Super Smash Bros. Melee Ranked Replays for Machine Learning

Anonymized Slippi replay files from Super Smash Bros. Melee ranked matches at platinum rank and above. The dataset is sharded by character and rank pair and was built for behavior-cloning and replay-driven machine learning projects. It was created by author erickfm and last updated on 2026-04-15.

Use Cases

• Train behavior-cloning agents based on anonymized replay data from high-level players.
• Analyze character-specific gameplay strategies based on the sharding by character and rank.
• Develop and benchmark replay-driven machine learning models, such as the MIMIC project mentioned in the description.

Strengths

• Data is sharded by 25 distinct character and rank pairs, facilitating targeted analysis.
• Replays are sourced from platinum rank and above, indicating a focus on high-level competitive play.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count and total dataset size are unknown, which may limit suitability assessment.

Provenance

Source

huggingface

Collection Method

Anonymized Slippi ranked replays, likely collected from online play.

Freshness

Last updated 2026-04-15 23:57:30; freshness should be verified.