Since August 2026, SENSEX no longer simply stops trading at 3:30. It closes through a closing auction session. From around 3:15 PM the published index freezes while the auction runs, and a single final price prints at about 3:29 PM.
On expiry day, that one print decides where every SENSEX option settles. Not the last traded price. Not a weighted average of the final half hour. One number, produced by an auction.
Three weeks ago I published data showing that CAS appears to have flattened ordinary intraday movement — smaller ranges, weaker directional bodies, fewer trend days. This article is the other half of that story. If the auction drained the ordinary day, did it also concentrate movement into the last minutes of expiry? And if so, is there a way to trade it with genuinely limited risk?
I spent a month testing that. What follows is the complete work: the history, the two structures that failed, the one that survived, every trade, the robustness checks, and an honest account of where it breaks.
Why I threw away the ordinary-day statistics
My first cut at this measured how much SENSEX moves in the last half hour across all trading days. That was the wrong question.
This structure is only ever placed on expiry day, and only expiry day has the property that matters: the auction print is also the settlement price. An ordinary Tuesday tells you something about market behaviour in general. It tells you very little about the specific afternoon on which you would hold six option legs into settlement.
So I rebuilt the study on expiry days alone.
1. What expiry day actually does in the last 20 minutes
166 SENSEX expiry days, May 2023 to September 2026. The move is measured from 3:10 PM to the close, and rescaled to today’s ~75,000 level so that a move in 2023 and a move in 2026 can be compared fairly.
One methodological note worth stating plainly: the SENSEX expiry weekday changed twice over this period — Friday until December 2024, Tuesday through August 2025, and Thursday since September 2025. Because of that, expiry dates are taken from the data itself rather than from any weekday rule.
| Era | Expiries | Median | 90th pct | Max | In the 150–550 zone |
|---|---|---|---|---|---|
| Friday era, to Dec 2024 | 85 | 56 | 161 | 246 | 14% |
| Tuesday era, Jan–Aug 2025 | 35 | 63 | 150 | 216 | 11% |
| Thursday, pre-auction | 40 | 41 | 102 | 241 | 8% |
| Thursday, auction era | 6 | 246 | 324 | 378 | 83% |
| All 166 expiries | 166 | 55 | 169 | 378 | 14% |
Three things stand out.
Before the auction, expiry day was boring. A typical expiry drifted 40 to 60 points in its last 20 minutes. Only 8 to 14 percent of expiries travelled 150 points or more. Compared with an ordinary day in the same era, expiry barely stood out at all — a ratio of 1.1x to 1.3x.
Since the auction, expiry day is a different animal. The median tail move is 246 points, and five of the six have cleared 150. In this era expiry moves roughly 1.9x an ordinary day.
The record books were rewritten in six weeks. The three largest expiry-day tail moves in the entire 166-expiry history all occurred since the auction began: 3 September −378, 27 August −270, 10 September +267.
The single most important number in this study
In 166 expiries, the last 20 minutes has never moved more than 550 points. The maximum on record is 378.
That matters because 550 points is the upper edge of the profit zone of the structure described below. The failure mode where a violent late move blows straight past the protective wings has not occurred once on a SENSEX expiry in three and a half years.
Never happened is not the same as cannot happen. But it is the reason this structure is shaped the way it is.
The last twelve expiries, measured 3:10 PM to close: 25 Jun −68, 2 Jul −14, 9 Jul +182, 16 Jul +31, 23 Jul +48, 30 Jul −32, 6 Aug +168, 13 Aug +225, 20 Aug +50, 27 Aug −270, 3 Sep −378, 10 Sep +267.
Data note: the spot history contains a gap from 3 March to 5 May 2026, so roughly nine pre-auction expiries are absent from these counts.
2. Two structures that did not survive testing
Before arriving at something worth publishing, two obvious ideas had to be discarded. I am including them because the failures are the reason the third structure looks the way it does.
The long strangle
Buy a cheap out-of-the-money call and put somewhere between 3:05 and 3:15, then exit on a combined profit target or a trailing stop. Simple, capped risk, and it should profit from exactly the kind of late expansion the auction seems to produce.
I tested 14,400 variations of entry minute, strike distance, target and trailing rule. Not one of them was profitable across all six auction-era expiries.
The problem is 20 August. The index moved 52 points that day. With a move that small, there is no exit at any moment of the session that recovers the premium paid — even choosing the best possible exit with perfect hindsight loses money. Every rule that looked good across the other five expiries was, on inspection, simply a rule that had been fitted to them. It broke the moment it met a quiet day.
The naked 1:3 ratio spread
Buy one at-the-money option and sell three further out, sized so the premium collected roughly matches the premium paid, and do it on both the call and the put side. This was profitable on all six expiries in the backtest.
It also carries unlimited risk. A 1,000-point move would cost somewhere between ₹20,000 and ₹35,000 per structure, and there is no upper bound on what a genuine shock could cost. A strategy whose worst case is undefined is not a strategy — it is a loan from the market that gets called in eventually. Rejected.
3. The structure that survived: 1:3 with wings
The fix for the ratio spread is to buy back the tail. Every SENSEX expiry, at 3:10 PM:
- Call side: Buy 1 ATM CE · Sell 3 CE approximately 400 points out · Buy 2 CE approximately 600 points out
- Put side: Buy 1 ATM PE · Sell 3 PE approximately 400 points out · Buy 2 PE approximately 600 points out
The rules that govern strike selection:
- Choose the short strike so the wings sit at least 500 points away from the index
- Wing distance is half the distance between the long strike and the short strike
- Hold everything to expiry settlement — there is no exit rule and no stop to manage
- For live use, derive the at-the-money level from option prices (strike + CE − PE, averaged across the nearest strikes) rather than from the published index, which is frozen while the auction runs
Why the loss is capped
Beyond the wings, one long option plus two wings exactly offset the three short options. The payoff stops changing and goes flat. The maximum loss is therefore the premium paid at entry — approximately ₹2,700 to ₹3,500 per structure at one lot of 20 — plus a small amount of exercise STT on extreme moves, around ₹7 per 100 points.
The profit zone is a move of roughly 150 to 550 points in either direction, peaking near the short strikes. A very quiet day loses the capped premium. A move beyond 550 points also loses it — though as established above, in 166 expiries that has never once happened.
4. Every trade
All results are per structure, after brokerage, STT, exchange transaction fees, GST, stamp duty, and ₹0.5 of slippage per leg.
| Expiry | 3:10 level → auction print | Move | Risk | Result |
|---|---|---|---|---|
| 6 Aug | 78,778 → 78,955 | +177 | ₹3,064 | +₹27 |
| 13 Aug | 77,846 → 78,080 | +234 | ₹2,668 | +₹2,924 |
| 20 Aug | 77,486 → 77,538 | +52 | ₹2,795 | −₹2,042 |
| 27 Aug (monthly) | 77,210 → 76,934 | −276 | ₹3,076 | +₹2,245 |
| 3 Sep | 76,537 → 76,153 | −384 | ₹3,531 | +₹3,403 |
| 10 Sep | 74,636 → 74,903 | +267 | ₹3,273 | +₹2,771 |
The full leg detail for each expiry:
6 Aug — CE: 78800 @240.10 / 3× 79200 @81.55 / 2× 79400 @41.70 · PE: 78800 @214.40 / 3× 78400 @80.85 / 2× 78200 @44.25
13 Aug — CE: 77800 @100.75 / 3× 78200 @4.15 / 2× 78400 @1.60 · PE: 77800 @30.65 / 3× 77400 @1.10 / 2× 77200 @0.70
20 Aug — CE: 77500 @76.20 / 3× 77900 @3.25 / 2× 78100 @1.35 · PE: 77500 @63.95 / 3× 77100 @3.30 / 2× 76900 @1.70
27 Aug — CE: 77200 @114.10 / 3× 77600 @9.90 / 2× 77800 @4.60 · PE: 77200 @56.75 / 3× 76800 @4.55 / 2× 76600 @1.95
3 Sep — CE: 76500 @178.70 / 3× 76900 @29.70 / 2× 77100 @12.90 · PE: 76500 @72.45 / 3× 76100 @12.95 / 2× 75900 @7.10
10 Sep — CE: 74600 @124.80 / 3× 75000 @27.40 / 2× 75200 @13.65 · PE: 74600 @156.65 / 3× 74200 @37.20 / 2× 74000 @17.60
Worst day was −₹2,042. The worst possible outcome on any day is approximately −₹3,700. Individual results are rounded to the rupee, so the six entries sum to ₹9,328 against the unrounded total of ₹9,329. Deriving the at-the-money level from option prices rather than the index produces +₹9,019 instead.
The 6 August result is worth dwelling on. A +177 point move sits just inside the profit zone, and the structure returned ₹27 — essentially nothing. That is what the edge of the zone looks like in practice, and it is a useful reminder that “landed in the zone” and “made money” are not quite the same thing.
5. Robustness checks
A backtest that only works with one exact set of parameters is a curve fit. Three checks:
- Entry timing and ATM derivation. Entering anywhere between 3:08 and 3:12, with the at-the-money level taken either from the index or from option prices, gives ten distinct versions. All ten produced totals between +₹6,800 and +₹9,800.
- Slippage. At ₹0, ₹0.50, ₹1 and ₹2 per leg the totals are +₹10,048, +₹9,329, +₹8,609 and +₹7,227. The cost model is internally consistent: each ₹0.50 of slippage per leg costs exactly ₹720 across six expiries, which is 12 legs × 20 quantity × 6 expiries.
- Independent re-derivation. Every trade was recomputed from the raw data by a separate process that did not use my code. All 36 legs and every result matched to within ₹1.
6. This is a base. Build on it.
I am publishing this as a starting point, not a finished system.
If you trade SENSEX expiries, take it apart and improve it. Different wing distances. A different entry minute. A filter based on the expected move implied by option prices, or on how far the index has already travelled that day. Position sizing that varies with the premium. Skewing the structure when one side is obviously richer than the other. I would rather ten people refine this than one person copy it blindly.
Three things to be clear about before you do.
The sample is small
Six expiries shows a shape. It does not prove an edge. Any conclusion drawn from six observations deserves to be held loosely.
The entire bet is on explosive moves in the CAS window
This structure needs SENSEX to travel roughly 150 to 550 points after 3:10 PM. Across the 160 expiries before the auction, that happened just 12 percent of the time — meaning this structure would have lost its capped premium in roughly nine expiries out of ten. In the six auction-era expiries it landed five times.
With an average win near ₹2,300 against a capped loss of ₹2,500 to ₹3,000, the move has to reach the zone roughly half the time simply to break even. Six weeks of a new market structure is nowhere near enough to conclude that it will keep doing so. If auction-day moves settle back down toward their pre-auction behaviour, the assumption underneath this entire strategy disappears and it loses money most weeks. The losses stay capped — but they are still losses.
Execution is not free
Six option legs and twelve lots have to be filled in the minutes before the auction closes. The backtest uses 1-minute closing prices and assumes settlement at the 3:29 print. Real fills will be worse than that, and the gap between a backtest fill and a live fill is precisely where many promising ideas quietly die.
What happens next
I am paper-trading this for the next 8 to 10 expiries before any real money goes near it. If you build something better, share it — I will run it against my data and publish what comes out.
I would rather share a small edge honestly than a big one I cannot defend.
Disclaimer. I am not registered with SEBI as a Research Analyst, an Investment Adviser or a Portfolio Manager. This article is backtest research shared for educational purposes only. It is not investment advice, not a recommendation, and not a call to trade. Options can lose money quickly, and a backtest is not a promise of future results. Please consult your own SEBI-registered financial adviser and test everything yourself before risking any capital.