Backtesting Archive
Offline evaluation snapshots for fast model benchmarking without running live challenges.
The archive provides quarterly snapshots of past challenges, including input context, ground truth, and the pre-registered forecasts from all participating models. This lets you evaluate a new model offline and report results in a paper, e.g. "evaluated on TS-Arena Archive Q1 2026".
Open on Hugging FaceUsage Guidelines
Temporal split. Use a training cutoff strictly before the start of the evaluation period to prevent data leakage.
Self-reported results. Clearly state that results are based on the TS-Arena Archive and are not official rankings from the TS-Arena platform.
Official leaderboard. Inclusion in the live rankings and future archive dumps requires participation in challenges at ts-arena.live.
Citation
@inproceedings{meyer2026tsarena,
title={TS-Arena -- A Live Forecast Pre-Registration Platform},
author={Marcel Meyer and Sascha Kaltenpoth and Henrik Albers and Kevin Zalipski and Oliver Müller},
booktitle={Proceedings of the 32nd ACM SIGKDD Conference on Knowledge Discovery and Data Mining V.2},
series={KDD '26},
pages={9558--9568},
year={2026},
publisher={Association for Computing Machinery},
doi={10.1145/3770855.3817515},
url={https://doi.org/10.1145/3770855.3817515},
}