#34 — S5 velocity/acceleration sweep

Hypothesis — what it tests

Whether S5-scale price velocity and acceleration predict the next move — i.e., whether the TR momentum idea works once the fill gap is engineered away.

Description

After the TR Momentum fill artifact was exposed, the user proposed rescuing the idea by replicating the signal with S5 (5-second) rolling windows — sub-pip trail resolution to eliminate the fill gap. Three experiments ran: a 1,296-config M5 velocity/acceleration confluence sweep (fast velocity above threshold + accelerating + slow velocity agreeing), a 288-config S5 fade strategy (because the IC study said the S5-scale signal is counter-trend, not momentum), and a 168-config-per-pair S5 replication of the TR entry with the trail updated every S5 bar (0.1-0.5 pip fill gap instead of 3-7 pips). Every one produced zero walk-forward survivors. The IC feature study on 8.26M EUR_USD S5 bars found IC = −0.026 (counter-trend), hugely significant statistically (t = −62.7) but economically nothing, with LightGBM walk-forward AUC of 0.509 — a coin flip. Even with 10x tighter fills, entry after a big bar is a random entry paying the spread.

Key result

Indicators

Algorithms

Code