The rubric
Robot capability is measured against a rubric published before entries open, scored by a party with no position in the outcome, with written reasoning behind every individual score. Manufacturer benchmarks are marketing because the scorer benefits from the score. A competition whose rubric is public and whose reasoning is written is evidence, because it can be argued with.
Four levels
The same scale the group's assessments use, so a competition result and a published assessment mean the same thing.
The condition is not met in any meaningful sense.
Present in principle, not at a reliability a business can plan on.
Met for most practical purposes, with a material caveat.
Met without a qualification a deployer needs to work around.
What is scored
All six conditions, including the four that are not about the machine. An arena that scored only embodiment and autonomy would be a sports event.
Embodiment
Can it physically act where the work actually is?
Reach, payload, dexterity and endurance sufficient for the task, in the environment the task happens in rather than in a demonstration cell.
Autonomy
Can it finish a task without a human in the loop at every step?
Teleoperation is a demonstration. Unattended completion, with recovery from the ordinary failures of a real environment, is a business.
Authority
Is it permitted to act, by whom, and within what scope?
Scored from what the entrant can demonstrate on the day: a recorded grant, a revocation that halts the machine, a named human. Most entrants will score zero here at first, and publishing that is the point.
Settlement
Can it be paid for work, and pay for what it consumes?
Scored from what the entrant can demonstrate on the day: a recorded grant, a revocation that halts the machine, a named human. Most entrants will score zero here at first, and publishing that is the point.
Accountability
When it damages something, who answers — by name?
Scored from what the entrant can demonstrate on the day: a recorded grant, a revocation that halts the machine, a named human. Most entrants will score zero here at first, and publishing that is the point.
Fleet
Can one operator run a hundred of them?
Unit economics live here. A machine that needs an engineer in attendance is a consultant with wheels, whatever it costs.
Every robot competition in existence measures what a machine can do. None measures whether it was allowed to, who answers for it, or whether it can be stopped. Scoring those in public is the only way the industry starts building them — and it is the reason this arena produces a benchmark rather than a highlight reel.
Standing rules
| Rule | Why |
|---|---|
| The rubric is published before entries open | A rubric written after the results is not a rubric. |
| Every score carries written reasoning | So a competitor can dispute one number rather than the whole result. |
| Scorers hold no position in entrants | Disclosed on every edition. A scorer with an interest is a judge with a stake. |
| Results are superseded, never edited | A published result stands with its date. A correction is a new publication. |
| Entrants keep their own data | The arena records scores and reasoning, not the entrant's engineering. |
The rubric derives from the six conditions, published by Botz Group. Safety requirements for entrant machines follow the industrial standard: ISO 10218.