title	BTC Forecast
emoji	📈
colorFrom	blue
colorTo	indigo
sdk	streamlit
sdk_version	1.33.0
app_file	app.py
pinned	false

BTC Forecast — AlphaI × Polaris Challenge

Predicts the 95% confidence price range for BTC/USDT one hour ahead using GBM with Student-t fat tails and rolling volatility clustering.

File guide

File	Where it runs	What it does
model.py	everywhere	Core GBM model — shared by all other files
backtest.py	Colab / local	Part A — 30-day walk-forward backtest
app.py	Streamlit Cloud	Parts B & C — live dashboard + history
BTC_Forecast_Colab.ipynb	Google Colab	Ready-to-run notebook for Part A
backtest_results.jsonl	generated	Output of backtest — commit to repo
requirements.txt	Streamlit Cloud	Python dependencies

How to run

Part A — Backtest (Google Colab)

1. Upload BTC_Forecast_Colab.ipynb to Colab
2. Change YOUR_USERNAME in Cell 2 to your GitHub username
3. Run all cells (takes ~2 mins)
4. Copy coverage_95 and mean_winkler_95 for the submission form
5. Download backtest_results.jsonl and upload it to this repo

Parts B & C — Dashboard (Streamlit Cloud)

1. Push all files to a public GitHub repo
2. Go to share.streamlit.io → Create app
3. Select this repo, branch main, file app.py
4. Deploy → get your public URL

Local development

pip install -r requirements.txt
streamlit run app.py