6 tasks, each one witnessed by the sources that watched the job — and behind every one, a prompt you can use tonight.
You spend time reading telemetry (match logs, item picks, win rates) and running tests to spot what's unfair. Mornings often mean checking recent patches and automated dashboards for spikes or regressions.
Afternoons you build or run test workflows in Docker containers, run quick SageMath scripts to model numbers, and write short reports that list assumptions, evidence, and recommended changes for designers to review. Meetings with designers or QA happen once or twice a week to agree next steps.
Expect to use Docker to run isolated test environments (replay servers, analytics pipelines) so experiments don’t break production. You’ll also run SageMath or similar math tools to model probability, damage curves, or economy inflation.
If your team uses an AI assistant like Claude, you may use it to draft summaries or generate test ideas — but you must check every output against your data and logs before trusting it.
Use SageMath when you need symbolic math, probability distributions, or to build a small simulation (Monte Carlo) for expected damage or match length. SageMath handles exact fractions, integrals, and random sampling scripts more cleanly than Excel.
You still export numbers to CSV for dashboards, but SageMath gives clearer proofs of why a change shifts win rate by X percentage points. Save the scripts and results in your structured logs for comparison.
Yes, but treat AI outputs as drafts. Use Claude to summarize long logs, suggest test cases, or rewrite reports for clarity. Always check any numerical claim against your raw telemetry or SageMath models and mark where the AI was used in your documentation.
Never change game values based on an AI suggestion alone. Record the AI prompt and the verification steps in your structured logs so reviewers can trace assumptions and evidence.
A data scientist builds production models and pipelines across the company; a balance analyst focuses on game-specific metrics (win rate by skill bracket, item pick rate) and short-term experiments. You often use simpler, targeted models and lots of controlled tests.
A game designer decides what should feel fun; you translate that intent into measurable models, run tests, and check assumptions. You’re the bridge: you document results, propose small numerical changes, and keep logs for comparison.
Learn statistics (mean, variance, confidence intervals), probability (distributions, expected value), and basic simulation (Monte Carlo). Practice by modeling simple systems in SageMath and running small experiments in Docker.
Also practice writing clear assumption checks and short reports. Employers want concrete examples: a Git repo with SageMath scripts, Docker test setups, and a few structured logs showing how you tested and documented one change.
No one can promise exact pay or review times; those vary by company and region. Typical ranges (depending on location and experience) are entry-level to mid roles that report salaries in public job listings — check your region’s job boards and Glassdoor for numbers.
You can expect decision timelines: small balance tweaks reviewed in 1–2 weeks, larger systemic changes take multiple sprints and more documentation. Always record review dates and outcomes in your comparison logs so you can track how long things actually took.