Volume breakout screener for Indian stocks

Price can drift to a new high on nothing. Volume cannot drift, because every share of it is somebody's money changing hands. That is why volume sits inside nearly every breakout scan worth running: it is what separates a level being cleared from a level being wandered past.

Defining high volume

Ten lakh shares is heavy trade in one stock and a rounding error in another, so any volume condition worth writing compares a stock to its own history:

where volume > 2x avg(volume, 20)

Twice the twenty-day average is the conventional threshold and the twenty-day window is the conventional baseline, long enough to smooth over one noisy week and short enough to reflect the stock's current regime. The shorthand for the common case is relative volume, rel_volume, which is today's volume over that same 20-day average. So rel_volume > 2 says the same thing.

The three standard scans

1. Volume breakout. Unusual volume with price up over 3% and holding above the short-term average:

where volume > 2x avg(volume, 20)
  and change > 3
  and close > sma(20)

The general-purpose one. It usually marks news, a result, or an institution starting a position. Run it: Volume breakout.

2. Volume shockers. The most-cloned scan on Chartink, and worth having for the reason people clone it: a shorter ten-day baseline and a bigger price move, tuned to answer what happened today rather than to time an entry. Run it: Volume shockers.

3. Channel breakout with volume. Volume confirming a price event that has a precise definition, clearing the previous day's 20-day Donchian channel top:

where close > donchian_upper[-1]
  and volume > 1.5x avg(volume, 20)

Note the [-1]. Today's high is part of today's channel, so the naive version triggers on itself. I wrote the naive version first and it matched almost everything, which is how I learned to distrust any breakout query that returns a comfortable number of names. Run it: 20-day channel breakout.

The mistakes that ruin volume scans

Screening the whole market. In the bottom half of the NSE by liquidity, twice average volume can be one operator having a busy morning, and the spike means nothing at all. Every scan above runs on the top 500 stocks by turnover by default. Widen the universe deliberately.

Ignoring where the volume closed. Twice-average volume on a day that closed flat and mid-range is an argument, and half of that volume was sellers. The price conditions, change > 3 and close > sma(20), exist to keep only the sessions the buyers clearly won.

Confusing churn with conviction. On the NSE you can check this directly, which most screeners never do. The exchange reports delivery percentage, the fraction of the day's volume actually taken home rather than squared off inside the session. A volume spike on 25% delivery was day traders passing shares around. The same spike on 65% delivery changed who owns the stock. The delivery scan family covers this angle, and delivery with a breakout puts the two ideas in one query.

Only looking at single days. Accumulation often shows up as many slightly elevated days rather than one loud one. On-balance volume catches that: OBV at a six-month high finds stocks where cumulative buying pressure has already broken out while price has not.

Test it before you trust it

Every scan linked above replays against the last year of sessions: for each historical trading day, who matched and what happened next. It is close-to-close with no costs or slippage, so treat it as a sketch. Volume breakouts have a specific known failure mode, where the spike day is the move and buyers of the close get the retrace, and the replay makes that visible per scan instead of leaving it to folklore.

Run the replay on Volume breakout at the 1-session and 20-session horizons and compare the two. If the first number is good and the second is not, you have found the retrace. The breakout hub has the variants side by side.