Reading the options tape: sweeps, blocks, and the four things a print never tells you
Every options print carries a strike, a size, a price and a timestamp. It does not carry intent. Here is how sweeps, blocks and splits are classified, how aggressor side is inferred, and the four pieces of missing information that make most unusual-activity alerts unreadable.
The options tape is a stream of executed trades. Each one carries a contract, a size, a price and a timestamp. That is it.
Everything else you see attached to a print (bullish, bearish, opening, institutional, smart money) is inferred. Some of those inferences are sound. Some are guesses presented with unearned confidence. Knowing which is which is most of the skill in reading flow.
Classifying the print
Raw prints get grouped before they mean anything. A single large order arrives at the tape as many executions, and the grouping is what turns noise into a readable event.
Sweep. One order filled across multiple exchanges within a few milliseconds. This is the most informative classification, because of what it implies about the trader's priority: they crossed several venues simultaneously and accepted worse prices on the later fills rather than wait. A sweep is a statement about urgency. Someone wanted the position now, at a cost.
Block. A single large execution, typically negotiated away from the continuous market and printed as one piece. Blocks imply size and institutional involvement, but not urgency. A block is often the calm opposite of a sweep: worked carefully rather than grabbed.
Split. One order broken into pieces on a single venue, usually to reduce market impact. This signals a desire not to be noticed.
Multi-leg. Executions across two or more contracts in the same instant, which almost always means a spread rather than a directional bet.
The grouping windows have to adapt to the tape. During heavy trading, hundreds of unrelated prints land in the same 500 milliseconds and a loose window glues strangers together. During quiet hours the opposite risk applies. Density-aware windows (tighter when the tape is fast, wider when it is slow, with a higher venue-count requirement in dense conditions) are how you avoid manufacturing sweeps out of coincidence.
Inferring the aggressor
The tape does not say whether a trade was buyer-initiated or seller-initiated. That has to be inferred, and the standard approach is the Lee-Ready algorithm, published in 1991 and still the reference method.
The logic is simple:
1. Quote rule. If the trade printed above the midpoint of the bid-ask spread, classify it as buyer-initiated. Below the midpoint, seller-initiated. The premise is that an aggressive buyer lifts the offer while an aggressive seller hits the bid. 2. Tick rule. If the trade printed exactly at the midpoint, the quote rule cannot decide. Fall back to comparing against the previous trade price: higher is a buy, lower is a sell.
This is a good algorithm with a specific failure mode worth understanding: it depends on the quote being contemporaneous with the trade. If the quote feed lags the trade feed under load, a book that has not caught up to a rising market places every print above a stale offer, and the classifier reports a wall of buying that is really a wall of latency.
The defensive check is a sanity condition rather than a threshold. A contemporaneous book brackets its own trade. If a print sits many spreads outside the quoted bid and ask, that is evidence about the quote, not about the buyer, and the classifier should distrust it and fall back to the tick rule. We learned this the direct way: a stale-quote period produced a 97% buy reading that dissolved once the check was in place.
The four things a print cannot tell you
Here is where most flow interpretation goes wrong. These limits are structural. No data vendor solves them, because the information is not in the tape.
1. Opening or closing. A 5,000-contract call trade might be a new bullish position, or someone closing a short call, or a long holder taking profits. The tape shows the transaction, not the intent. Open interest updates the next morning and settles it after the fact, which is too late to have informed the alert you saw.
2. Which side the customer was on. A print shows a trade occurred. Inferring the aggressor from quote position is an inference with a known error rate, not an observation.
3. Whether it is one leg of something larger. This is the biggest one. A huge call purchase can be the long leg of a vertical spread, the buy-to-close of a covered call, a hedge against a short stock position, or the call side of a conversion. Each has a completely different meaning, and several are not bullish at all. Multi-leg detection catches the cases that print together, and misses the ones legged in over minutes or executed in different venues.
4. Whether it is directional at all. Volatility trades, dispersion trades and hedges all generate large, aggressive-looking options prints with no view on direction whatsoever.
What "unusual options activity" actually measures
Unusual activity screens flag prints that are large relative to some baseline: volume against average volume, volume against open interest, premium above a threshold.
That is a measure of notability, not of information. The screen is answering "is this print big?" and the label implies it answered "does someone know something?"
Two adjustments make the read more honest:
Measure premium, not contracts. Ten thousand contracts of a two-cent lottery ticket is $20,000 of conviction. Two hundred contracts of a deep in-the-money call can be millions. Contract counts systematically overweight cheap out-of-the-money options, which is exactly the part of the chain where noise lives.
Compare against the name's own baseline. Volume that is remarkable for a mid-cap is a quiet minute in a mega-cap. An absolute threshold produces a screen dominated by whatever is liquid.
How to read a print properly
Given all of that, a defensible reading of a single print sounds like this:
A sweep across four venues, $1.4M premium, at the offer, in a strike with
existing open interest well above today's volume, 12 days to expiry.
And the honest interpretation:
Someone paid up for urgency in size. Whether this opens or closes a position
is unknown until tomorrow. Whether it is one leg of a spread is unknown. The
aggressor read is an inference from quote position.
That is less satisfying than "smart money is loading up on calls." It is also the only version supported by what is in the data.
The transferable rule
Every label on a print is either in the trade record or it is an inference.
Know which, for every field you are looking at.
Strike, size, price, timestamp and venue are recorded. Sweep versus block is a sound inference from grouping. Aggressor side is an inference with a known failure mode. Opening versus closing, hedged versus naked, and single versus multi-leg are frequently not knowable at all.
Flow data is genuinely useful. It is useful for what it records, and treating the inferences as records is how a good dataset produces bad conclusions.
Related: what dealer gamma is and how to read gamma levels.