Open dvr333 opened 7 years ago
On closer inspection i no longer suspect raw move:
There are a lot of incredibly narrow misses against seemingly straight moving Warriors but again on closer inspection they seem to be imperfectly straight world lines. They stutter, make pauses, and make small heading changes. That these small heading changes are sufficient to dodge may be an issue.
But i'm not sure i want to see map-spanning lines of rockos again.
Is this still a problem after Hoko projectile speed buff f315e830ab2d2fcff3cebc856350bf12e4502d9c ?
Engine. Engine leading is wrong: https://github.com/spring/spring/blob/d5771b9cfdf35491853f6130106535585bf5b6bb/rts/Sim/Weapons/Weapon.cpp#L1244
A shoots B with a projectile, 10 speed B walks towards C at 5 speed Projectile ETA from A to B = 17 time 17 time * 5 speed = 85 distance (point D) Projectile ETA from A to D = 11 time Projectile arrives at D 6 time units before the target. Overshoot.
So raw move made units move in lines that are straight enough to reveal aiming issues?
I think it might rather be lines not straight enough over time. Rocko seems to be able to lead just at the edge of the hitbox so any tiny changes in heading can cause a miss. Raw move can apply such tiny changes when spamming orders but regular move just keeps going along the 8 cardinal directions.
Apparently Bandit can fail to hit Glaive, unsure if the same root cause: https://zero-k.info/Battles/Detail/1192709
https://github.com/ZeroK-RTS/Zero-K/blob/master/LuaRules/Gadgets/weapon_projectile_retarget.lua now exists and reproduces the engine behaviour. Maybe it could implement a proper solution and then any problematic units (rocko etc) could be handled there.
Just so we don't lose it.
Niarteloc — Yesterday at 14:38 100% accurate single sqrt flight time for projectile leading equation:
t = (-2*dot(v_unit, dist_to_unit) + sqrt(4*dot(v_unit, dist_to_unit)^2 + 4*(v_projectile^2 - v_unit^2)*dist_to_unit))/(2*(v_projectile^2 - v_unit^2)
Engine ticket beyond-all-reason/spring#746
I was playing a game on Intersection, and while I built Rockos to deal with snoke's Warriors, the Rockos kept over-leading their shots slightly and barely missing. I'm not sure if this is a speed issue, a unitAI issue, a collision volume issue, or an issue with Warriors using the pathfinder instead of ignoring it for raw move (Anarchid's suspicion).
Game: http://zero-k.info/Battles/Detail/444850 Rocko-Warrior engagements start around 11 minutes in.