# Chess Move Prediction on Lichess Puzzle Database (held-out positions)

Source: [Chess Move Prediction on Lichess Puzzle Database (held-out positions)](<https://www.sota2.com/research/sota/chess-move-prediction-on-lichess-puzzle-database-held-out-positions>)

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

Data updated: 2026-05-14T06:55:00.603Z

Evaluation of model performance on predicting optimal chess moves using held-out positions from the Lichess Puzzle Database.

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 | Chess Move Prediction |
| Dataset (including any stated split) | Lichess Puzzle Database (held-out positions) |
| Metric | Top-1 Accuracy |
| Unit | PERCENT |
| Direction | Higher is better |
| Results loaded | 6 of 6 |
| 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

- [Top-1 Accuracy](<https://www.sota2.com/research/sota/chess-move-prediction-on-lichess-puzzle-database-held-out-positions.md?metric=Top-1%20Accuracy>): Higher is better; unit: PERCENT.

- [Elo Rating](<https://www.sota2.com/research/sota/chess-move-prediction-on-lichess-puzzle-database-held-out-positions.md?metric=Elo%20Rating>): Higher is better; unit: SCALAR.

- [Coverage](<https://www.sota2.com/research/sota/chess-move-prediction-on-lichess-puzzle-database-held-out-positions.md?metric=Coverage>): Higher is better; unit: PERCENT.

## Recorded results

| Method | Top-1 Accuracy | Result date | Report paper |
| --- | --- | --- | --- |
| VPS | 54.5 | Not recorded | 6a05570cdbc836b87d828c09 |
| SFT + GRPO | 54.1 | Not recorded | 6a05570cdbc836b87d828c09 |
| VPS | 53.6 | Not recorded | 6a05570cdbc836b87d828c09 |
| SFT + GRPO | 53.4 | Not recorded | 6a05570cdbc836b87d828c09 |
| SFT only | 21.5 | Not recorded | 6a05570cdbc836b87d828c09 |
| SFT only | 21.2 | Not recorded | 6a05570cdbc836b87d828c09 |

### VPS

Top-1 Accuracy: 54.5. Result ID: 6a05701347f34ce924ebd036.

[Reporting paper](<https://www.sota2.com/research/paper/6a05570cdbc836b87d828c09.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a05570cdbc836b87d828c09.md>)

Recorded evaluation configuration: {"Backbone":"DeepSeek-R1-Distill-Llama-8B"}. This record does not assert independent reproduction.

### SFT + GRPO

Top-1 Accuracy: 54.1. Result ID: 6a05701347f34ce924ebd035.

[Reporting paper](<https://www.sota2.com/research/paper/6a05570cdbc836b87d828c09.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76a91e209131cea6e3eb.md>)

Recorded evaluation configuration: {"Backbone":"DeepSeek-R1-Distill-Llama-8B"}. This record does not assert independent reproduction.

### VPS

Top-1 Accuracy: 53.6. Result ID: 6a05701347f34ce924ebd033.

[Reporting paper](<https://www.sota2.com/research/paper/6a05570cdbc836b87d828c09.md>)

[Method origin paper](<https://www.sota2.com/research/paper/6a05570cdbc836b87d828c09.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-8B"}. This record does not assert independent reproduction.

### SFT + GRPO

Top-1 Accuracy: 53.4. Result ID: 6a05701347f34ce924ebd032.

[Reporting paper](<https://www.sota2.com/research/paper/6a05570cdbc836b87d828c09.md>)

[Method origin paper](<https://www.sota2.com/research/paper/68fa76a91e209131cea6e3eb.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-8B"}. This record does not assert independent reproduction.

### SFT only

Top-1 Accuracy: 21.5. Result ID: 6a05701347f34ce924ebd034.

[Reporting paper](<https://www.sota2.com/research/paper/6a05570cdbc836b87d828c09.md>)

Recorded evaluation configuration: {"Backbone":"DeepSeek-R1-Distill-Llama-8B"}. This record does not assert independent reproduction.

### SFT only

Top-1 Accuracy: 21.2. Result ID: 6a05701347f34ce924ebd031.

[Reporting paper](<https://www.sota2.com/research/paper/6a05570cdbc836b87d828c09.md>)

Recorded evaluation configuration: {"Backbone":"Qwen3-8B"}. This record does not assert independent reproduction.

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