RED-2400: rejection events of an automated trading agent on Solana DEXs, with post-rejection samples

Canonical source and DOI: this is a mirror of the data files of RED-2400 on Zenodo. Please cite the concept DOI 10.5281/zenodo.19989074. The data files are byte-identical to every Zenodo version (v1 to v4). The documentation below includes corrections made in September 2026.

Data descriptor: Kamat, A. U. (2026). RED-2400: A Public Benchmark of Algorithmically-Rejected Trading Events with Outcome Labels. arXiv:2605.12151. Its Section V matched comparison is withdrawn (see below).

Author: Arati Uday Kamat (ORCID 0009-0000-4781-312X) · License: CC BY 4.0

What it is

A public registry of the decisions an automated trading agent made to reject newly launched tokens on Solana decentralised exchanges, together with the tokens' market data after each rejection. It supports research on auditing screening decisions: what happened to the candidates a rule turned down.

File	Rows	Contents	SHA-256
rejections.csv	6,660	One row per rejection: timestamp, source (source_a / source_b), mint, symbol, reason (anonymised filter_1 to filter_8), timeSlot	ae773c88da68c95b1e7d9ba5adda0c41bafdd123e776d5a754e896771c373a66
rejection_outcomes.csv	169,123	Post-rejection samples: sampleTs, mint, symbol, rejectReason, rejectTs, ageMin, priceUsd, liquidity, volume24h, dexId, pairAddress, source	50181e433903ee72814b456dfd7d2ada10d20e740fbc8ba931c55225f7f1d204
graveyard_lifecycle.csv	1,837	Lifecycle-state transitions of rejected tokens: ts, mint, symbol, from, to, liquidity, ageDays	28fc461d1952f6654f52e9bb7233e7ea6f4faf1797b62724ea9fc4707ab95960

The rejections span 2026-04-10T21:10Z to 2026-05-02T21:48Z (22 days) and cover 1,076 tokens.

How the post-rejection samples were collected (read before use)

• Cadence: a tracker sampled each tracked token every 10 minutes. 99.51% of the gaps between samples of a token are 9.5 to 10.5 minutes.
• Start: the tracker started at 2026-04-11T20:26:53Z, about 23 hours after the first rejection. Most earlier rejections have no samples.
• One record per token: a new rejection of a tracked token restarted its record, so an event's samples usually stop at the token's next rejection. The label in rejectReason can carry an earlier rejection's filter.
• Horizon: sampling ends 1,440 minutes after the rejection.
• ageMin is minutes since rejectTs (0 to 1,440), not the token's age.
• liquidity and volume24h are USD values quantised to powers of two.
• Empty fields: 916 rows of rejection_outcomes.csv have an empty sampleTs; 348 of them also lack rejectReason and rejectTs.

Do not read single-sample events, or an event's last sample, as token death or disappearance. They mostly reflect a re-rejection of the same token. Half of the rejection events (3,333 of 6,660) match an outcome record exactly. Most of the others were re-rejected within minutes, were logged before the tracker started, or carry an earlier rejection's label.

The lifecycle file

• A separate lifecycle tracker recorded state changes from 2026-04-23T18:07Z to 2026-05-02T16:53Z.
• to takes the values alive_active, alive_dormant and gone. from is empty on a token's first observation.
• ageDays is days since the tracker first observed the token (0 to 8.7), not days since the rejection.
• gone is a tracker state recorded with liquidity 0. It is not terminal: 167 of the 377 tokens that reach it later return to an alive state.

Withdrawn

The earlier matched comparison of "early-death" and other tokens reaching gone (48.9% vs 57.6%) is withdrawn. Its "early-death" label marks events whose only sample was followed within 10.5 minutes by a re-rejection of the same token (1,235 of 1,236).

Loading

from datasets import load_dataset

rejections = load_dataset("AratiKamat/RED-2400", "rejections", split="train")
outcomes = load_dataset("AratiKamat/RED-2400", "outcomes", split="train")
lifecycle = load_dataset("AratiKamat/RED-2400", "lifecycle", split="train")

How to cite

Kamat, A. U. (2026). RED-2400: A public benchmark of algorithmically-rejected trading events with outcome labels [Data set]. Zenodo. https://doi.org/10.5281/zenodo.19989074

@dataset{kamat_red2400_2026,
  author    = {Kamat, Arati Uday},
  title     = {{RED-2400}: A Public Benchmark of Algorithmically-Rejected
               Trading Events with Outcome Labels},
  year      = {2026},
  publisher = {Zenodo},
  doi       = {10.5281/zenodo.19989074},
  url       = {https://doi.org/10.5281/zenodo.19989074}
}