Dev Blog: Why We Rebuilt the Netcode from Scratch

Halfway through Season 4, we made a call that delayed a content update by six weeks: rebuilding the game’s networking model from the ground up. Here’s why.

The problem with our original netcode

Our launch netcode prioritized low bandwidth over precise hit registration, which worked fine at low player counts but started producing inconsistent peeker’s-advantage complaints as our population grew and average ping variance increased.

What rollback-assisted hit registration actually changes

Rather than trusting only the server’s delayed view of a fight, the new system rewinds a short window of player positions to better match what each player actually saw on their screen, then reconciles the result. This doesn’t eliminate lag, but it makes fights feel fair even between two players with different ping.

The trade-off we accepted

This approach costs more server compute per match than our old model, which is part of why ranked queues can take longer during off-peak hours in smaller regions – we’re running fewer, more expensive server instances rather than many cheap ones.

What’s next

We’re still tuning reconciliation windows for very high ping matches (300ms+). If you’re in a region with limited server coverage, our next infrastructure update targets exactly that.