Money

Volume was the only thread in our biggest movers, and it was unknowable at entry

By Strota Newsroom · 2026-08-17 · How Strota reports

backtestrelative volumeintradayresearchIndian equities

Ranking every trade by how far it actually ran left one variable standing: relative volume. The catch is that the measure counts a whole session's turnover, most of which arrives long after the position is already open.

We wanted to know what the trades that actually ran had in common. Not the ones that closed a shade green, but the handful that travelled far enough to matter. So every one of 391 completed intraday trades was ranked by how far it moved in its favour while it was open, and the leading 30 were pulled out for inspection. Each of those had run at least 5.3%. Twenty-seven of the thirty were long positions.

One variable did nearly all of the separating, and it was volume. The big movers traded, at the median, 8.9 times their twenty-day average turnover. Everything else in the sample managed 2.1 times. Filter the trade history on that single condition - anything doing five times its normal volume or more - and the win rate climbs from 57% to 69%, while the share of trades that became big movers reaches 49%.

Here is where the good news has to be interrupted. Entries in this system fire at around 9:30 in the morning, and the volume figure being praised is the whole session's volume, roughly 90% of which changes hands after the trade is already on. A large move and a large day's turnover are produced by the same event, at the same time. Sorting trades by the completed day's volume is therefore close to a tautology dressed as a signal: stocks that ended up trading enormously also, unsurprisingly, ended up moving a lot.

The descriptive picture is still worth having, because it says what a big mover tends to look like. Among the leading thirty longs, 81% had broken above their highest price of the previous twenty sessions, against 43% of the rest. Some 73% were trading above the volume-weighted average price - the running average of the day's trades weighted by size, which many desks treat as the day's fair value - versus 51% elsewhere. And 87% were sitting at the session's high when measured, against 70%. The month leading in was quiet rather than hot: a median prior twenty-day return of +0.8% for the movers, +3.4% for everything else.

Two things people expect to see were absent. Gap size did not separate anything: 60% of the big movers opened more than 1% above the previous close, and so did 63% of the also-rans. More surprisingly, the internal tag for a sudden one-minute volume spike was less common in the movers, 67% against 84% - a deficit of 18 percentage points. What distinguishes a stock that travels is sustained heaviness through the session, not one violent print at the open.

To test whether that picture predicts or merely describes, the long trades were re-run under each condition separately. The unfiltered baseline is 284 trades, a 57% win rate, an average of +0.201R, and 10% of trades ending as big movers. R here is profit measured in multiples of the risk taken on that trade, so +1R means the position made exactly what it stood to lose. Requiring the twenty-day breakout leaves 115 trades at 60% and +0.336R, with the big-mover share at 19%. Requiring heavy volume instead leaves 39 trades at 69% and +0.80R, with 49% big movers.

Combining the filters adds nothing. Breakout plus volume gives 34 trades at 68% - indistinguishable from volume on its own, and no better. Layering the quiet-base condition on top of both cuts the sample to 18 trades and drags the win rate back down to 61%. The reading is that the breakout, the position above fair value, the close at the day's high and the sleepy month beforehand are all downstream of the same thing. They are correlates of heavy volume, not independent sources of edge.

Which figure from that exercise deserves trust is a separate question. The average of +0.80R is driven by the right tail - a couple of outsized outcomes doing most of the lifting - and would not survive their removal. The 69% win rate is the sturdier of the two. Both are in-sample results, produced by hunting for the filter in the same data used to score it, which is exactly the process that inflates numbers. Earlier work in this programme watched an in-sample 80% hit rate settle at 55% once it was run forward. And the samples are thin: 39 trades for the volume filter, 18 for the fully stacked version.

There is a concentration problem underneath all of it. A single trending session accounts for close to 59% of the entire profit in the trade set, and that same day clustered the big movers together. The obvious check - remove it and see what survives - has been carried on the to-do list through several rounds of this work and still has not been run. Until it is, every number above is partly a description of one very good day.

A second idea tested in the same pass came back flatly negative, which is worth reporting because it closes a door. The question was whether a sudden price skip on the one-minute chart carries information a fast system could trade. Across the derivatives-eligible universe, 772 such skips were identified, each a one-minute move larger than 4 times the stock's trailing volatility and at least 0.25% in size.

Continuation after those skips ran between 40% and 51%, which is a coin flip or slightly worse, and the mean signed move failed to clear trading costs at every horizon examined. Every bucket showed immediate reversion in the following minute, with downward skips snapping back hardest at 37% to 43%. Larger skips did not sustain better than small ones. That pattern reads as a mechanical gap in liquidity rather than news reaching the tape, and entering on the skip amounts to buying a local extreme that fades. The planned live detector for it was abandoned.

What would turn the volume result into something real is already identified: relative volume measured over the first 15 to 30 minutes of trading, compared with a typical first 15 to 30 minutes for that stock. That version is knowable at the moment of entry and moves with the full day's figure, so it carries the information without the look-ahead. It has not been tested yet. As things stand, this is a finding about measurement rather than a live opportunity - the study located a strong hint about where to look, and then disqualified its own evidence.

Volume was the only thread in our biggest movers, and it was unknowable at entry

More money stories