About Six7 Alpha

Most retail traders fail not because they lack access to good information, but because they apply good information in the wrong order. They find a stock before diagnosing the market regime. They pick a strategy before checking whether current conditions support it. They build a trade plan without accounting for how the broader sector is behaving.

We built Six7 Alpha to enforce the correct sequence: diagnose first, then strategize, then screen, then rank, then plan. Every session follows the same pipeline. The result is a research process that is consistent, explainable, and improvable — rather than a different set of impulses every morning.

The AI in our platform is not generating stock picks from thin air. It is synthesizing real quantitative data: market breadth from the McClellan Oscillator and Summation Index, volatility regime from VIX, sector rotation from relative strength rankings, and ticker screening from live Finviz and yfinance queries. The language model explains the reasoning and flags risks — it does not replace the data.

The 44+ strategies in the platform are not a curated list of names from trading textbooks. They are implemented end-to-end: each strategy has a Finviz pre-filter, a yfinance verification pass, specific entry and exit criteria, and a backtest record across a diversified 20-ticker benchmark going back five or more years.

We run a continuous research pipeline we call autoresearch — a structured experiment loop inspired by Andrej Karpathy's work on autonomous research agents. Each strategy is treated as a hypothesis. Parameters are tested, filters are evaluated, and changes are merged only when they survive both quantitative benchmarking and human review. The result is a library that improves over time rather than remaining frozen at its initial implementation.

We maintain a public changelog of every significant parameter or logic change to any strategy. If a filter was removed because it was silently blocking valid signals, that is documented. If a stop was widened because the prior setting was unrealistic for volatile names, that is documented. We think that transparency is essential to taking the platform seriously as a research tool.