Game Jam Post-Mortem: Everything Must Ghost

Everything Must Ghost was created during the 2025 Out Of Office Game Jam event with the theme “Everything Must Go!” Our team had 7 days, from Aug 22 at 9 PM EST through Aug 29 at 9 PM EST, to design, build, and submit a playable game. After brainstorming and voting on 14 pitched ideas, we landed on the concept of a ghost competing against another ghost by stealing items from a home with everything, even the house itself, up for grabs. Our goal was to make something fun, goofy, and charming with a competitive edge, even when playing against an AI opponent.

Team & Roles

The team varied in experience from fresh grads to industry vets. While we had dedicated roles, many teammates wore multiple hats and contributed outside their primary area.

• Team Size: 8 developers (plus external playtesters)
• Roles:
  •  3 Programmers 
    • AI
    • UI
    • Gameplay (me)
  • 2 Artists
    • Environment
    • Prop
  • 2 Designers
    • Gameplay/System
    • Level 
  • 1 Sound Designer / Composer 
  • 1 QA & Narrative Writer 
  • 1 Producer (me) 
  • 5 Dedicated external playtesters
• Tools Used: Unreal Engine 5, GitHub, ClickUp, Maya, Discord, Logic Pro

What Went Well

The team met before the jam to align on strengths and preferred roles, which helped us get off to a fast start. Our team’s mix of recent graduates and industry veterans created a balance of strong practices and new learning opportunities. By the end of day 3, we had a playable prototype, allowing playtesters to highlight important design adjustments early. Dedicated QA support caught critical bugs along the way and kept polish from slipping through the cracks. Using a mix of premade and custom assets allowed us to build a wider variety of collectible items.

Key positives:

• Clear role assignment based on strengths and interests
• Dedicated QA leading to early bug fixes and polish
• Strong idea that fit the jam theme perfectly (collecting everything, including walls)
• Industry experience enabling quick iteration
• Asset packs combined with custom work for item variety

What Went Wrong

Even though our concept was straightforward, we tried to do too much for the time we had. We originally scoped fourteen rooms across multiple floors but had to cut it down to just the first floor and reduce the timer to keep players focused. Midway through, one teammate had to leave unexpectedly, and real-world responsibilities like family, and vairas problems such as an SSD failure, and a multi-day power outage slowed progress. Near the end, we were scrambling to finalize the build, and although QA flagged a critical bug, my rushed response under deadline pressure meant it still slipped into the release.

Main challenges:

• Over-scoping the project (14 rooms and 2 floors planned, only the first floor shipped)
• Losing a teammate midway through development
• Unexpected disruptions: hardware failure and power outage
• Repository issues that consumed time and caused delays
• Miscommunication with QA leading to a game-breaking bug in release
  

Another unique technical challenge came from how we handled item pricing and weight. Each object in the game was designed to have its own randomized values, even when items were duplicates. Everything worked fine in the editor, but in test builds many objects reverted to default values. After a day of troubleshooting, we discovered that every item was pulling data from the same sheet at spawn, overloading the system. The solution was to defer value assignment until the first time an item was looked at or interacted with, reducing thousands of simultaneous calculations to just one or two (if the AI was interacting with a different item simultaneously) at a time.

What We Learned

Looking at other jam submissions, I noticed that some of the most impressive projects had very simple ideas but executed them extremely well. That showed me the value of keeping scope small and polishing a few features. Communication also proved to be an area for growth: tasks were clearly assigned, but breakdowns led to duplicated work and lost time. Around the halfway mark, I suspected our level was too large but relied on optimism instead of data, which was a mistake. Finally, working with someone highly experienced in QA gave me a much deeper appreciation for the discipline, and how QA extends far beyond bug reporting.

Key takeaways:

• Keep scope small and polish a limited set of features
• Ensure task assignments are not only clear but also followed consistently
• Trust data and progress tracking over optimism when evaluating scope
• Appreciate QA as a critical role that shapes the overall player experience

Conclusion

Everything Must Ghost was ambitious, and while we had to cut scope, the final game captured the goofy, competitive spirit we set out to create. Strong brainstorming, clear role alignment, and dedicated QA allowed us to deliver a polished submission despite real-life setbacks and technical challenges. Most importantly, the project reinforced the value of focusing on a well-executed core, maintaining consistent communication, and treating QA as a true partner. These lessons will carry forward into future jams and projects, helping us balance creativity with practicality while still making games that are fun and memorable.