About this game North & South
Development notes
Only orders cross the network
The server and the browser run one shared simulation. The server sends each accepted deploy or order with its tick number instead of unit positions, which keeps matches light on mobile connections. So every machine gets the same result, the simulation avoids Math.random, the clock, and functions like sin and pow that can differ between JavaScript engines; a checksum every two seconds catches drift.
Fair by rotation, not mirroring
Each map stores only its bottom half; the top is that half rotated 180 degrees, so both players see the same screen. Deploys snap to the grid in each player's own view, because snapping in world coordinates sent tile-edge clicks to different tiles for each side, and one side won noticeably more in AI-versus-AI games.
Supply that helps the side behind
When each held point added 30% supply, the first to capture also built faster: in 40 bot games the winner led from 18 seconds on, and only 3 games turned around after the first minute. I cut the bonus to 10% and gave the side holding fewer points 30% per missing point. In the 2026-10-02 run the winner emerged around 61 seconds, 20 games turned around, and Hard still beat Easy as often.
Testing balance with bots
A Sharpshooter-only bot beat Normal and Hard AI every game, since nothing chased a range-8 shooter outside the point. On 2026-10-09 Cavalry began charging ranged units within 9 tiles and the Sharpshooter dropped to range 7, 300 HP and a 3-second shot, cutting that bot to 21% and 13% wins. The AI levels once differed only in thinking speed, and Normal lost to Easy 41–59; with a plan per level, Normal now beats Easy 99–1.
Updated: