Relative volume is a ratio, and everybody publishes the top half

Relative volume is the number this whole site is built on, and it is a ratio. Current volume on the top, expected volume on the bottom.

The top half is trivial. Anyone can count shares.

The bottom half is where every screener quietly differs from every other screener, and almost none of them will tell you what they used.

Here is why it decides the answer.

One stock, one day, measured minute by minute

This is $AAPL, from our own volume baseline. Average shares traded in a single minute, by time of day.

Read the three marked points, because they are the argument.

• 9:30am, the opening minute: 411,895 shares.
• 12:00pm: 17,489 shares.
• 3:59pm, the closing minute: 638,118 shares.

The opening minute is 23.5 times the noon minute. The closing minute is 36.5 times it. None of that is unusual activity. That is Apple having a completely ordinary day.

Now the premarket point, which is the one that catches people. At 9:00am, normal for $AAPL is 995 shares. The opening minute is 414 times that.

What goes wrong when the denominator is flat

Suppose a screener divides current minute volume by a simple daily average instead of by the average for that specific minute.

At the open it flags everything. Every liquid stock in the market is doing twenty times its daily-average minute at 9:30, because that is what 9:30 is. A list of "unusual volume" built that way is just a list of large caps, reprinted every morning.

At noon it flags nothing. A stock genuinely doing three times its normal midday trade is still below its daily average in absolute terms, so it never clears the filter. The real signal is invisible precisely when the tape is quiet enough for it to mean something.

In premarket it is nonsense in both directions. $AAPL doing 10,000 shares at 9:00am is ten times its actual 9am norm and worth a look. Against a daily average it rounds to zero.

A flat denominator does not measure unusual volume. It measures what time it is.

Our baseline, including the part that is thin

We could have left that last row out. It is the sort of number nobody publishes about their own product.

But a baseline built from a single day is not a baseline. It is one day's volume wearing a baseline's name, and dividing by it produces a confident-looking multiple with nothing behind it. You should know which cells those are, and the honest answer is that they cluster in exactly the places where a thin tape makes any ratio unstable: illiquid names, and the quiet minutes at the edges of the session.

This is the same shape as the rest of this series. A Baseline is a denominator, and a denominator you cannot inspect is a number you cannot check.

Five questions for any relative volume figure

1. Is the denominator matched to the time of day? If it is a daily average, the number is measuring the clock. This is the one that matters most.
2. How many days is the average built from? Ask for the count, not the window. A 60 day window with 3 days of data in it is a 3 day average.
3. Does it handle premarket and after hours separately? A session-hours average applied to 8am traffic is meaningless.
4. Is it split adjusted? A stock that did a reverse split has an old share count that is not comparable to today's, and an unadjusted baseline will read the split itself as a volume event.
5. What does it do when there is no history? A new listing has no baseline. The honest behaviours are to say so or to show nothing. Printing a confident multiple is not one of them.

Our RVOL scanner runs on the minute-matched baseline above, and the other stock scanners share it.

The rest of the series takes the same approach to three other numbers: the four that decide a short squeeze, why almost nobody can source a float, and how old a 13F already is.

Every one of them is the same lesson. The number on the screen is a fraction, somebody chose the bottom half, and if you cannot find out what they chose then you are not reading a measurement. You are reading a decision that somebody else made for you.