading, opposite trades.

The prebuilt version adds one more refinement, waiting for the turn rather than buying the reading:

where rsi(14) crossed above 30 within 3 bars and close > sma(200)

Crossing back above 30 is the first measurable evidence the selling stopped. Run it: RSI oversold, turning up.

The level nobody screens for: 50

In a real uptrend, RSI rarely reaches 30 at all. Pullbacks bottom near 40 to 45 and turn. Waiting for the textbook 30 means missing every routine dip in every strong stock, which is exactly what I was doing until I went and looked at where the dips in strong NSE names actually bottom. Two scans cover the trending case.

• RSI reclaiming 50: the midline cross fires earlier and far more often than the 30 line inside uptrends. It marks "pullback over" rather than "washout over".
• Pullback in an uptrend: RSI under 40 with the 200-day intact. Rare by construction, so the list is short and worth reading on the days it isn't empty.

For the hair-trigger version, Stochastic RSI oversold applies the stochastic formula to RSI itself and reaches extremes on dips plain RSI barely registers. It fires often. Treat it as a shortlist builder.

The other side

RSI above 70 in a stock you hold is not automatically a sell. Strong trends spend weeks above 70, and "overbought" is often just "working". What it does give you is a reason to move a stop up, and in weak stocks, a fade candidate. RSI overbought keeps the strong-trend context with close > sma(50), so the list reads as extended rather than doomed. The wider reversal family has the CCI, Williams %R and stochastic variants of both sides.

Check the levels against reality

Every threshold in this guide is convention: 30, 40, 50, 70. Convention is checkable. Each scan page carries a hit-rate replay across the past year of NSE sessions, close-to-close and without costs, showing how often the signal resolved upward, over what horizon, and what it cost when it failed.

Spend ten minutes replaying the oversold-turn scan beside a bare RSI-below-30 query. The gap between those two numbers is the argument this whole post is making, and you will believe it more from your own screen than from me.