TastyAlgo

Regime-aware algorithmic trading dashboard that actually works (most of the time)

what this does

Built this to experiment with different trading strategies and see how they perform across various market conditions. turns out market regime matters way more than i initially thought.

• fetches real market data from yahoo finance
• detects market regimes using statistical methods or k-means clustering
• tests 4 different trading strategies with proper backtesting
• shows performance breakdown by market regime
• interactive charts and csv export for deeper analysis

strategies included

MA crossover - classic trend following using moving averages. buy when short MA crosses above long MA, sell when it crosses below. works great in trends, gets chopped up in sideways markets.

momentum - buys the strong stuff, sells the weak stuff. uses n-day returns vs configurable thresholds. can work really well in trending environments but watch out for nasty reversals.

vol breakout - trades when volatility spikes above historical norms. idea is that high vol often precedes big moves. catches some good ones but also plenty of false signals.

pairs trading - market neutral approach that trades mean reversion between correlated stocks. uses cointegration analysis to find statistical arbitrage opportunities.

quick setup

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

project structure

TastyAlgo/
├── .streamlit/config.toml     # for configuring defualt streamlit theme
├── app.py                     # main streamlit dashboard
├── data_fetcher.py            # yahoo finance data handler
├── regime_detection.py        # market regime classification  
├── backtesting.py            # trade execution and portfolio tracking
├── metrics.py                # performance calculations
├── visualizations.py         # plotly chart generation
├── utils.py                  # helper functions and utilities
└── strategies/               # individual strategy implementations
    ├── moving_average.py
    ├── momentum.py
    ├── volatility_breakout.py
    └── pairs_trading.py

key features

• real market data - uses yahoo finance, not synthetic/fake data
• regime detection - statistical and machine learning approaches to classify market states
• proper backtesting - includes commissions, proper trade execution, realistic assumptions
• comprehensive metrics - sharpe ratio, max drawdown, win rate, etc. all split by regime
• interactive charts - plotly-powered visualizations with hover info and zooming
• csv export - download detailed results for further analysis

how regime detection works

statistical method - uses rolling returns and volatility thresholds to classify market states into bull/bear/sideways

k-means clustering - groups similar market conditions automatically using multiple features (returns, vol, rsi, volume)

the regime stuff actually helps a lot - most strategies perform very differently depending on market conditions.

performance metrics

calculates the usual suspects:

• total and annualized returns
• sharpe ratio for risk-adjusted returns
• maximum drawdown (worst peak-to-trough decline)
• win rate (percentage of profitable trades)
• volatility and other risk measures
• all metrics broken down by market regime

also compares against simple buy & hold benchmark.

data notes

uses yahoo finance via the yfinance package. data gets cached for 1 hour to avoid hitting rate limits.

for educational purposes only - dont use this for actual trading without doing your own research and testing.

known issues

• sometimes yahoo finance data is flaky or missing
• pairs trading strategy could use more sophisticated hedge ratio calculation
• regime detection can be a bit noisy during transition periods
• commission model is simplified
• probably some other bugs i haven't found yet

possible extensions

could add:

• more sophisticated regime detection (maybe using ML)
• portfolio optimization across multiple assets
• real-time data feeds
• more trading strategies (mean reversion, breakout, etc)
• walk-forward analysis and parameter optimization
• better transaction cost modeling
• risk management overlays

disclaimer

this is for educational and research purposes only. Do your own research if want to use it for real trading.

Built this while learning about quantitative finance - definitely learned a lot in the process. probably has some quirks but seems to work reasonably well for what it is.