Agent, model and search profiles in public evidence
The exact execution-profile mixture behind the current Standard aggregate—without pretending historical matches all used one latest model.
| Player 1 profile | Player 2 profile | Search iterations | Games | Share | Date range |
|---|---|---|---|---|---|
ALPHAZERO-v240 | ALPHAZERO-v240 | 1600 | 7,891 | 40.5% | 19 Jul 2026 – 22 Jul 2026 |
ALPHAZERO-v1606 | ALPHAZERO-v1606 | 1600 | 6,533 | 33.5% | 3 Aug 2026 – 2 Sept 2026 |
UNKNOWN | UNKNOWN | 400 | 2,313 | 11.9% | 19 Jun 2026 – 9 Jul 2026 |
ALPHAZERO-v1875 | ALPHAZERO-v1875 | 1600 | 1,730 | 8.9% | 10 Jul 2026 – 13 Jul 2026 |
UNKNOWN | UNKNOWN | 1600 | 947 | 4.9% | 9 Jul 2026 – 10 Jul 2026 |
ALPHAZERO-v798 | ALPHAZERO-v798 | 1600 | 77 | 0.4% | 23 Jul 2026 – 28 Jul 2026 |
ALPHAZERO-v555 | ALPHAZERO-v555 | 1600 | 4 | 0.0% | 22 Jul 2026 – 22 Jul 2026 |
ALPHAZERO-v1953 | ALPHAZERO-v1953 | 1600 | 4 | 0.0% | 13 Jul 2026 – 16 Jul 2026 |
Comparison boundary
This table is provenance, not a model-strength comparison. A valid comparison requires pinned architectures, checkpoints, serving settings, the same deck schedule and side-balanced direct games.