How to read a week of SPY options flow

What actually changes between Monday and Friday

A week of SPY options flow is not five comparable readings. Some fields are recomputed from zero at every open, some accumulate, and one of them, open interest, is only correct after the market has closed and OCC has processed the session. Confusing the three is the single most common cause of a weekly read that sounds coherent and is not.

Before comparing any two days, sort the fields by cadence. The comparison only means something when both sides of it are measured on the same clock.

The overnight open-interest settle rewrites yesterday

OCC publishes open interest once a day, after the close, once opening and closing activity has been reported and exercise and assignment have been processed. Nothing intraday, from any vendor, is a real open-interest figure. That single cadence fact is what makes a week-scale read structurally different from a day-scale one.

The practical consequence is that Tuesday's flow cannot be fully interpreted until Wednesday morning. A strike that absorbed heavy Tuesday volume can show a large open-interest increase, a small one, or a decrease, and each of those turns the same tape into a different story. A week gives you four overnight settles to check four days of assumptions, which is exactly the advantage a single session does not have.

• Large volume, large open-interest increase: consistent with new positioning at that series, though it does not say by whom or in which direction.
• Large volume, roughly unchanged open interest: consistent with transfers between participants, one side opening while the other closed.
• Large volume, open-interest decrease: consistent with closing activity, including rolls out of that series into another expiry.
• Any of the above with a matching change at a second strike or expiry: consistent with a spread or roll rather than an outright position.

The 0DTE mix is not constant across the week

SPY lists an expiration on every trading weekday, so every session contains same-day contracts. What changes is how much of the week's remaining structure sits alongside them. Monday's tape carries four more SPY expirations ahead of it inside the week; Friday's carries none, and it also carries the weekly and, at month end, monthly series expiring into the same close.

That mix changes what a raw call-versus-put number means without anyone changing their opinion about the market. A Friday reading blends same-day activity with the closing, rolling and exercising of everything that expires that afternoon. Comparing Friday's mix directly with Monday's compares two different calendars, not two different moods.

• Separate same-day expiry activity from the rest of the chain before comparing any two sessions.
• Expect closing and rolling activity to be heaviest into the sessions where large series expire.
• Treat a shift in the 0DTE proportion as a calendar effect first and a behaviour change only if it persists after the calendar is controlled for.
• Read short-dated gamma exposure as a model output that changes with spot through the day, not as a level the week is working towards.

A five-session read in the order the evidence arrives

The reading order matters more than the number of screens. Start with the field that is already settled, then move to fields that are still provisional, so the least reliable evidence never gets to frame the most reliable.

• Open with last night's settled open interest by strike and expiry, since it is the only confirmed positioning number you have.
• Add the week's cumulative premium and trade count by day to see whether activity is concentrated in one session or spread across five.
• Overlay the underlying price path, so a build in flow can be checked against whether price actually moved with it.
• Mark the strikes that appear in more than two sessions; single-session appearances are noise more often than they are structure.
• Check for dated catalysts inside the week, since scheduled events explain a large share of what otherwise looks like conviction.
• Only then write anything down, and write it as observations with their unresolved alternatives attached.

How to write a weekly note that contains no prediction

A defensible weekly note records what was observed, what the observation is consistent with, and what would change the reading. It does not contain a view about next week, because nothing in the data supports one. Flow describes activity that has already happened, side is an estimate, and no field in the dataset is forward-looking.

Written this way, the note stays useful when you return to it. A note that says put open interest at a given strike rose across three sessions while price held above it remains a factual record. A note that says the market is positioning for a decline is an interpretation that will be either embarrassing or accidentally flattering, and neither outcome teaches you anything.

Where a weekly read goes wrong

The most common failure is treating the week as a narrative with a beginning and an end. A week is an arbitrary window; positions routinely span expiries far outside it, and a large Thursday print may be the second leg of something that started three weeks earlier in a series you never looked at.

The second failure is asymmetric attention. It is easy to notice the sessions that agreed with an early impression and to skim the ones that did not. Reviewing the week in fixed order, with the settled data first, is a cheap structural defence against that.

• The week boundary is arbitrary; positions do not respect it.
• Estimated side is an inference from quote location and stays an inference no matter how many days you aggregate.
• Repeat activity at one strike may be one participant, several, or a market maker managing inventory.
• Realtime Options does not backtest, so a weekly note is a record to reread, not a strategy you can test in the product.

Questions people ask about this