The Auction Ledger · preregistration

The pre-open alignment, measured from day one.

From Monday 7 September 2026 NSE aligns its equity pre-open session with the closing auction: a limit-only phase with a random close replaces the last minutes of open order entry. Nobody has history for the new mechanism, so this page fixes the measurement, the control and the bars before the first aligned session exists. The number publishes here either way. Written 5 September 2026; baseline measured 6 September; the git timestamp on this file is the proof the bars preceded the result.

baseline sessions, both exchanges

33.6 bp

NSE median open dislocation, baseline

41.9 bp

BSE, the control

40 + 40

sessions each side before the bars are read

What changes, in NSE's words

phase	before	from 7 September
order entry, all types	09:00 to 09:08	09:00 to 09:05
order entry, limit orders only	none	09:05 to 09:10, random close in the last 2 minutes
matching	09:08 to 09:12	09:10 to 09:12
transition to continuous	to 09:15	09:12 to 09:15

The load-bearing change is the limit-only phase. Market orders entered in the closing seconds were the cheapest way to push a pre-open print, and from Monday they cannot be entered at all in the final five minutes. Execution priority is unchanged: market against market by time, then residual market against limit, then limit against limit, by price and time. NSE also extended the closing-auction reference-price and price-band rules to index futures from the same date (circular NSE/CMTR/76170, 3 September 2026), a separate change on the same day that is not part of this measurement. The alignment itself is circular NSE/CMTR/74969 of 1 July 2026, with a mock session on 5 September under NSE/CMTR/76187.

The claim being tested

If the limit-only phase makes the pre-open print harder to push, the opening price should be a better estimate of where the stock actually trades, and the market's correction of it in the first minutes should get smaller.

The measurement, defined now

open_dislocation = (first continuous traded price at or after 09:15 / official open) - 1,
                   in basis points

official open: the raw NSE CM bhavcopy open, which is the pre-open equilibrium price. Raw, not the adjusted panel, for the same reason the closing-auction series uses raw: adjusted opens against raw intraday prices measure the corporate action, not the auction.
first continuous price: the close of the first minute bar at or after 09:15 with volume above zero. A stale bar is not a price.
Both legs are computable for any historical session, which is the whole point: the pre-period goes back as far as the minute bars do, and the measurement is identical on both sides of 7 September. The change is the treatment; the ruler does not move.

Secondary, same definitions: the 09:15 to 09:30 return conditional on the sign and size of the open dislocation. A print that was pushed should revert; a print that was fair should not.

Universe: the F&O-eligible list, the same names the closing-auction series measures.

The control

BSE, same symbols, same days, same two-leg definition. BSE runs its own pre-open and is not covered by this NSE circular, so a change that shows up on both exchanges is not this circular's doing. This is the same cross-exchange control the closing-auction series already uses, for the same reason: without it, any market-wide change in September reads as a treatment effect.

The bars, fixed before the data exists

Comparison windows: the 40 sessions before 7 September against the first 40 sessions from 7 September, medians per session, difference in differences against BSE.

The headline claim publishes only if the median absolute open dislocation on NSE falls by at least 10% relative to its pre-period median AND the same statistic on BSE does not fall by as much, with the NSE-minus-BSE difference outside a bootstrap 95% interval over sessions.
The reversal claim publishes only if the 09:15 to 09:30 reversal coefficient on the open dislocation shrinks toward zero on NSE and not on BSE, again outside the interval.
If neither clears, that is the published result. "The alignment did not measurably change the honesty of the opening print in its first 40 sessions" is a finding, and it goes on the auction ledger beside the closing-auction numbers with the same prominence.
No trading claim is made from this either way. This measures a price formation mechanism. A tradable claim would need its own preregistration, its own costs and its own gate.

What could make this uninformative, said now

40 sessions is 40 observations per side on a session-median statistic. If the effect is small, this cannot see it, and "no measurable change" then means the study lacked power rather than the change lacked effect. Power will be stated with the result.
September is one month. A regime that moves opens for unrelated reasons contaminates both arms, which is what the BSE control is there to absorb, imperfectly.
The index-futures price-band change lands the same day. It does not touch this measurement's two legs, but it is a same-day confound for anything wider, and nothing wider will be claimed.

Baseline, measured 6 September 2026, one trading day before the change

69 sessions, 15 June to 4 September 2026, the F&O universe. The bars above are unchanged by this; it is the pre-period the first 40 aligned sessions will be compared against.

exchange	symbol-days	sessions	symbols	median of session medians, abs open dislocation	session range	median abs 09:15 to 09:30 return
NSE	14,487	69	210	33.6 bp	24.5 to 65.4 bp	35.6 bp
BSE	14,348	69	208	41.9 bp	29.1 to 70.3 bp	36.6 bp

Paired NSE and BSE symbol-days: 14,348; the correlation of the two exchanges' dislocations across them is 0.81, which is what makes BSE usable as the control: the two opens move together on the days that move them, so a change on NSE alone is visible against it.

Clarification, 6 September, before any aligned session; the bars are unchanged

An audit of the baseline tried the first traded bar's open as the first-continuous-print leg and found it equal to the bhavcopy open on 98% of symbol-days on both exchanges: the official open is, by construction, the first print of continuous trading, so a leg built on it is identically zero and measures nothing. The preregistered leg, the close of the first traded minute bar at or after 09:15, is therefore what the measurement is, and its honest description is: the price at the end of the first continuous minute against the pre-open equilibrium price. It carries that minute's drift by design; that drift is the market's first correction of the auction print, which is the object of the study.

The capture that cannot be repeated

The order-book evolution inside 09:00 to 09:12 exists only while it happens. A tick recorder captures 08:58 to 09:20 every session from 5 September, which covers the whole new pre-open including the random close. That capture is not required by the bars above, which run on bhavcopy and minute bars; it exists because Monday's opening book is a one-time object and the same reasoning that justified the closing-auction capture applies here.

Publication: this page and the auction ledger, after 40 aligned sessions, around early November 2026, and not before. The series is built nightly by the same pipeline as the closing-auction record. Nothing here is investment advice; this measures a market mechanism, not a strategy. Terms · Privacy