A half-ball hit is a useful visual checkpoint: the incoming cue-ball centre line points toward the lateral edge of the object ball. It describes contact thickness, not a complete three-cushion solution.
Read ball thickness at the instant of contact
Set contact thickness first. Then evaluate speed, side spin, throw and table conditions separately.
Which ball travels at 30 degrees?
For equal balls in the ideal no-throw geometry, half-ball overlap gives an object-ball departure angle of 30° relative to the incoming aim line. This is not a claim that the cue ball leaves at 30°, or that the two outgoing balls are separated by 30°.
The cue ball’s later path depends on its spin at impact, speed and the collisions that follow. A stroke delivered at centre height can acquire forward roll before impact; “centre ball” does not automatically mean a stop shot.
Make the reference visible
Start with a clear ball-first position and no obstruction. From behind the cue ball, identify the left or right object-ball edge that sends the object ball in the desired direction. Keep your head alignment and chosen sighting method consistent.
First judge the contact and the object-ball direction, before judging the eventual carom. A missed point after three cushions does not by itself show that your half-ball aim was wrong.
A three-column practice record
Record intended thickness, observed first-ball direction and cue-ball route separately. Repeat one setup with unchanged speed and tip position, then compare a slightly fuller and a slightly thinner contact. Establish a repeatable baseline before adding more variables.
Only after establishing a repeatable baseline should you vary one spin setting. If the first-ball direction changes too, inspect the actual contact: adding side spin may change the delivered line as well as the cushion response. Do not attribute every difference to the cushion.
Where half-ball stops being enough
Positions between familiar fractions still need intermediate aim adjustments. Half-ball is not a universal rail-system starting condition, and a preset with a convenient name is not proof of a measured 50% collision. Use replay to inspect what happened, then check the same distinction on a real table.