Third Eye Trading
Equity research terminal and proprietary pattern-detection engine — 130+ endpoint FastAPI backend, an automated market-data pipeline, and a published track record
- published, auditable track record
- proprietary pattern engine
- subscriber research product
Overview
A subscriber research product built around a proprietary edge. The terminal covers ~290 symbols with 15 data tabs each — fundamentals, valuation, 10-year financials, ratio trends, estimates, dividends, momentum, and ownership — fed by an automated Seeking Alpha ingestion pipeline with a four-stage fallback chain. On top of it sits the Wick-Wick model, a candlestick pattern detector for monthly and quarterly liquidity-purge setups, backtested across 786 enriched patterns from 2020–2025 and published as an auditable track record. Rounded out with a pattern scanner, screener, backtesting engine, options-surface capture, portfolio and play tracking, a CFA learning hub, and a tiered subscriber funnel with admin CRM.
The challenge: Retail investors get news and opinion, not an edge with its results shown. Third Eye pairs a research terminal with a proprietary pattern model — and publishes every play, including the ones that did not work.
Quantitative Modeling
Pattern detection over monthly and quarterly candles with parameterized range thresholds, fundamental quality enrichment, validation against known ground-truth plays, and a backtest harness reporting hit rates per target and horizon.
Resilient Data Ingestion
Authenticated Playwright scraping with cookie persistence on a 24-hour TTL, resumable multi-hour batch runs with a per-symbol audit trail, and a four-source fallback chain so no single upstream outage takes the terminal down.
Python API Design
FastAPI with Pydantic models across 20 route modules covering stock data, scanning, backtesting, portfolio, plays, newsletter, knowledge base, billing, and admin — deployed to Render, with the Vite frontend on Vercel.
Honest Analytics Presentation
Track-record surfaces where every claim carries its own window and denominator, built through repeated passes specifically to remove overstated tiers, redundant metrics, and framing that flattered the results.