Monitoring runs — Docs · Verdict Strategy Builder
Verdict Strategy BuilderNT8

Operations

Monitoring runs

The Runs screen (sidebar → Runs, or Cmd/Ctrl+3) shows the whole fleet: every live run across every strategy, then recent finished ones. A strategy's workspace shows the same view scoped to that strategy under its Run stage.

The fleet view

Each live run gets a card: which strategy it belongs to, its State, generation progress and best fitness while evolving, and the gauntlet stages while validating. Filter chips narrow the fleet (All / Running / Validating / Queued / finished), and the slots meter shows how much of the scheduler is in use (e.g. slots 2/3).

Runs execute in parallel up to the slot count (Settings → scheduler); extras queue with a visible position and start automatically as a slot frees — see The strategy library.

State Meaning
Queued Accepted, waiting for a scheduler slot.
Running The optimizer is evolving generations.
Validating Evolution finished; the top finalists are in the validation gauntlet — the card shows each stage's progress.
Done Finished — results ready.
Failed Errored; the run detail shows the error text.
Cancelled Stopped by you.

The live run view

Click a run to open its detail. While it's working you'll see:

  • Stat rowGeneration, Best fitness, Median fitness, Diversity, and total Evaluated candidates, updating as generations complete.
  • Fitness per generation — best (green) and median (gray) lines. Best climbing = still improving; both flat = converged.
  • Population diversity — how genetically varied the population remains. See How evolution works for reading these curves.
  • Top strategies so far — a snapshot of the current leaders, refreshed as generations complete ("Evaluating — best strategies appear as generations complete.").
  • Validation gauntlet — once evolution ends, the stage list replaces the generation counters. Reading the gauntlet stages.

The top bar shows a compact live badge (gen 12 · best 1.87) wherever you are in the app — click it to jump back to the run.

Cancel (top right) stops the run; finished evaluations aren't lost, but the results reflect only what completed. Early generations are the slowest (cold caches) — give a fresh run a couple of generations before judging its pace.

When it finishes

A completion banner sums up the run — "Run complete — 45,000 evaluations over 60 generations." — with a View Results button, and the full results table is embedded below the live charts. From there: Reading the results.

Runs and their results are kept — the Runs screen is your history, and each strategy's Results stage picks from any of its finished runs. Everything persists locally in the app's database, and run/result metadata backs up privately to your account (Account & sync).