Applying Karpathy's autoresearch pattern to trading. We scout community strategies, enrich them with 39 signal series, and validate everything with walk-forward testing. No gut feelings. No overfitting. Just data.
Real-time market regime, actionable calls per asset, and signal intelligence. Updated continuously.
A systematic pipeline from raw community strategies to validated, signal-enriched trading research.
Crawl TradingView's public strategy library. Parse Pine Script, extract logic, normalize parameters. Currently tracking 254 community strategies across multiple asset classes and timeframes.
Layer 39 external signals — macro indicators, sentiment data, on-chain metrics, sector rotations, and CoinGecko feeds — onto each strategy. Test which combinations improve Sharpe ratio and reduce drawdown.
Every strategy-signal combination runs through walk-forward validation. No peeking at future data. No curve-fitting. If it doesn't hold out-of-sample, it doesn't ship.
Each strategy is tested with and without these signal overlays to find genuine edge — not noise.
Fed funds rate, CPI, PMI, yield curves, unemployment, GDP, and more. The big picture that moves everything.
Fear & Greed index, put/call ratios, VIX derivatives, social sentiment scores. What the crowd is feeling.
Exchange flows, whale wallets, MVRV, SOPR, hash rate, active addresses. The blockchain doesn't lie.
Sector ETF relative strength, rotation models, defensive vs cyclical ratios. Where money is flowing.
Market cap dominance, DeFi TVL, trending tokens, volume anomalies. Crypto-native intelligence.
Systematic research means letting compute do the heavy lifting while humans define the constraints.
Every component is chosen for reliability, reproducibility, and speed.
Core research engine in Python. Strategy logic parsed from Pine Script for faithful reproduction.
Rolling window validation prevents overfitting. Train/test splits that respect time series causality.
Strategy metadata in SQLite. Time series data in Parquet for fast columnar queries.
Automated ingestion from FRED, CoinGecko, alternative.me, and custom scrapers. Daily refresh.
Interactive dashboard for exploring strategy performance, signal correlations, and validation results.
LLM-powered strategy summarization, anomaly detection, and research report generation.
See exactly what our system sees — market regime, per-asset BUY/SELL/HOLD calls with reasoning, 46 signal readings, and watchlist items. Plain English, not quant jargon. Free while we're building. Premium tiers coming soon.
View Live Signals & Calls →Building in layers. Each phase adds capability while maintaining research rigor.
Crawl TradingView, parse Pine Script, normalize and catalog 254+ community strategies.
● LIVE39 signal series from macro, sentiment, on-chain, sectors, and CoinGecko. Daily automated ingestion.
● LIVERolling-window backtests for every strategy × signal combination. Out-of-sample performance tracking.
◐ BUILDINGAI-powered meta-analysis layer. Pattern recognition across strategy performance and market regimes.
○ PLANNEDTop-performing strategies deployed to Alpaca paper trading. Real-time validation without real capital.
○ PLANNEDGraduated strategies with proven paper-trading track records deployed with real capital. Position sizing, risk limits, circuit breakers.
○ PLANNED