avg_dollar_volume multiplies today's price by a three-month average volume

[Ben05-sys]

In app/universe.py:

df["avg_dollar_volume"] = df["price"] * df["avg_volume_3m"]

That is today's price against a three-month average volume — two different clocks in one number — and it is then classified STATIC_SAFE in app/screen.py, meaning a screen can prefilter on it before live quotes are fetched.

The intent is defensible and documented: it is a liquidity floor, not a reading of the tape. avgdvol > 20m means "this name normally trades enough to get out of", and half a day of price drift genuinely does not change that answer. That is why it is static-safe while today's dollar_volume deliberately is not.

But the docstring does not say that the two inputs come from different periods, and a reader checking whether the field means what they think will not find out from the code. For a name whose price has halved since the quarter began, the figure overstates historical liquidity by 2x.

What to decide

Either is a good outcome — pick one and argue for it in the PR:

• Document it. Extend the comment in derive() to state the mixed time base explicitly and why it is the right trade for a liquidity floor. Cheapest, and possibly correct.
• Fix it. Use a longer-run price so both factors describe the same period. Needs a price series, which is the same missing piece as Indicators inside the expression engine (RSI, MACD, ATR) #3 — so it may be worth doing after that.

If you change the number

Anything that shifts what avgdvol returns changes which rows every liquidity screen matches, so:

• Add a test with hand-computed values showing old versus new for a name whose price has moved a lot.
• Check the dvol dropdown in FILTER_SPECS still compiles to thresholds that mean the same thing.
• tests/test_screen.py already asserts avgdvol > 20m narrows before re-pricing. Keep that property or explain why it should change.

Small, self-contained, and it makes you read the most interesting part of the screener — the static-versus-live split.