Every single playtester of my turn-based tactics game quit at the same place: chapter 3, mission 2. Not 'struggled and pushed through it' — quit, closed the game, sometimes apologized to me afterward like they'd let me down. For six weeks I threw fixes at it. I lowered enemy health. I added a revive item. I cut the spawn count. I rebalanced the loot drops so players would be better equipped going in. After all of that effort, roughly 80 percent of testers still hit that mission and stopped cold.
The cruelest part of being too close to your own systems is that you eventually stop being able to see them. I had stared at chapter 3 for so long that the wall had become wallpaper. I knew the problem was real and reproducible, but my brain had filed it under 'just the way it is' and quietly stopped looking for the cause. I was tuning symptoms because I'd lost the ability to perceive the disease.
Dumping the whole system into the doctor
I ran the Game Balance Doctor — Diagnose the Difficulty Spike, Prescribe the Fix prompt on Grok 4.1 Fast Reasoning and, for once, I held nothing back. Game type: turn-based tactical RPG. System: player attack versus enemy defense scaling. Current values: I pasted my actual formulas, the literal numbers, the per-chapter scaling curves, all of it, ugly variable names included. Feedback: 'chapter 3 felt like hitting a wall, ran out of ways to deal damage.' Desired feel: 'tense but always feeling capable.' I picked the Fast Reasoning model on purpose because I wanted to iterate — it turns answers around quickly enough that I could test three different what-ifs in the time a slower model would take to produce one.
The one sentence that ended six weeks
The diagnosis was almost rude in how fast it landed. My enemy defense was scaling quadratically — squared against level — while player attack scaled linearly. By chapter 3 those two curves had crossed into an asymptote: raw player damage was bumping a hard ceiling while enemies kept getting exponentially tankier. I had spent six weeks adjusting enemy health, which is additive and basically irrelevant next to the actual driver. I'd been bailing water out of a boat that had a hole in a completely different compartment, and feeling virtuous about how much water I'd moved.
- Changed enemy defense from a quadratic curve to a 1.5 exponent with a soft cap at chapter-3 levels
- Left enemy health almost untouched — it had never been the problem
- Re-tested with 8 fresh playtesters: 7 of them cleared chapter 3 on their first attempt
- The PRINCIPLE section taught me to graph two scaling curves against each other before shipping any system
The risk warning that saved my next playtest
The systemic-risk section flagged something I would have walked straight into and faceplanted on: softening the defense curve would make my chapter 5 enemies, which had been tuned against the old broken math, suddenly feel trivially easy. It was completely right. I pre-emptively re-tuned chapter 5 in the same sitting instead of discovering the problem the embarrassing way, two weeks later, in front of a fresh cohort of testers who'd wonder why the late game was a cakewalk. That one warning probably saved an entire feedback cycle and a chunk of my dignity.
There's a deeper lesson in that risk section, which is that a balance change is never local. Every number you touch is connected to other numbers downstream, and the difference between a senior designer and a flailing one is mostly the habit of asking 'what else moves when this moves?' The prompt asks that question for you, every time, and it asks it before you ship instead of after.
Honest caveat: this prompt is only as good as the numbers you feed it. The very first time I ran it, I described my system in vibes — 'enemies get strong really fast' — and got a vague, hedgy answer back that was no better than my own guessing. Paste the real formulas. Paste the actual curves. It reasons about math, so if you give it math it gives you a diagnosis, and if you give it adjectives it gives you adjectives.
If a playtest cohort keeps dying in the exact same spot and you've genuinely run out of guesses, grab the Game Balance Doctor — Diagnose the Difficulty Spike, Prescribe the Fix prompt on Prompt Dock and feed it your real curves, not your feelings about them. Then, once the numbers behave, I run the encounter through my Full Questline Generator — Objectives, Twist, Branching Endings, and Rewards prompt to make sure the mission's pacing matches its new difficulty. Fixing the math is half the job; making the fixed math feel intentional is the other half.