Options sweep

How a sweep is detected

US options trade on many exchanges. An order that wants more contracts than the top-of-book venue offers can be routed to several venues at once, producing a series of prints on the same contract, same side, within milliseconds. A flow tool groups those prints and shows them as one sweep. Three parameters define the grouping: the time window between consecutive prints, whether all prints must share contract and side, and whether the feed covers every venue where the pieces printed.

A sweep is therefore an inference about execution, not a field in the raw data. Change the window and the same tape yields a different number of sweeps.

What a sweep suggests, and what it does not

• It suggests urgency: the sender took the available liquidity rather than resting at one price.
• It does not identify the sender, whether the position is opening or closing, or whether other legs were executed elsewhere.
• A large single print is not a sweep; it is one execution. A negotiated block is not a sweep either, even when it is bigger.
• A volume-to-open-interest ratio cannot verify that a sweep occurred; only the print sequence can.

Sweeps and the underlying's order flow

A sweep in options and an imbalance in the futures footprint are separate observations from separate markets. If a session record pairs them, it should note the option event time, the converted futures level and the subsequent footprint response, and keep the cases where nothing followed. That is how the pairing can be tested rather than remembered selectively.

In Senzoukria

The Option Flow module groups prints into sweeps with an explicit default rule: at least three executions on the same contract and the same side, each arriving within 800 ms of the previous one and printed on at least two distinct exchanges, form one group and are tagged SWEEP in the table. When the group's cumulative premium reaches $500K the tag becomes GOLDEN. A separate rule tags MULTI when different contracts of the same underlying print with the same size within a 25 ms window, which is the signature of a spread's legs. These thresholds describe grouping, not intent, and the feed's delay applies to every tag: 15 minutes on the Alpaca free tier, 5 to 15 minutes behind on Databento historical OPRA, as the module header states.

Common mistakes

• Counting sweeps from a feed that covers only some venues and treating the count as complete.
• Reading a sweep of calls as bullish without checking whether it was bought or sold, and whether a put leg printed at the same time.
• Comparing sweep counts between two tools that use different time windows.
• Treating the premium tag threshold as a statement about institutional identity.