High Ground
The wall works. E7 fixed pathfinding through the chokepoint, with a physics constraint as backstop. But the emulated map was still flat — every tile either walkable or wall, no elevation, no tactical geometry. SC2 isn’t flat.
E8 adds height.
The terrain model extends from two values to four: HIGH, LOW, RAMP, and
WALL. The 64×64 emulated map now has real topology — enemy staging above the
wall line (y > 18), nexus side below (y < 18), chokepoint gap as a ramp rather
than just a hole in the wall.
I made one deliberate tactical choice: enemies attack from high ground. My ranged units defending from below have a 25% miss chance when firing uphill. It’s the harder scenario — no natural defensive advantage — but it’s the correct one. High ground matters in SC2 because of exactly this penalty.
The miss check lives in resolveCombat(), not in DamageCalculator. The
calculator stays pure — damage formula only. The guard fires before
computeEffective():
private boolean missesHighGround(Point2d attackerPos, Point2d targetPos, UnitType attackerType) {
if (terrainGrid == null) return false;
if (SC2Data.attackRange(attackerType) <= 1.0f) return false; // melee — never penalised
Height ah = terrainGrid.heightAt((int) attackerPos.x(), (int) attackerPos.y());
Height th = terrainGrid.heightAt((int) targetPos.x(), (int) targetPos.y());
if (ah != Height.LOW || th != Height.HIGH) return false;
return random.nextDouble() < 0.25;
}
Testing probability logic without a mocking framework: inject an anonymous
Random subclass with a fixed nextDouble() return — 0.0 for always-miss,
1.0 for always-hit. Random.nextDouble() isn’t final. No Mockito, no
seed-hunting — the boundary condition is explicit.
One subagent per task, two reviews each
I had Claude execute E8 as subagent-driven development — seven tasks, each dispatched as a fresh subagent so context didn’t bleed between them. After each: spec compliance review, then code quality review, before moving on.
The review gates caught things. A quality reviewer flagged java.util.Random
written fully qualified when java.util.* was already imported — inconsistent
with every other java.util type in the file. Another caught @Tag("browser")
on the new Playwright test but not on two existing pixel-sampling tests, meaning
those tests ran in the default surefire suite but were excluded from the playwright
profile. Both fixed before the next task began.
378 surefire tests. The wall-physics and pathfinding tests from E7 still pass.
The map now looks like something
Warm tan/brown above the wall line — HIGH ground. Dark canvas below — LOW ground. The wall stripe with the ramp gap visible through it. It looks like a proper topographic map now.
The next step is fog of war. I parked it out of E8 deliberately — full fog of war is a system-level feature, and half-measures leave an inconsistent game model. It gets its own epic.