Reward-Adaptive Iterative Discovery: A Case Study on Automated Game Testing for NHL26

Source

• Source name: arXiv
• Source kind: paper
• Authors / organization: Florian Fuchs, Jessy Gosselin-Grant, Boris Skuin, Michele Petteni, Alessandro Sestini, Joakim Bergdahl, Amir Baghi, Linus Gisslén
• Venue: arXiv: cs.LG
• Published: 2026-07-08
• Collected: 2026-07-09
• URL: http://arxiv.org/abs/2607.07498v1
• PDF: https://arxiv.org/pdf/2607.07498v1
• DOI: UNKNOWN

Plain-Language Brief

The authors present Reward-Adaptive Iterative Discovery (RAID), a new approach for automated game testing using Reinforcement Learning (RL) in the gaming industry. The case study is based on EA SPORTS NHL 26's development version, focusing on testing the goalie AI for behavioral exploits. RAID aims to reduce re-testing efforts by finding multiple diverse high-quality solutions instead of overfitting to a single solution.

What the Document Says

• The text discusses a case study about EA SPORTS NHL 26's development version.
• Human playtesters test the goalie AI for behavioral exploits in the game.
• Reward-Adaptive Iterative Discovery (RAID) is proposed as a new approach to reduce the effort of re-testing the goalie AI.

Key Concepts Mentioned

• Testing
• Gaming industry
• Reinforcement Learning (RL)
• Reward-Adaptive Iterative Discovery (RAID)

Claims Made by the Document

• RAID can reduce the effort of re-testing the goalie AI.

Evidence Used

• The text mentions that RAID was able to find six hockey scoring exploit strategies within a single experiment.

Methods / Approach

• RAID uses an iterative Reinforcement Learning (RL) approach

Limitations Stated by the Authors

The authors mention that RL algorithms tend to overfit to a single solution, and the proposed extension aims to find multiple diverse high-quality solutions.

Limitations Noted by the Agent

Agent interpretation — not from the source.

It's unknown if there are any limitations mentioned by the authors regarding the application of this approach in different game contexts or with other types of AI.

Open Questions

• How effective is RAID compared to manual testing for finding various exploits across multiple games?
• What is the scalability of RAID when applied to larger and more complex games?

Notes for Future Q&A

May include interpretation. Labeled where it does.

The text suggests that RAID finds qualitatively similar strategies as those found through hours-long manual testing sessions, but it doesn't provide information about its efficiency or effectiveness in other contexts.

Raw Abstract / Excerpt

Testing is a major effort for the gaming industry, requiring a significant part of development budget and people power. We present a case study on a development version of the ice hockey game EA SPORTS NHL 26, for which human playtesters test the goalie AI for behavioral exploits. To reduce the effort of re-testing the goalie AI after every game or behavior modification in the development phase, we propose Reward-Adaptive Iterative Discovery (RAID), a novel approach to automatically find exploits using an iterative Reinforcement Learning (RL) approach that trains a population of goal scoring agents. While previous approaches can already successfully find exploits, RL algorithms tend to overfit to a single solution. We introduce a simple extension on top of existing RL algorithms, such that they find multiple diverse high-quality solutions. For our first deployment of this approach, within a single experiment we were able to find six hockey scoring exploit strategies that were qualitatively similar to those that playtesters had found in hours-long manual testing sessions.