Methodology
Public, versioned methods. Every result shows protocol + platform + GameVersion + date + uncertainty — or it doesn’t ship.
No fabricated shipping values. When real measurements are absent, we show method only and report population as BLOCKED_EXTERNAL.
Vehicle
Top speed, acceleration, braking, handling — two-way mean, 10 runs, flat road.
“How fast is this vehicle, really?”
View method →
Police
Recognition, witnesses, wanted behavior, search duration, escalation.
“How does the wanted system behave?”
View method →
NPC / World
Density, persistence, reactions — controlled observation windows.
“How does the world simulate?”
View method →
Economy
Payout per hour, costs, efficiency — with fees and variance noted.
“What pays best per hour?”
View method →
Performance
FPS, frame time, mode comparisons per platform/GPU.
“How does it perform?”
View method →
Patch Retest
Retest policy: critical families within 7 days, others within 14 days of a new GameVersion.
“What changed in this patch?”
View method →
Retest hook for M13
Every Result is version-pinned. A new GameVersion creates a new Result row; VersionChange diffs the two (absolute + percent). Patch diff pages are generated from VersionChange, not by diffing copy. See
computePatchDiff in src/lib/testing.ts.