Gap-up mornings, day-of-week myths, profit locks, and the closing auction

Across 2,905 trading days since 2015, after a ≥1% day followed by a ≥1% same-way gap (23 cases) the day itself fades more often than not — open-to-close positive only 39% of the time — but the gap holds: only one in four such sessions traded back below the previous close within three days. Two, three or four consecutive up closes carry no information at all; the next day is up 53–54% of the time, the same as any day. The mirror (down day plus down gap) bounces intraday and resumes lower over the week.

The gap is the payoff. Positions carried into those mornings ended up averaging ₹22.7k per signal, 82% winners — money already in the bag at 09:15. Measured forward from that open to the trail exit they were a coin flip around zero, with 55–60% exiting the same day on the first pullback. Signals entered on those mornings — chasing the gap — were the worst entries in the book (−₹2,637 per signal versus +₹1,153). We later tested skipping those entries as a rule: the whole benefit sat in 2020's crash-rebound days; since mid-2023 the filter costs money. Rule not adopted.

Not in 2025–26. Ninety-five such days versus 258 ordinary ones: nothing two standard errors apart, signs flipping between the two years, no early "pinning" to strikes, no predictive link to the expiry day. Expiry day itself closed down 62% of the time in these two years — a lead for a longer test, not a conclusion.

In the last two years Wednesday was the strong day (+0.14%, 56% up, 2.2 standard errors) and Tuesday the weak one. Out of sample on 2015–2024 Wednesday was nothing and Tuesday was the best weekday. The weekday effect is real structure that moves — it followed the expiry day when NSE moved it from Thursday to Tuesday — so a fixed weekday rule would have been wrong-footed at the switch. One fact for risk rather than edge: the weekend night carries about 1.5× the gap of any other night, both ways.

We tested a "lock and pause": once a position is three boxes in profit the stop ratchets to half the best profit, and after a lock exit the same-direction re-entry waits for a new pivot. It was the first exit rule in the programme to beat the plain trail on both data sources — about +7%, every rolling two-year window, every volatility regime, and statistically sound once the day-shifting effect in the comparison is handled (t ≈ 2.3 weekly). It still was not adopted: it loses 5% on BankNifty, a quarter of random lock rules do as well, almost all of the gain was on one side, and — the decisive point — on the forty-five biggest days it is identical to the plain trail on forty-one of them. The trail already exits on the first pullback of a burst day; the lock has nothing left to protect. Being in before the gap is what pays; how cleverly you leave barely matters.

No — NSE's Closing Auction Session. Since at least October 1, continuous trading in F&O stocks ends at 15:15; the index freezes until the auction result prints at 15:29, and equity derivatives trade until 15:40. A stop crossed by the auction close is a legitimate stop at the official closing level, and because option prices reprice after the auction, the option sold at 15:29 fetched more than its post-auction value. Over October's fifteen auction sessions the frozen window changed no exits in our engine; the same check for an intraday gap-and-go system showed its gap reference (the 15:15 close) now differs from the official close by 20–110 points, changing the day's classification on three of fifteen days. That one goes to the next review with more data.