DIE AGAIN: Behind the Scenes

The part of Die Again we are most excited about is something the player should not notice directly, but should feel in the quality of the experience.

We are planning a data-driven balancing and testing layer using:

• quantitative modeling

• simulation

• Reinforcement Learning

• Imitation Learning

• UE5 Learning Agents

In practical terms, this layer would help us with:

1. Difficulty and pacing tuning

Use quantitative / stochastic methods to understand how:

• spawn rates

• enemy health

• weapon damage

• cooldown timings

• support systems

affect the player’s core pressure-to-power loop.

We also plan adaptive support systems such as ammo or med-kit assistance based on how the player is actually performing.

2. Progression freshness

We want each new map to feel familiar, but not repetitive. RL / IL agents can help us test whether a level, map layout, or mission is actually producing different behaviors and strategies, instead of repeating the same solution.

3. Automated QA and repeated playtesting

RL / IL agents can support:

• repeated automated playthroughs

• exploit detection

• difficulty spike detection

• dominant strategy detection

• balance validation across settings and maps

This is especially valuable for a two-person team.