For game balancing, I let bots play my browser game over and over again. New gear, multipliers and upgrades can all change how difficult a level becomes. That is why I need many test runs to find unusual difficulty spikes. In this Short, I show how local bots help with that work.
This devlog looks back at the development stage shown in the Short published on 11 September 2026.
Why I need so many test runs
Each new mechanic can change the difficulty. An extra multiplier, an upgrade or a new piece of gear can affect the rest of a run. As a result, one successful attempt is not enough to judge how everything works together.
However, I cannot play through my own game a million times. Bots therefore handle many of those attempts. In the video, I talk about 100 bots and hundreds of thousands of runs. The scale of those tests helps reveal balancing problems.
What the bots do in local test mode
These tests run locally in test mode. Their purpose is to help develop the game. They do not artificially increase visits to my website.
The large number in the video title therefore refers to the scale of testing. It does not mean a million real players or a million different bots. What matters is the number of repeated runs used to investigate different difficulty levels.
How the results reveal game balancing problems
The level results are particularly useful. In the Short, I describe points where the bots suddenly take much longer than in other sections. As a result, clear spikes appear in the chart.
These spikes show where I need to look more closely. For example, a level may stand out sharply from the progression around it. I can then ask whether that jump is intentional or whether the balance needs more work.
The goal is not simply to make every level easier. Instead, the game needs a clear sense of progression. Challenges can become harder as players learn how the mechanics work.
Many tests give clues rather than a complete answer
An unusual result first points to a section that deserves attention. On its own, it does not explain why human players might find that level difficult or frustrating. The actual playing experience therefore remains part of the judgement.
Above all, the bot tests help me locate those sections. I can then focus on a particular part of the game instead of judging every change through only a few runs of my own.
The development stage shown in the video
By this stage, I had added gear, new multipliers and upgrades. The bots helped investigate how those additions affected the difficulty and progression curve. Unusual levels still needed further work.
The next step from that test stage was therefore to examine the difficulty spikes more closely and keep improving the balance. The video did not announce a fixed release date or a completed total of corrected levels.
You can try the current free Undies vs Undead browser prototype. For another look behind the scenes, read the devlog about my Blender weight painting problem with a new outfit.
What bothers you more in games: a sudden difficulty spike or long sections without a fresh challenge? Share your experience in the comments.