Red Weather provides SP Story, a multimodal dataset that pairs 90+ years of S&P 500 price data and visual graphs with crowdsourced free‑text narratives, binary market direction predictions, and respondent demographics. Data scientists can integrate this human‑derived context into financial forecasting models to enhance prediction accuracy.
Funding
Funding not disclosed
Founders
Product
Problem
Predicting stock market movements relies primarily on quantitative price data, which often lacks contextual insight into how humans interpret market trends. This limits the ability of AI/ML models to capture narrative-driven signals that could improve forecasting accuracy.
Solution
Red Weather offers SP Story, a curated dataset that combines historical S&P 500 price data, visual price‑movement graphs, free‑text narratives, and respondent demographics. The narratives are collected via a uniform, repeatable crowdsourcing process from over 2,100 participants across all U.S. states, each screened for English proficiency. Each story is linked to the corresponding price history and a binary up/down prediction, providing a rich, multimodal source of human‑derived market insight. Data scientists can use the integrated files—price CSV, graph images, story and prediction CSV, and demographic data—to train or augment predictive models, either standalone or alongside other economic and news data. A comprehensive User’s Guide details the dataset structure, enabling straightforward incorporation into existing pipelines.
Target Audience
Primary customers are data scientists and AI/ML model developers focused on financial forecasting, including quantitative research teams in hedge funds, investment banks, and fintech firms, as well as academic researchers studying market behavior.
Features
- Daily S&P 500 closing price CSV covering over 90 years (23,594 trading days)
- 50 JPG price‑movement graphs representing random periods of 26–101 trading days each
- CSV linking each graph to its full price history for easy data alignment
- Survey results CSV with 2,156 rows containing free‑text stories (≈45 words), up/down predictions, and respondent demographics (age, state, gender, education)
- Uniform, repeatable crowdsourcing methodology ensuring diverse, non‑PII participant data
- No personally identifiable information; all participants assigned randomized identifiers
- Included User’s Guide describing each file and field for rapid onboarding