Now licks the first pole 12f faster than last WIP.
Sliding down is still unoptimized, very high jump which probably can't go any lower. Inputs sadly can't be distributed.
Jumping down the void while grabbing the pole might save a rotation step.
(10561/512, 117.3620605)u faster in XZ-plane at frame 943. Higher speed and lower angle too. Not jumping into the wall at frame 746 is better.
Sometimes jumping can alter speed, either in one direction, about 0.00001u.
When ledge-boosting delaying? movement in the direction perpendicular to the wall can give a higher speed; Delay by increasing optimality in the other, X or Z, direction, usually by changing angle.
Changing angle after boost can slow down speed decrease.
TODO: Optimize stuff from above, maybe 0.2u can be saved.
Old TAS can be converted to a bk2 file.