Quant Memo
Core

Random End Times And Auction Extensions

Some auction designs deliberately end the call window at a randomized moment, rather than a fixed clock time, specifically so no one can profitably game the very last instant before lock — and separately, exchanges can extend an auction beyond its scheduled end if the order imbalance is too large to clear safely.

Prerequisites: MOC Cut-Offs And Late Order Rules, Gaming The Auction Imbalance

If everyone knew an auction's call window ended at exactly 10:00:00.000, a trader could time an order to arrive at 09:59:59.999 — late enough that no one else can react, early enough to still count. Several auction designs, especially European periodic auctions and dark-pool-adjacent venues, close that loophole by making the end of the call window itself unpredictable.

Randomized end times

Instead of a fixed close, the auction ends at a random instant drawn from a short window — for example, uniformly at random somewhere in the last 2 seconds of a nominal call period. No participant, no matter how fast, can know in advance whether their order at any given millisecond is "the last one in" or arrives after the lock, because the lock time itself isn't fixed until it happens. This removes the incentive to race for the literal final tick, since racing to a moment you can't identify in advance provides no edge — the best strategy becomes submitting a genuinely representative order early, not gaming the clock.

Auction extensions

Separately, if the order imbalance at the scheduled end of a call window is unusually large, several exchanges automatically extend the auction by a short, fixed period (commonly a small number of minutes) rather than locking in a price at a moment when the book clearly hasn't finished adjusting. The extension gives liquidity providers more time to see the imbalance and respond with offsetting orders, which — per the same logic as imbalance-to-price-impact — reduces how far the eventual print has to travel from the pre-auction price.

Worked example: comparing the two mechanisms

MechanismTriggerEffect
Randomized end timeEvery auction, alwaysEnd instant drawn randomly from a short window (e.g., last 1-3 seconds); prevents last-tick order timing games
Auction extensionImbalance exceeds a threshold at scheduled close (e.g., NYSE's imbalance-triggered extensions)Call window extends by minutes, giving offsetting liquidity time to arrive before the price locks

A stock with a normal, modest imbalance at its scheduled close experiences only the randomized-end mechanism — the auction locks within the usual short random window, unaffected by the extension rule. A stock with a large end-of-window imbalance (say, well above its typical closing imbalance) triggers the extension instead, and the auction runs several extra minutes while new orders adjust the balance, before a randomized-end lock is applied to that extended window.

random lock window scheduled end extension if imbalance too large extended lock
A randomized window governs the exact lock instant on every auction; a separate, larger extension only triggers when the imbalance at the scheduled close is too big to clear safely.

What this means in practice

Both mechanisms are anti-gaming design choices that trade a small amount of timing certainty for a more robust, harder-to-manipulate print. For anyone building execution algorithms around an auction, it means the exact lock time can't be treated as fixed and known in advance — order entry needs to account for a randomized window and the possibility of an extension, not a single deterministic clock time.

Randomized end times prevent gaming the literal last tick before an auction locks, while imbalance-triggered extensions give the book extra time to clear when the imbalance at the scheduled close is unusually large — two separate anti-gaming mechanisms addressing two different problems.

Related concepts

Practice in interviews

Further reading

  • European periodic auction and dark pool operator rulebooks (Cboe, Aquis, Turquoise)
ShareTwitterLinkedIn