Ten lots or five, no month to skip, no festival to fear, and the one condition that loses

The edge per lot per trade was ₹633 in 2015–22 and ₹140 in 2023–26, a step rather than a drift (p 0.019), so sizing from the long-run average would size from a market that may no longer exist. Every debater and the judge landed on fixed five lots with an anniversary rule: step up to ten only after twelve months in which no top-up was needed and measured costs stayed at or under 0.3% per side. Ramps gated on recent profit were tested and whipsaw — they add lots at the top and cut them at the bottom.

Rejected ten to nothing, 0 of 36 cells. After a top-decile month the next month averaged +₹86k against +₹48k otherwise — momentum, not mean reversion, because the size of gaps is autocorrelated (0.47) and the book earns on gaps. Month cards, a fixed allocation by calendar month, were rejected ten to nothing as well: movement explains the months (wild-gap months +₹92k against +₹16k for calm ones), the pattern breaks in 2023–26, and drawdown is independent of the month.

Neither. Post-holiday days look good only because holidays pile up news: 44% of them gap 0.5% or more against 25% normally, and once the gap is controlled for the holiday effect is ₹2.9k with p 0.27. Budget day lost 5 of 6 trades; exit-poll and result days earned ₹33.9k and ₹27.6k, but on two trades each. Thirteen event tests, none survives correction, no live change.

On the profit side, no. A pre-registered bar — at least 25 traded days, win rate at or above 80% or at or below 20%, confidence interval clear of 60/40, both halves of history agreeing, surviving an extra 0.2% per side of cost and a label-shuffle null — was applied to roughly 330,000 pre-entry conditions, including 39 pieces of desk folklore about gap fills, expiry day, Mondays and Fridays, budget, RBI, VIX and the first 30 minutes. No single feature moves the win rate outside 34–62% from a base of 45%. The chain's money is bigger wins on big-gap days — gaps of 0.75% or more pay about three times the average per trade — not more frequent wins, and gaps under 0.25% have been break-even since 2023.

One condition, found independently by three reviewers and confirmed by hand: the door on a 0.2–0.5% gap when the prior close is at least 2.5% above its 20-day average. 67 trades since 2015, 13% winners, about −₹7.2k per trade at five lots; 27% winners in 2015–20 and 1 of 31 since 2021, and out of sample on 2023–26, with the rule frozen on data to 2022, 25 trades at 20%. It is not fragile to the threshold (2.0–3.5% gives 13–26%), and the mirror image — stretched below — is normal at 47–52%, so the rule stays one-sided. Skipping it would have taken the chain from ₹33.4 lakh to ₹38.3 lakh and the drawdown from ₹2.42 lakh to ₹2.02 lakh. Because a search this large produces about 50 such "passes" by chance, it is a candidate skip rule to be paper-logged, not adopted today.

Reversing — buying the opposite option at the same minute with the same stop and trail — turns 61 trades at 15% winners and −₹4.41 lakh into 49% winners and +₹4.83 lakh. But almost all of that came in 2019–22; in 2023–26 the reverse is roughly break-even (+₹24k on 16 trades), so nearly all the benefit comes from not taking the losing trade. A reverse is a second data-mined bet stacked on the first; if anything is adopted it is the skip.

Three tests, ten reviewers, and the stock data earns its keep only as a caution. As a filter on the index chain, breadth, heavyweight confirmation and sector models all failed — heavyweight "confirmation" would have cut the door's best days, which are the ones where the heavyweights gap against the index, and the one breadth result that looked good was an artefact of the stock universe growing from 48 to 117 names in September 2024. Trading the stocks themselves, gap-and-go lost ₹715–787 per trade net and gap fades about ₹1,000–1,200; stocks do give gaps back, but before any signal fires, and only 0–10 basis points remain against a 0.9% option round trip. A low sector-dispersion condition at entry reproduces in both halves of history and goes on the paper log. Stock gap trading is dropped; the cheap, liquid index chain stays the edge.