Gamma blast on expiry day: what really happens to ATM options in the last 30 minutes

Every expiry week the same clips circulate: an at-the-money option going from ₹20 to ₹150 in the last ten minutes, captioned gamma blast. The clips are real. What nobody publishes is the base rate underneath them. We hold every NSE index option chain at ten-second resolution through the expiry endgame, so this guide shows what the last 30 minutes actually do to the at-the-money option, how the settlement price that drives it is built, and what our terrain page had on screen on 8 September 2026 while a NIFTY put went from ₹27 to ₹95 in 52 seconds and back to ₹15 seven minutes later.

Why the last 30 minutes are different from the rest of the day

Two things change at 15:00 on an expiry day. First, the option is about to become a fixed number: at settlement a call is worth the index close minus the strike or zero, a put the reverse. Every rupee of time value has to leave the premium by 15:30, so an at-the-money option that still carries ₹30 of extrinsic value at 15:00 must lose all of it, whatever the index does. Second, the settlement price is not the last tick. NSE computes the index close as the weighted average of the index over the final 30 minutes, which is why we call this window the settle walk: from 15:00 the market is no longer trading a price, it is trading an estimate of an average that is still being formed.

Those two facts explain the shape of the whole window. Gamma, the rate at which an option's delta changes with the index, is at its lifetime maximum for the at-the-money strike in the last half hour. A 20-point NIFTY move that would shift an ATM premium by ₹8 on a Monday shifts it by ₹15 or more at 15:15 on expiry, and by the same amount in the other direction a minute later. Writers who are short that strike see their hedge requirement swing with every tick. When the strike is slightly in the money and open interest is large, some of them stop hedging and start closing, and closing a short option means buying it back at whatever the market asks. That is the blast.

Pin risk: the setup that produces it

The dangerous configuration is specific. The strike with the largest open interest on one side sits within a few points of the settle walk, on the wrong side for its writers, with under fifteen minutes left. On 8 September 2026 that strike was the NIFTY 23650 put: 22 million contracts of open interest, the largest put wall on the chain, with the index walking through 23640, ten points in the money. A writer short that put faced assignment at settlement unless the index recovered above 23650 or the position was closed. The first closing wave began quietly at 15:05, the strike losing about 400,000 contracts a minute while its price stayed near ₹30. Between 15:11 and 15:16 the pace tripled: 6.9 million contracts closed, the premium rose from ₹28 to ₹53, then a dump to ₹12 as the buying paused and fresh writers sold into the dip. At 15:20:03 the second wave arrived. In 52 seconds ₹152 crore of that put was lifted at the ask, open interest fell from 22.0 million to 13.9 million, and the premium printed ₹95. The index did not move. NIFTY futures traded 23744 at the start of that minute and 23741 at the end. By 15:27 the put was back at ₹13, which was roughly its intrinsic value against the settlement, and everyone who had bought the ₹95 print had lost 85 percent in seven minutes.

That is the whole mechanism in one strike: not a directional move in the index, but a forced exit by one crowd, priced against the other crowd's estimate of a settlement that had not yet been computed.

What 21 expiries say about the odds

We measured every NIFTY Tuesday and SENSEX Thursday expiry in our tick data from 30 June to 8 September 2026. For each index-day we took the strike nearest the index at 15:00, both the call and the put, and followed their premiums at ten-second resolution to 15:30. The table is what the at-the-money option did relative to its 15:00 price.

Read the four numbers together and the shape is clear. Big multiples exist and they are not rare: a third of legs doubled at some point, a third of days produced a three-bagger. But the median leg went nowhere good. Only 31 percent ended the half hour above where they started it, and the losses are not gentle, they are near total, because the losing leg's destination is zero. The clips show the seven days. The other fourteen do not make clips.

Two details from the per-day record matter for anyone reading a screen in that window. First, the doubling and the halving happened to the same leg within the same half hour only once in 42 cases, the 8 September put, but the survivors often doubled and then gave most of it back before 15:30, because the settle walk keeps moving after the spike. Second, the large winners were nearly all puts on days the index was drifting lower into the close, and calls on days it was drifting higher. The blast did not fight the walk; it rode it.

What Gamma Terrain shows while it happens

Our terrain page rebuilds the option structure every 30 seconds from the chain and reads the tick feed for every print at the strikes around the index. In the settle walk five readings on it carry the story, and on 8 September they carried it in this order.

The pin and the walls. From 14:00 the pin, the strike whose gamma-weighted open interest holds price hardest, sat at 23700, with the put wall at 23600 and the call wall at 23650 or 23700 depending on the minute. At 12:05 the pin had already moved from 23700 down to 23650. A pin that migrates toward the largest in-the-money wall late in the day is the terrain's version of pin risk.

GRIP. The chip beside the pin reads how the pin's hold is changing minute to minute. Through the afternoon it read holding and firming; it is not built to call a blast, only to say whether the anchor is weakening.

BOOK. The read that mattered most. It measures the resting depth at the two put strikes below the pin and the two call strikes above, as a percentage of their own baseline, and names the side that is thinning. From 15:07 it read calls above thinning. At 15:16:50 it flipped: puts below at 26 percent of normal, calls above at 112 percent, the put book emptying three and a half minutes before the second wave. By 15:20:50 it read put book 8 percent, call book 780 percent. A flip is not a direction call, across 45 sessions a flip moved with its side less than half the time, but in a settle walk a vanishing book on the in-the-money side is the vacuum forming.

The ledger. Every print above the size bar lands in the event feed with its side. At 15:15 the feed showed 35 buy prints in one minute on the 23650 put against 26 sells; at 15:20 it showed 56 buys, ₹152 crore, against 29 sells. Aggressor side on the option itself, not on the index, is what distinguishes covering from a market move.

The crater. At 15:20:56 the terrain flagged a crater exposed below, a band of strikes with almost no gamma-weighted open interest beneath the pin, the range price could fall through with little friction if the walk turned. It did not turn; the put collapsed instead. The crater told the truth about the structure, and the structure is not the outcome.

The honest note about our own detector belongs here too. The Gamma Blast state machine on the page, which watches for a coiled at-the-money premium and a trigger, sat coiled from 14:00 with a score of 2.8 against its gate of 3, and switched itself off at 15:20:00 by a design rule that ends its hunt at 15:20. It went dark sixty seconds before the move it was built for. We are extending it through the settle walk, in shadow first, and the result of that will be published here when it has enough expiries behind it.

What the tape does not tell you

None of this predicts which side blasts. The readings above describe a structure and a crowd under pressure; they say a vacuum is forming and which side it is on, and the data says that even with that knowledge, less than a third of at-the-money legs end the half hour higher. The premium you see at 15:20 on expiry is a settlement guess plus panic, and implied volatility in that window is a number back-solved from that panic, not a measure of anything. If a page tells you a gamma blast is a strategy, ask it for the 21-day table. TBTflow shows the readings, live, tick by tick, and leaves the decision where it belongs.