Power law breakout
Flags breakouts above recent highs that keep growing stronger, behind a 0.5% floor SPY one-minute bars never cleared.
Over the last window bars, each close is measured against the highest high of the bars before it in that window. A close above that high gives a breakout percentage, and closes below it count for nothing. The test takes the average of the first two breakouts and of the last two, and it fires only when the last pair is more than 1.3 times the first pair and above 0.5%, returning the ratio between them. It needs at least three breakouts in the window before it can test them.
How to read it
- Markers sit on the candle that completes an accelerating breakout run; a plain new high does not mark it.
- The window is 12 bars by default; the 0.5% floor applies to the average of the last two breakouts, each a close above the earlier highs in that window.
- On SPY one-minute bars with defaults there were no markers over a full month, so expect a mostly empty chart.
What it outputs
- Accelerating breakout
- Drawn on the candle completing a run where the last two breakouts average over 0.5% and exceed 1.3 times the first two. The ratio is not drawn. Never appeared on the SPY month sample at defaults.
When it fires
Fires on the candle completing a window where the last two breakouts above the earlier highs average over 0.5% and are at least 30% stronger than the first two. It marks a sharp push above the window's highs that is still gaining strength. Never fired on the SPY month sample at 12, 40 or 120 bars.
Settings
- window
- Bars the test looks across. Longer windows give more breakouts to compare, so the accelerating shape shows up in more windows (6,050 of about 20,000 at 120 bars, 1,223 at 12), yet no length tried reached the 0.5% floor. Shorter windows show it less.
Pitfalls
- The 0.5% floor is fixed in the code. On the SPY month sample the strongest recent breakout average was about 0.22%, so no window from 6 to 480 bars fired.
- The window changes how often the accelerating shape appears, not the strength it reaches, so the presets raise the shape count but still give no markers on SPY one-minute bars.
- Each bar re-scans the window, so cost grows with the square of the window and long windows are slow to compute.