The PSX Gap-Down Study: One of the Only Positive-Median Trades We Have Ever Measured

Most trading patterns pay through a thin tail of winners while the typical trade loses. This study found a rare exception on the Pakistan Stock Exchange — and, on the way, a textbook example of how backtests manufacture fake edges.

First, the mirage: an "18x edge" that was pure arithmetic

The first measurement pass asked: after a deep gap-down open, how often does the HIGH reach the open × 1.10? Answer: 46% for gaps ≤ −10% — an apparent 18.14x lift, the biggest number of the whole study. It was mechanical. A stock gapping down 10% opens at 0.90× yesterday's close, so "open × 1.10" is only 0.99× yesterday — the stock merely has to recover its own gap. Re-measured against yesterday's close × 1.10 — the honest question:

Against a 3.674% base rate the deep gap-down is 1.95x, not 18x — and the direction inverts: for a genuine +10% move, gapping UP is the better predictor (3.59x).

The real finding: buy the gap-DOWN open, sell at the close

A −5% to −10% gap-down open, sold at the close, carried a median of +3.1% to +3.5%. And the median low from the open is 0.00% — the open IS the day's low in the typical case, so waiting for a better price does not work.

The mirror is just as important: buying a gap-UP open measured −2.14% by the close, after first trading 4.26% below the entry. Chasing morning strength on PSX is measurably wrong.

Fillability — checked on the minute tape, and it passes

From 627 gap-down sessions with minute-level tape (Jun 30 – Aug 13, 2026): a median 210 of ~360 minutes traded at or below the open, with a median Rs 17.13mn of value available there; 71% of sessions cleared Rs 2mn at or below the open. On those same sessions, close-vs-open: median +2.92%, mean +3.66%, positive on 75%, worst −7.60%, best +22.26%.

The traps

• The sweet spot is −7% to −10%, not the deepest gaps. Same-tape medians: −10..−7 → +6.47% (n=146); −7..−5 → +4.09%; −5..−3 → +2.08%; but ≤ −10% → +0.00% — several of those names locked at the lower circuit and never traded (Rs 0.00mn at or below the open). The gap is real; the trade is not.
• A −5% to −10% gap usually means news. The 0.00% median low says the average one holds the open; the average is not the one that hurts. Position sizing and a hard stop are not optional.
• The edge is intraday only. A separately measured "gap-down that recovered intraday" signal FAILS on day 2 — holding overnight gives the edge back. Sell at the close.
• The literal dream — "buy morning, +10% by afternoon" — is not available. Real same-day +10% is a 3.7% base-rate event and no gap bucket lifts it past ~13%.