sn28-miner-richtao

Model Card for raulcel/sn28-miner-richtao

This model predicts short-term movements of the S&P 500 Index (SPX) for use in the Bittensor Subnet 28 – S&P 500 Oracle.
It integrates macroeconomic indicators, futures data, and optional news sentiment to forecast the spot price 5–30 minutes ahead.

---

Model Details

Model Description

• Developed by: Raúl Cel
• Funded by: Self-funded research project under the Bittensor ecosystem
• Shared by: Raúl Cel (@raulcel on Hugging Face)
• Model type: Linear / tabular regression (macro-financial)
• Language(s): English (numeric-based model)
• License: MIT
• Finetuned from: Nullnotfound/Auren-plus-E4B-it (used as a reference placeholder)

Model Sources

• Repository: https://huggingface.co/raulcel/sn28-miner-richtao
• Dataset: https://huggingface.co/datasets/raulcel/sn28-miner-richtao-dataset
• Demo: Internal deployment in Bittensor SN28 network

---

Uses

Direct Use

Predicts the expected S&P 500 spot value (Yahoo Finance ^GSPC) up to 30 minutes ahead.
Used directly by a miner node in the Bittensor SN28 subnet to respond to validator challenges.

Downstream Use

Can be reused for:

• Intraday trading simulation and risk monitoring
• Short-term macro-sentiment analysis
• Ensemble financial prediction systems

Out-of-Scope Use

Not suitable for:

• High-frequency trading execution
• Predicting individual stock prices
• Long-term portfolio forecasting

---

Bias, Risks, and Limitations

Predictions depend on live Yahoo Finance feeds. Sudden data outages or macro releases (e.g., CPI, NFP, FOMC) can distort results.
The model assumes normal market conditions; it may underperform in low-liquidity or event-driven regimes.

Recommendations

Monitor RMSE/MAE metrics regularly (auto-computed in metrics.json).
Recalibrate weights if RMSE > 25 pts for more than 50 samples.

---

How to Get Started with the Model

from huggingface_hub import hf_hub_download
import json
weights_path = hf_hub_download("raulcel/sn28-miner-richtao", "model/weights.json")
weights = json.load(open(weights_path))
print(weights)