Volume acceleration
Volume bursts out of a quiet stretch, marked on the candle; most fire outside regular hours.
The name suggests acceleration, but it only looks at volume, never price. It takes simple averages of volume over the last 3, 10 and 20 bars, each including the current bar. A bar is marked only when all five tests pass: volume is over 1.8 times the 3-bar average; that average rose more than 10% on this bar and also on the bar before; the 10-bar average sits over 10% above the 3-bar average; and the 20-bar average sits over 5% above the 10-bar average.
How to read it
- Markers sit on candles only. The indicator draws no line, and a marker says nothing about which way price moves.
- Each marker is a volume burst: that bar traded well above the recent 3-bar average, which had started climbing.
- In the September sample most markers came before 09:30 or after 16:00, so judge them against the thin volume around them.
What it outputs
- Volume burst
- Marks bars where volume bursts out of a quiet stretch. The marker's value is the bar's volume as a multiple of its 3-bar average.
When it fires
Fires when volume is over 1.8 times its 3-bar average, the 3-bar average rose over 10% on this bar and the bar before, and the 10-bar average sits over 10% above the 3-bar average with the 20-bar over 5% above that. It means a volume burst out of a quiet stretch. Price direction is not tested.
Settings
- short_window
- Length of the fast volume average, which includes the current bar. At 1 the output never fires; at 2 it fired only 5 times in the September sample. At 4 or 5 it fired 71 times, against 48 at the default 3.
- medium_window
- Length of the middle volume average, which must sit at least 10% above the fast one. At 5 the September sample had no markers; at 15 it had 53. Longer values keep older volume in the baseline, so the test passes more often.
- long_window
- Length of the slow baseline, which must sit at least 5% above the middle average. A value equal to the middle window means it never fires. Moving it from 20 to 30 or 60 changed the September count only a little (49 or 51, against 48).
Pitfalls
- The code comment says volume is expanding across timeframes, but the gap tests need the 3-bar average lowest and the 20-bar highest. That looks inverted; read the markers as the code behaves.
- Of 48 markers in the September 2026 SPY sample, 47 fell outside 09:30 to 16:00. It mainly picks thin extended-hours bursts, and regular-session volume almost never passes.
- Equal windows never fire: a medium window equal to the short one, or a long window equal to the medium one, makes a gap test impossible. The first long-window bars of a run show nothing.