← How the Lab works How we built it

We tested our own system, and fixed what failed.

The short version, told straight.

  1. We started with hand-scored battles

    When we tested them on clear mismatches, they only scored about 55. That was too flat, so we said so and kept going.

  2. We built LAB IMPACT ourselves

    From raw line-up data, instead of borrowing anyone else's metric.

  3. We trained Lab Offence and Lab Defence on it

    And checked them on seasons they'd never seen.

  4. We anchored the scale to MVP seasons

    So 70–30 means something.

  5. We locked the rules before every result

    We published the misses and changed nothing to suit an answer. When a fix we wanted failed its own test, we didn't use it.

← Back to How the Lab works