Methodology
Last updated: 2026-05-27 12:36 UTC
c8alpha is a quantitative research publication from Camp8 LLC. We combine data-driven scanners with LLM-led investigation to publish a daily ranked list of stock picks. This page describes the system, the validation discipline, and what we deliberately don't do.
How it works
c8alpha is a three-stage pipeline:
- Mechanical scanners continuously evaluate the equity and options universe against validated patterns. Scanners surface candidates; they don't make recommendations on their own.
- LLM investigation runs on every candidate. For each pick, the advisor checks for catalysts, recent news, sector context, and known risk factors, and assigns a conviction score from 0 to 100. The investigation paragraph is published alongside the pick.
- Daily ranking combines the mechanical signal with the LLM investigation into a single sortable list. Higher-conviction picks appear first. Every pick links to the audit trail showing exactly how it was produced.
The result is research that's reproducible (the same mechanical signal on the same data produces the same pick) and inspectable (the audit trail shows the investigation behind the conviction score).
Validation discipline
We don't ship a strategy because a single backtest looked good. Before any strategy contributes to the daily ranking, it has to clear three gates:
- Permutation testing. Historical returns are shuffled 10,000 times to confirm the signal beats random at a Benjamini-Hochberg-corrected p-value below 0.05.
- Walk-forward out-of-sample testing. The historical sample is split 60/40 by date. Strategies are calibrated on the first 60% and measured on the last 40%. Out-of-sample performance must hold.
- Minimum sample size. Every strategy must accumulate a minimum number of historical signals before being trusted.
After going live, strategies are continuously monitored against their validation benchmark. When forward performance diverges from expectations for long enough, strategies are demoted or retired. We don't promote narratives over numbers.
Scope
c8alpha currently runs 13 active strategies across these categories: cross-asset momentum confirmation, dealer-positioning-driven option premium selling, deep mean-reversion oversold conditions, insider open-market accumulation, institutional options flow, options flow regime shifts, and supply-chain ripple after earnings. Categories are added when a new strategy clears validation, and removed when no validated strategy remains in that category.
What we don't do
- We don't trade real money. Every position reported on c8alpha is held in a paper-trading account. No client capital is at risk because there is no client capital.
- We don't take compensation from companies we cover. No paid promotions, no sponsored picks, no affiliate relationships with brokers tied to a specific name.
- We don't personalize. The same daily ranking is shown to every reader. We don't know your portfolio, your tax situation, or your risk tolerance, and we don't pretend to.
Limitations of mechanical validation
Even a strategy that clears all three gates above is measured under specific assumptions. It does not capture:
- Real-world execution costs. Historical performance is computed on bar-to-bar fills with no slippage or commission.
- Market impact at scale. Strategies that work at small size may not at large size.
- Regime shifts the in-sample period didn't see. A strategy validated on past data may not work in a regime that hasn't happened yet.
- LLM investigation errors. The narrative paragraph attached to each pick is generated by a large language model and may contain factual errors. Verify the narrative independently before acting.
Past performance does not predict future results. All performance shown on c8alpha is from paper-trading simulations; no real money has been deployed. See the full disclaimer for the complete risk disclosure.