Momentum acceleration
Marks bars where the recent return is stronger than its fair share of the longer-run move.
It takes the return over the last short_period bars and subtracts the return over the last long_period bars, scaled to the short window's length. A bar is marked when that difference is positive, that is, when the recent return is more positive than its fair share of the longer one. The test ignores direction, so a falling price whose drop is easing also gets a marker.
How to read it
- Markers sit on the candles where the test is positive. No line is drawn.
- At defaults (3 and 8 bars) about half of all bars are marked, so a single marker says little.
- Runs of consecutive markers mean the recent move keeps beating its longer pace; gaps mean it has evened out.
- Read a marker against the trend: it can mark a rally speeding up or a selloff losing steam.
What it outputs
- Acceleration
- Marks each bar where the recent return beats its fair share of the longer return. Blank otherwise.
When it fires
A bar is marked when the recent return is more positive than its fair share of the longer return. That covers rallies speeding up and drops slowing down. At defaults it switches on about 144 times a day in the September SPY sample, so one marker is an ordinary event, not a breakout.
Settings
- short_period
- The length of the recent window being measured. Raising it gives smoother, slower runs of markers. Lowering it toward 1 compares only the last bar or two, which is noisier. Keep it below long_period; equal values never fire.
- long_period
- The length of the longer window that sets the fair pace. Raising it averages that pace over more bars, so it moves less with each bar's noise. Lowering it toward short_period makes both windows overlap, so the test compares nearly the same return.
Pitfalls
- It ignores direction. A price that is falling but falling more slowly is marked just like a rally that is speeding up, so read markers against the trend.
- At defaults about half of all bars are marked, so the marker is close to a constant condition rather than a rare event. Markers alone do not show an edge.
- Equal short_period and long_period never fire, because the two returns cancel to zero. Despite the name, it is a two-window return comparison, not a growth-curve fit.