Predictive aiming

Three questions for the mechanic

• What movement does the boss sample?

  Use observed position and recent velocity at a defined sample time. Limit the look-ahead interval and clamp the forecast to reachable arena space; do not secretly read future player input.

• When does the prediction stop changing?

  Show the forecast marker during aiming, then freeze it at an audible and visible release cue. Spawn the shot on that committed line and keep its velocity fixed unless a separate homing rule is announced.

• Can an ordinary course change clear it?

  Test from several approach speeds, distances, and directions. Include the whole player body and projectile radius in clearance checks; leave enough time to reverse or sidestep after the lock.

Design mistakes

• Prediction uses hidden future input

  A perfect forecast that reacts to a turn made after commitment feels like input reading. Derive it from a past sample and make any later correction a separately signaled mechanic.

• The marker is too subtle or too late

  If players cannot compare their current position with the projected point before release, the attack reads like an unexplained miss or unavoidable hit. Use geometry and timing, not color alone.

• Lead time outgrows the response window

  An extreme forecast can cover every reachable turn or point beyond the arena. Cap the horizon and test the response at low and high player speeds.

Tune the attack

• Predict from data the boss could observe

  Estimate velocity from two recent positions, smooth noise if needed, and compute target = position + velocity × lead time. Make the sample window and lead time explicit tuning parameters.

• Keep forecast and live threat separate

  The dashed marker is a destination cue, not a damage zone. Collision belongs to the moving shot; it should not fire merely because the player touches the future marker.

• Budget motion, not only distance

  Check a standing player, a steady runner, and a runner who reverses just after commitment. Tune projectile travel time and player acceleration together so the intended response stays possible.

• Vary one axis at a time

  Later cycles may shorten the lead time or add a second forecast, but preserve the same lock cue. Combining faster prediction and longer tracking at once hides what changed.

Arena

• What the mechanic communicates

  Kern reads the momentum of Tavi's run and stamps a rune where that route would lead. The stone shot obeys the completed mark rather than chasing Tavi's later decision.

Player upgrades

• Adapt it to the player verbs

  Tavi needs only ordinary movement to change course after the lock. If a dodge, dash, or guard is added, decide whether it creates a new escape or merely shortens a window that was already fair without it.

Do not confuse this with

• Keep the mechanic distinct

  Target lock fixes the player's current position; predictive aiming projects from observed movement. Source tracking keeps turning the firing source, while homing changes a projectile after release. This shot does neither once Kern commits.