We have released RumiCar Simulator v8.1.0. The release notes below are also available on GitHub.
v8.1.0
We measured whether a leader can defend position by blocking in a touge-style race — start and finish in different places, several cars at once (Stage AX, tabletop × precise v2 engine, measured on v8.0.0. Note the tabletop default is the dynamics model, so this is not what you get by leaving the defaults alone). The answer is no — not one of the four leader tactics raised the win rate significantly above simply taking the fastest line, in any condition (0 of 66, full sweep of 10,368 races; still 0 when the tactic’s wins are restricted to races with no contact and no close approach). What defends position is the road width, not the tactic: simply racing wins 85% on the touge courses as shipped, 85% at 4.5 car widths (flat), 100% at 3 car widths and 100% at 2 car widths — narrower than 3 car widths and it reaches 100%. Blocking the inside does not work because in a hairpin the reachable corridor narrows to a single point, so racing and inside-blocking go through the corner in the same place (the only difference is hugging one side on the straight, which leaves the other side open), and 23 of the 60 points in the inner band are unreachable for the blocker too. We also added 18 narrowed variant courses ([4.5/3/2-Car Width]) at the end of the course list, with the road width squeezed down to the ratio of a real mountain pass. The built-in sample programs struggle to finish on the narrow ones, and the course descriptions say so. The physics notes (“11. When Drift Does / Doesn’t Hold”) and the in-app design Q&A were updated too. Not one byte of the driving physics changed (record reproducibility is preserved).
Verifying this release yourself
cd Simulator
node wf_run_all.mjs # the full assertion-gate suiteThe suite runs from a fresh clone with no dependencies beyond Node.js.
The real-browser checks under Simulator/browser/ additionally need npm install and a display.