How to Read Time and Sales
To read time and sales, filter for a minimum trade size and watch where trades are executing relative to the bid and ask. It records completed transactions rather than intentions, which makes it the only display showing what participants actually did.
Time and sales is the list of completed transactions: price, size and time, one line per trade. It is the only display that shows what participants actually did rather than what they said they might do.
Before you start
A filter for minimum trade size, because the unfiltered feed is unreadable. An active instrument produces more prints per second than anybody can process.
An understanding that the aggressor side is inferred, not reported. A trade at the ask is usually a buyer taking liquidity, and “usually” is doing real work in that sentence.
A decision about what you are watching for, since the feed does not pause. Large prints, activity at a specific level, or the rate of trading. One thing.
The steps
1. Set a size filter first
A floor that removes the majority of prints. Most small trades are automated and carry no information you can act on in the time available.
2. Watch where trades print relative to the quote
Trades at the ask suggest buyers crossing the spread; at the bid, sellers doing the same. Trades between the quotes are usually a different order type entirely.
3. Note the large prints
A single trade many times the typical size happened, and it cannot be withdrawn. That is a different category of information from a large order resting on the book.
4. Watch the rate, not just the prices
A sudden acceleration in prints per second is an event even before you know its direction. Quiet periods and bursts are the feed’s most legible signal.
5. Compare it against the book
A large bid on the book that never appears in the trade record was never hit. The tape is where you find out whether the book’s promises were kept.
6. Do not trade from it alone
The feed has no structure, no levels and no history beyond the last few minutes. Entries and stops come from the chart; this informs how you execute them.
7. Check whether your feed is consolidated
Many feeds show one venue. In a fragmented market that is a minority of the actual trading, and a quiet tape can mean a quiet venue rather than a quiet instrument.
How to tell it worked
A size filter is set, removing at least 90 percent of prints.
You can state the 1 thing you are watching for, rather than watching generally.
0 entries came from the tape alone, without a structural level.
And the feed’s coverage was checked once, so you know whether it is one venue or many.
Why the aggressor side is an inference
The feed reports price, size and time. It does not report who initiated. The convention is to infer from whether the trade printed at the bid or the ask, and that convention is right most of the time.
It fails when the spread is wide or the quote is stale. A trade printing between the quotes, or at a price the quote has already moved past, cannot be classified with any confidence.
The tape against the book
The book is intent; the tape is execution. An order on the book can be cancelled at no cost. A trade on the tape has already happened and cannot be undone.
Which makes the tape the more trustworthy of the two, and the less predictive. It tells you what has occurred, with certainty, and nothing about what comes next.
Reading them together is the actual technique. Size resting on the book and then appearing in the trade record is a participant following through. Size resting and then vanishing without appearing in the record is a participant who never intended to.
What a size filter should be set to
Start from the instrument’s typical trade size, not from a round number. A hundred-share floor is meaningless on something that trades in thousands and removes everything on something that trades in tens.
A useful floor removes the large majority of prints and leaves a readable stream. If lines are still scrolling faster than you can read them, the filter is too low.
Then check what you filtered out. Sometimes the small prints carry the information — an iceberg order refilling shows as a run of identical modest trades at one price, and a high filter hides exactly that.
Which means one filter does not serve every question. A floor for spotting large participants and a floor for spotting persistent absorption are different numbers, and switching between them deliberately is better than picking one and hoping.
The original data
Of the 24,971 unique videos in research/search-study-corpus.jsonl, 0 mention time and sales in the
title. The order book appears in 3 at a median of 2,367 and depth charts in 0. Volume generally
appears in far more, but the execution record itself has no coverage at all. The counts come from
site/corpus_count.py.
0 videos on the only display that shows what actually executed. Every trade any strategy takes appears here, and the instructional coverage of the entire microstructure group is three videos across 25,000 — which is the largest single gap this corpus contains.
The answer to the question on that chart is that a large print is a completed transaction, not a direction. Somebody bought and somebody sold, and the tape reports the event without telling you which side was informed.
When it fails
The failure is trading the tape without structure, and it produces entries at arbitrary prices. A burst of large prints at the ask looks decisive, so a position goes on. There is no level behind it, so there is nowhere obvious for the stop, so the stop goes at a round distance instead. The trade is now sized from a number that has nothing to do with the market, entered at a price chosen by a burst of activity that lasted four seconds.
The second failure is no size filter. The feed is unreadable.
A third is treating the aggressor side as fact. It is inferred.
A fourth is reading one venue as the market. Trading is fragmented.
A fifth is reading a large print as direction. Both sides transacted.
And a sixth is watching it with no specific question. The feed never pauses.
Related
The tape covers the feed itself. Order flow is the broader idea this is one view of. And volume analysis is the aggregated version of the same data.
The filter is what made this usable. Unfiltered, an active instrument produces hundreds of prints a second, most of them tiny and automated, and watching it is watching static. Set a size floor and the display becomes a short list of things that might actually matter.
— Michael Whitman
This page is educational, not financial advice. Test every idea on your own charts before risking money.