Datasets:

mospira
/

open30-equity-features

license: mit
pretty_name: Open30 30-Minute Open Equity Features Dataset
task_categories:
  - tabular-classification
  - tabular-regression
tags:
  - finance
  - trading
  - equities
  - intraday
  - tabular
  - time-series
  - xgboost
  - walk-forward-validation
size_categories:
  - 100K<n<1M

Open30 30-Minute Open Equity Features Dataset

Assembled modeling table for Open30, a research project studying short-horizon equity behavior after the U.S. market open.

The rows are candidate trade instances keyed by (date, ticker, side). Features include prior daily price context, volatility and liquidity proxies, opening-minute behavior, market alignment features, calendar features, mean-reversion regime features, and Alpha Vantage news sentiment aggregates. The table also includes supervised outcome labels for multiple reward/risk targets.

Author GitHub: mospira
Project repo: mospira/ml-open30
Project site: Open30 Research
Research report: Open30 Research Report

Files

dataset_open30m.parquet

Dataset Details

Rows: 195,166
Columns: 77
Date range: 2010-04-28 through 2026-02-27
Universe size: 25 tickers
Candidate sides: long, short
Entry assumption: 09:31 ET
Label scan window: 09:31 through 09:59 ET
Reward/risk multiples: 0.5, 1.0, 1.5, 2.0

Column Groups

Identifier columns:

date
ticker
side

Feature groups include:

prior daily price context
volatility regime
liquidity proxies
first-minute open-window features
SPY/QQQ market-alignment features
calendar features
mean-reversion regime features
news sentiment aggregates and interactions

Label columns follow this pattern:

y_type_m_<multiple>
y_R_m_<multiple>
y_hit_minute_m_<multiple>
y_ambig_m_<multiple>

Outcome encoding:

0: stop loss
1: take profit
2: time exit
3: ambiguous same-bar stop/target touch

Usage

from datasets import load_dataset

ds = load_dataset("mospira/open30-equity-features", split="train")
df = ds.to_pandas()
print(df.head())