On the New York Stock Exchange, a market-on-close order locks ten minutes before four o'clock, and you cannot cancel it after that. It is a hard wall. Miss it, and your order fills at whatever price the auction produces.
Most people picture the close as a single event. It is really a sequence of deadlines, each one shutting a different door, and the sequence runs across the final ten minutes of the session. All of it is timed against the bell that ends the regular session, and the exchange builds every cutoff backward from that one point.
Knowing the order of those doors is the difference between choosing your exit and accepting one.
There is one narrow exception. Past the deadline for on-close orders, the exchange still accepts new market-on-close orders, but only on the side that offsets a published imbalance. If the auction is showing far more sellers than buyers, you can add a buy. You cannot add another sell.
That asymmetry is deliberate. The rule lets late liquidity fix a lopsided auction without letting anyone make it worse in the closing minutes.
Closing offset orders run on the same logic and push it further. They are limit orders that execute only against the opposite side of an imbalance, they never add to one, and they yield to every other kind of competing interest. Quiet, patient orders. They fill when the auction needs them and sit still when it does not.
Those extra ten minutes of optionality matter more than they sound, because D Orders carry the auction. In September 2025 they made up roughly 60 percent of NYSE closing auction volume. Market-on-close and limit-on-close orders were about 20 percent each.
And the brokers push later every year. More than 60 percent of D Orders now arrive in the last two and a half minutes.
The shift is sharpest at the end. Between August 2024 and September 2025, the share submitted in the final ten seconds of trading climbed from 4.45 percent to 13.4 percent. Over the same stretch, floor broker use of third-party order management systems went from 69.3 percent to complete adoption.
Better software, later orders. The technology removed the reason to submit early, so brokers stopped.
Read it carefully, though. An imbalance printed at ten minutes to four reflects a book that has not yet received most of its D Order volume, so the early picture is often nothing like the four o'clock picture. The number you act on can invert twice before the auction runs.
At four o'clock the exchange runs the match. Every eligible order is paired at one price, and that price becomes the official closing print for the stock, which is the figure index funds and pricing services take for the day.
There is a second problem with acting late. NYSE's own analysis found that order marketability declined through the final ten minutes as submissions clustered nearer the bell, meaning the orders arriving last were priced further from where the auction was likely to clear.
That is the trade every closing auction participant makes. Wait, and you see more of the imbalance data before you commit. Wait too long, and you are chasing fills against a crowd doing the same thing, at prices nobody has much incentive to improve.
None of this transfers cleanly to Nasdaq. Its version of the auction is called the closing cross; it runs on a separate rulebook, and it sets its own entry and cancellation cutoffs for on-close orders. There is no floor broker equivalent of the D Order there either, so the late-arrival pattern that dominates the NYSE close does not shape the Nasdaq one in the same way. If you trade listings on both exchanges, you need two sets of deadlines in your head, and the one that matters is the one attached to the venue where the stock is listed.
Most people picture the close as a single event. It is really a sequence of deadlines, each one shutting a different door, and the sequence runs across the final ten minutes of the session. All of it is timed against the bell that ends the regular session, and the exchange builds every cutoff backward from that one point.
Knowing the order of those doors is the difference between choosing your exit and accepting one.
On-close orders stop being yours at ten to four
Market-on-close and limit-on-close orders can be entered, changed, or pulled right up to ten minutes before four. After that, the exchange freezes them. You cannot move the price, you cannot cut the size, and you cannot walk away.There is one narrow exception. Past the deadline for on-close orders, the exchange still accepts new market-on-close orders, but only on the side that offsets a published imbalance. If the auction is showing far more sellers than buyers, you can add a buy. You cannot add another sell.
That asymmetry is deliberate. The rule lets late liquidity fix a lopsided auction without letting anyone make it worse in the closing minutes.
Closing offset orders run on the same logic and push it further. They are limit orders that execute only against the opposite side of an imbalance, they never add to one, and they yield to every other kind of competing interest. Quiet, patient orders. They fill when the auction needs them and sit still when it does not.
Floor brokers now wait until the final ten seconds
The rest of the auction runs on a different clock. Closing D Orders, which only NYSE floor brokers can submit, stay open until ten seconds before four, and they carry discretion to trade at an undisplayed price inside their limit.Those extra ten minutes of optionality matter more than they sound, because D Orders carry the auction. In September 2025 they made up roughly 60 percent of NYSE closing auction volume. Market-on-close and limit-on-close orders were about 20 percent each.
And the brokers push later every year. More than 60 percent of D Orders now arrive in the last two and a half minutes.
The shift is sharpest at the end. Between August 2024 and September 2025, the share submitted in the final ten seconds of trading climbed from 4.45 percent to 13.4 percent. Over the same stretch, floor broker use of third-party order management systems went from 69.3 percent to complete adoption.
Better software, later orders. The technology removed the reason to submit early, so brokers stopped.
The imbalance feed shows you who needs to trade
Ten minutes before four, the exchange starts publishing what it knows. The significant imbalance feed goes out, and from that point the informational imbalance updates every second whenever the numbers change. It gives you the size of the unmatched interest and tells you which side is short of a counterparty.Read it carefully, though. An imbalance printed at ten minutes to four reflects a book that has not yet received most of its D Order volume, so the early picture is often nothing like the four o'clock picture. The number you act on can invert twice before the auction runs.
At four o'clock the exchange runs the match. Every eligible order is paired at one price, and that price becomes the official closing print for the stock, which is the figure index funds and pricing services take for the day.
There is a second problem with acting late. NYSE's own analysis found that order marketability declined through the final ten minutes as submissions clustered nearer the bell, meaning the orders arriving last were priced further from where the auction was likely to clear.
That is the trade every closing auction participant makes. Wait, and you see more of the imbalance data before you commit. Wait too long, and you are chasing fills against a crowd doing the same thing, at prices nobody has much incentive to improve.
None of this transfers cleanly to Nasdaq. Its version of the auction is called the closing cross; it runs on a separate rulebook, and it sets its own entry and cancellation cutoffs for on-close orders. There is no floor broker equivalent of the D Order there either, so the late-arrival pattern that dominates the NYSE close does not shape the Nasdaq one in the same way. If you trade listings on both exchanges, you need two sets of deadlines in your head, and the one that matters is the one attached to the venue where the stock is listed.
