When you press the Patmos Jiggle button, Coot runs a multi-stage rigid-body domain placement algorithm against the current map. It first does a limited molecular placement search (Crowther rotation function + Phased Translation function) to establish a starting position. Then it runs 1000 random jiggle trials against a heavily blurred version of the map (B-factor ~300 Å²): each trial applies a random rotation (uniform quaternion sampling) and a random translation of up to ±3.3 Å, with an annealing schedule that gradually reduces the amplitude as the trial count climbs. The top 10 scoring trials are carried forward into a rigid-body refinement step and re-scored, and the best result is accepted if it beats the starting placement. Finally, a short second jiggle pass (12 trials, ±0.4 Å) runs against the unblurred map to fine-tune the fit. The blurred-first, sharp-second strategy is the distinctive feature: the smooth density landscape of the blurred map lets the random search escape local minima, while the sharp map does the final precise fitting.
