Sub-tick splits movement into denser command slices; full stop is a narrow velocity zero band, not a frozen model. Counter-strafe windows depend on reverse-input duration and speed magnitude—train on impacts and spread, not crosshair UI or fixed millisecond tables.
Old muscle memory treats counter-strafing as “see a silhouette, freeze, shoot.” In CS2, sub-tick breaks movement into tighter command slices. The server registers your velocity vector first; crosshair spread only tightens once inaccuracy catches up—what your eyes see and what the hitbox state reads are often half a beat apart.
Zero band: the vector stops, not the camera
After sub-tick, friction and acceleration no longer settle on whole 64-tick frames. The frames where you release W or tap S get chopped into multiple client submissions; when the server stitches continuous displacement, full stop usually means horizontal speed drops into a narrow zero band—not that the model on screen has gone perfectly still.
Common misread in matches: you’re flush to a wall, footsteps sound quiet, but diagonal inertia remains. On a Mirage mid micro-peek, releasing W alone without clearing A/D leaves diagonal friction in a long half-stop. The first shot still eats movement penalty. Treat the counter-strafe window as reverse-input duration × current speed magnitude, not a fixed 80 ms or 120 ms chart.
Reverse hysteresis: don’t share one key rhythm on diagonals
Pure forward/back counter-strafe is one key pair biting the other. Angled exits—Dust2 short wall hug, Ancient outer diagonal cuts—often hinge on release order. Forward first then clear strafe, versus strafe first then release forward, produces different velocity curves under sub-tick buffering; second-shot ownership can shift by half a slice.
Pro-note drill: use sv_showimpacts or demo review for bullet clusters, not crosshair animation alone. Recovery animation is visually eased; spread rebind tracks velocity epsilon more closely. Under high zoom, the reticle can “look stable” while the server still flags light movement. In pro flash swing windows, the second shot is often budgeted for confirmed zero band follow-up—not for “crosshair finished its shrink circle.”
Crosshair recovery: read spread, not UI
However fine your custom crosshair, it’s still UI. Whether your first shot is standing-accurate depends on whether current inaccuracy has already received the standing reward. Under sub-tick movement, short reverse taps (tap-strafe style) sometimes drop into the zero band faster; sometimes they create jagged speed. Map materials, ramps, and step edges amplify the gap—Inferno balcony arc, Anubis hall steps are classic “feels stopped, register didn’t stop” spots.
Training loop: same angle, repeat peek → single tap → peek again, only varying reverse key length and diagonal clear order. Judge the window from vertical spread in impact clusters, not kill feedback. Gun inspect animations on a C5GAME detail page can look smooth; they cannot replace in-match epsilon—inspect runs on a third timeline; don’t calibrate stops there.
Week 39 wrap (hour 14)
This week’s cut-in at hour 14: sub-tick command slices + diagonal friction + spread rebind—read all three together before you write a reproducible team rule for stop windows. Default utility lines can transfer across maps; stop thresholds cannot be hard-coded to one millisecond set across patches. Trust velocity entering the zero band, then trust the crosshair.