# Human-blunder alignment on rated bullet games

Source: [Human-blunder alignment on rated bullet games](<https://www.sota2.com/research/sota/human-blunder-alignment-on-rated-bullet-games>)

Collection index: [llms.txt](<https://www.sota2.com/research/llms.txt>)

Data updated: 2026-04-01T08:42:42.294Z

Evaluation of model performance regarding human blunder alignment, measured by conditional probabilities of model blunders given human actions, and the resulting Lift.

A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage.

| Scope | Recorded value |
| --- | --- |
| Task | Human-blunder alignment |
| Dataset (including any stated split) | rated bullet games |
| Metric | P(Model Blunder \| Human Blunder) |
| Unit | PERCENT |
| Direction | Lower is better |
| Results loaded | 3 of 3 |
| Coverage | All results currently returned for this benchmark |

Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol.

## Metrics

- [P(Model Blunder \| Human Blunder)](<https://www.sota2.com/research/sota/human-blunder-alignment-on-rated-bullet-games.md?metric=P(Model%20Blunder%20%7C%20Human%20Blunder)>): Lower is better; unit: PERCENT.

- [P(Model Blunder \| Human No Blunder)](<https://www.sota2.com/research/sota/human-blunder-alignment-on-rated-bullet-games.md?metric=P(Model%20Blunder%20%7C%20Human%20No%20Blunder)>): Lower is better; unit: PERCENT.

- [Lift](<https://www.sota2.com/research/sota/human-blunder-alignment-on-rated-bullet-games.md?metric=Lift>): Higher is better; unit: SCALAR.

## Recorded results

| Method | P(Model Blunder \| Human Blunder) | Result date | Report paper |
| --- | --- | --- | --- |
| Maia-2 rapid | 43.3 | Not recorded | 69cca69e0636e5bf4afcbc11 |
| Maia-2 blitz | 44.8 | Not recorded | 69cca69e0636e5bf4afcbc11 |
| V2.2 final | 49.2 | Not recorded | 69cca69e0636e5bf4afcbc11 |

### Maia-2 rapid

P(Model Blunder \| Human Blunder): 43.3. Result ID: 69ccda4147f34ce924b69989.

[Reporting paper](<https://www.sota2.com/research/paper/69cca69e0636e5bf4afcbc11.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75751e209131cea68118.md>)

Recorded evaluation configuration: {"CPL threshold":"\> 100"}. This record does not assert independent reproduction.

### Maia-2 blitz

P(Model Blunder \| Human Blunder): 44.8. Result ID: 69ccda4147f34ce924b6998a.

[Reporting paper](<https://www.sota2.com/research/paper/69cca69e0636e5bf4afcbc11.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa75751e209131cea68118.md>)

Recorded evaluation configuration: {"CPL threshold":"\> 100"}. This record does not assert independent reproduction.

### V2.2 final

P(Model Blunder \| Human Blunder): 49.2. Result ID: 69ccda4147f34ce924b69988.

[Reporting paper](<https://www.sota2.com/research/paper/69cca69e0636e5bf4afcbc11.md>)

[Method origin paper](<https://www.sota2.com/research/paper/69cca69e0636e5bf4afcbc11.md>)

Recorded evaluation configuration: {"CPL threshold":"\> 100"}. This record does not assert independent reproduction.

Page 1 of 1. This is an index page, not the full collection.