Read the explanation
Battle AI system

How PokeSynergy tests Battle AI

PokeSynergy tests tactics and team matchups on separate, complementary axes, promotes only after bounded reproducible evidence passes, and sends failures back into the next candidate rather than hiding them.

Read the explanationReviewed · PNG download · Updated 21 October 2026
How PokeSynergy tests Battle AI: stages and decision branches, explained in text below.
Two complementary evidence axes · bounded reproducible gates · failures loop back · Read the equivalent explanation below, or open the PNG to inspect every label.

Scope and inputs

Intended reader: Competitive players and product stakeholders who want to understand what Battle AI tested means without reading benchmark code.

Two complementary evidence axes · bounded reproducible gates · failures loop back.

Follow the stages below. Branches name their destination; a loop returns to the named stage.

Stages and branches

  1. Test 12 Tactical Views

    Six fixtures, original plus mirrored.

  2. Test 12 Source-Pinned Teams

    66 pairings, 660 replay-verified battles.

  3. Keep Both Test Axes Separate

    Complementary evidence. No 12×12 run.

  4. Tune on Visible Panels

  5. Run Reproducible Qualification

    Lineup and turn tracks stay separate.

  6. Reproduce Owner-Found Exploits

    Current gate remains owner-deferred.

  7. Test the Live Product Run

    Run evidence, never certification.

  8. Keep Promotion Evidence Sealed

    No tuning on holdouts.

  9. Is This Version Ready?

  10. Promote With Rollback

  11. Fix One Mechanism

Outputs

  • Promote With Rollback
  • Fix One Mechanism

Limits and uncertainty

  • 12 tactical views ≠ 12 teams · Product run ≠ certification · Fixed cohort ≠ ladder odds

What this does not prove

A bounded evaluation is not perfect-play proof or a guarantee of results outside its declared protocol.

Glossary

orientation
the same test viewed from the opposite battle side
regret
bounded counterfactual gap under a disclosed cutoff, not perfect-play proof
no mirror
one unordered A-v-B matchup; its paired battle schedule still flips sides
holdout
a separately named, untouched partition reserved for a gated decision

Sources and review

Reviewed for publication . Source observations retain their original dates.

  • Tactical fixtures · observed 2026-08-07
  • Regret panels · observed 2026-08-07
  • League partitions · observed 2026-08-07
  • Team corpus · observed 2026-08-29
  • Qualification artifact · observed 2026-08-29
  • Product limits · observed 2026-08-24
Source references and revision

Source SHA-256: ba00a4c462b8211bba90020618ea3453f215ca1a8ffa76590f894861dc3d85f7

Originating chart: POK-99

  • Tactical fixtures: scripts/calibrateChampionsBattlePolicy.ts
  • Regret panels: scripts/benchmarkChampionsDecisionRegret.ts; scripts/benchmarkChampionsLeadRegret.ts
  • League partitions: scripts/benchmarkChampionsPolicyLeague.ts
  • Team corpus: src/server/battle/championsTeamScoringCorpus.ts
  • Qualification artifact: data/bundled/champions/Champions Doubles M-B S3/battle-analysis/team-scoring-qualification-v9.json
  • Product limits: shared/battleAnalysisJobs.ts