All Projects

Project Overview Page

This page documents how the system works and shows live sample data pulled from it. The full trading dashboard — with portfolio, signals, and controls — lives at a separate URL.

View Full Dashboard

Swing Trading Analytics Dashboard

A professional-grade automated stock analysis system combining technical indicators, AI forecasting, and sentiment analysis to generate actionable swing trading signals.

Core logic lives in a private repository.

CatBoost LightGBM Time Series Plotly FastAPI

Strategy Rules

Market filter: only trade when SPY > moving average
Entry: RSI oversold + MACD bullish + Volume spike + Positive sentiment
Trailing stop: ATR-based, triggered when price retreats from peak
Hard stop: fixed drawdown limit from entry
Take profit: fixed gain target from entry
68%
Mean Directional Accuracy
Low 43% · High 97%
across 138 tickers
3
Model Ensemble
CatBoost · LightGBM · Theta
Daily Scan Cadence
30d
Forecast Horizon

Alpha Intelligence (Watchlist)

Click any ticker for deep-dive technical charts.

Latest Scan Cycle

Apr 18, 2026 07:51 PM

Top Quant Opportunities

Sieve: Total P/L > 30% (+ Core Assets)
Ticker Price Score Signal & Insights Backtest Performance
NFLX Fixed ⚠️
$97.31
62
No Signal
Awaiting next algorithmic scan...
25.0% Win Rate / 8 Trades
SIGNAL GROWTH: -13.1%
0.87x
AAON Fixed
$93.98
56
No Signal
Awaiting next algorithmic scan...
80.0% Win Rate / 5 Trades
SIGNAL GROWTH: +90.9%
1.91x

Why This Is Hard

Stock prices approximate a random walk. Most ML models overfit on historical patterns and collapse out-of-sample. This system guards against that with a market-regime filter, strict multi-condition signal gating, and walk-forward validated backtests — so signals reflect real edge, not curve-fitting.

Signal Gate

A buy signal fires only when a proprietary set of technical, volume, and sentiment parameters simultaneously agree. This strict multi-condition gate filters the vast majority of candidate setups — keeping only high-conviction entries where several independent signals converge.

Production Infrastructure

Fully automated pipeline: a task scheduler fires scans 4× daily, a message broker queues the jobs, and live prices plus news headlines are ingested automatically — no manual intervention. When a signal fires, I receive an email alert and act on it.

Rigorous Model Evaluation

Each model's directional accuracy is tracked independently on out-of-sample data. The best-performing model per ticker is highlighted in the charts below — the ensemble isn't blind averaging, it's informed selection backed by live hit-rate tracking.

How Score Is Calculated

The Score combines Signal Strength, Model Confidence, and Win Rate as a weighted value. Higher scores indicate higher conviction signals where multiple components align — tickers with scores 70+ represent the most promising entry opportunities.

Watchlist Filter (Sieve)

Tickers in the Alpha Intelligence table are filtered to show only those with historical backtest performance above 30% total P&L, plus core portfolio holdings. This ensures you're focusing on setups with proven statistical edge and established trade history.