- Home
- Skills
- Game Development
- Game Balance Tuner
Works with the AI Agents you already use
Game Balance Tuner
Turn a CSV or JSON of playtest sessions into exact numeric tuning changes, with expected effects and confidence levels.
$14.99
Game Balance Tuner
Example session with this skill installed
Analyze my playtest data, compare completion against a 60 to 80 percent band, and give me the five highest priority tuning changes with expected effects.
- Read your context and instructions
- Compiled the game balance tuner
Tuning report: Level 3 is below its target band. Recommendation: reduce enemy HP by 25 percent on level 3, from 120 to 90, projected to lift completion from 42 to 49 to 55 percent, high confidence, n equals 306. Levels inside the band get no recommendation. Projections are hypotheses to test on the next build, never promises.
Connects securely to your tools. The creator never sees your data.
About this skill
Turn a CSV or JSON of playtest sessions into exact numeric tuning changes, with expected effects and confidence levels.
For a finished playtest round, a balance patch about to ship, a level that feels too hard or trivial, or an economy that feels broken. Get per level completion funnels, difficulty curves against a target band, economy earn versus spend balance, and specific recommendations such as reduce enemy HP by 25 percent on level 3, from 120 to 90, projected to lift completion from 42 to 49 to 55 percent, high confidence, n equals 306.
Why it matters: Gut feel tuning overshoots. This skill gives numbers instead: before and after values, evidence, a projected range, a confidence label tied to sample size, and the sample size itself. Vague advice such as make level 3 a bit easier is treated as a bug.
What it does:
- Runs bundled Python scripts, standard library only, no installs, on your playtest CSV or JSON.
- Computes per level completion rate, median deaths, median clear time, quit rate, earn to spend ratio, inflation drift, and starting currency adequacy.
- Compares completion against a target band, default 60 to 80 percent, adjustable for genre. A roguelike should not use the same band as a casual puzzler, so it asks.
- Recommends only when a level sits outside its band or the economy shows a clear imbalance: ratio outside 1.0 to 1.5, starting currency below one opening level spend, or drift of 25 percent or more. Levels inside the band get no recommendation.
- Caps changes at 25 percent per patch, 10 percent for levels with under 30 sessions, so a patch cannot overshoot. When the cap binds, it says so, and that it will take more than one round.
- Handles lever direction correctly: higher makes it harder, such as enemy HP, or lower makes it easier, such as player damage or starting health, writing the correct reduce or raise verbs into the report.
- Produces a tuning report in a fixed order: title and summary, level funnel table, difficulty curve, build comparison when several builds are present, economy section, and recommendations.
- Presents the five highest priority recommendations and asks which to lock in.
Honest limits:
- It does not judge fun, pacing feel, art, or story.
- It never invents lever values. Without current lever values per level, changes are given as percent changes on an index of 100.
- Confidence follows sample size: under 30 sessions is low, 30 to 99 is medium, 100 or more is high. A projection resting on the assumed default effect size, not your own data, is downgraded one step. A level with a median of 0 deaths is always low confidence for difficulty changes.
- Under 30 sessions the target band is widened by 10 points each side. A level outside the nominal band but inside the widened one is labeled inconclusive and gets no recommendation.
- Projections are hypotheses to test on the next build, never promises.
- Missing optional columns simply switch off the metrics that need them. No currency columns means no economy section. Missing time data skips clear time metrics.
- Genre presets are not in version 1. Genre is accepted and noted, but no preset band is applied automatically.
Requirements: Python 3 and a CSV or JSON playtest file, one row per attempt at one level, with player id, level id, completed, deaths, and time seconds. Optional currency, session length, build version, and quit columns unlock more metrics.
Example prompt: Analyze my playtest data, compare completion against a 60 to 80 percent band, and give me the five highest priority tuning changes with expected effects.
What is in the package: SKILL.md, LICENSE, references including metrics and schema, scripts for analysis and reporting, and sample data and report.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before any payment.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 23 days ago
Requires