Stock picks with their receipts.
Mechanical scanners that clear permutation testing, walk-forward out-of-sample, and minimum sample size before they ever produce a pick. LLM investigation on every signal. Every recommendation ships with its code.
How we know it's not noise.
Most stock picks are pattern-matching dressed as research. We validate every strategy against four gates before any signal contributes to the daily ranking, and we keep measuring them after they go live.
Permutation testing
Historical returns are shuffled 10,000 times to confirm the signal beats random. The strategy has to clear a Benjamini-Hochberg-corrected p-value below 0.05 — a real significance bar, not a number we made up.
Walk-forward out-of-sample
The historical sample is split 60/40 by date. The strategy is calibrated on the first 60% and measured on the last 40%. Out-of-sample profit factor must exceed 1.0 and out-of-sample win rate must exceed 50%.
Minimum sample size
Strategies have to accumulate real evidence before they're trusted with capital. 200 closed trades to graduate from playground tier; 500 to reach paper tier. Small-sample backtests don't ship.
Forward drift monitoring
After going live, every strategy is continuously measured against the validation benchmark. When forward performance diverges from expectations for long enough, the strategy is demoted or retired — and its historical picks stay in the track record.
Mechanical first. AI second. Audit trail always.
Three stages, in order. The mechanical signal has already cleared the gates above; the LLM investigates context the signal can't see; the daily ranking is what you get.
Mechanical scanners
Continuously evaluate the equity and options universe against the validated patterns above. Scanners surface candidates; they don't make recommendations on their own.
LLM investigation
For every candidate, the advisor checks for catalysts, recent news, sector context, and known risk factors. Outputs a conviction score and a written thesis — published alongside the pick.
Daily ranking
The mechanical signal and the LLM investigation combine into a single sortable list. Higher-conviction picks appear first. Every pick links to the audit trail behind the score.
Every pick ships with its code.
We don't expect you to take "we tested it" on faith. The methodology, the audit trail, and the strategies we've killed are all public.
Audit trail per pick
Every pick links to the mechanical signal that fired and the investigation that scored it. You can verify the path from raw data to recommendation.
Published methodology
The strategies, the validation discipline, and what we deliberately don't do — all on the methodology page. No black box.
We publish the kills too
Most mechanical research is finding what doesn't work. When a strategy fails forward validation, we retire it — and don't quietly remove it from the track record.
A real pick from two weeks ago.
Every pick includes the mechanical signal, the LLM-investigated thesis, a conviction score, and the validated exit rule. Here's one we published recently.
Free to start. $290/yr for the full feed.
Three tiers. Free gets you the methodology, track record, and a look at what we publish. Standard is the full daily ranking. Pro is the advisor on demand.
What we don't do.
The negative space of the product is the trust signal.
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 goes to every reader. We don't know your portfolio, your tax situation, or your risk tolerance, and we don't pretend to.