Every performance claim is an argument. Here's how to judge one.
This page isn't about any particular competitor — it's about practices. The retail trading industry runs on a handful of presentation tricks that make noise look like edge. Once you can see them, you can't unsee them. We hold ourselves to the right-hand column, and you should hold us to it too.
| The claim | What you usually get | What evidence looks like |
|---|---|---|
| "90%+ win rate" | A win rate with no trade count, no average win/loss sizes, and no baseline. High win rates are often mechanical — a property of the trade's shape, not of any skill. | The full distribution: every closed pick shown, losers included, with n, average win, average loss, and the answer to "how often would random produce this?" (We killed our own 97.5%-win-rate strategy. The write-up is public.) |
| "Backtested and verified" | One number from one backtest — no out-of-sample split, no permutation test, no disclosure of how many variants were tested to find it, price data too coarse to see the stops. | A 10,000-shuffle permutation test with multiple-comparisons correction, a walk-forward out-of-sample split, a minimum sample floor, and an exit sweep — before a strategy produces its first pick. The full gauntlet is published. |
| "Look at these winners" | A highlight reel. The losers are cropped out; the strategies that died are deleted; the account that went bad quietly disappears. | The record is append-only. Killed strategies stay published with the reason they died — 24 so far, against 6 that survive. Picks that lost stay in the track record forever. |
| "Trust me" | Credentials, screenshots, testimonials, urgency. Nothing you can independently check. | A published methodology, a per-pick audit trail, every pick cryptographically timestamped at publication, and real entry and exit prices in the track record. You're not asked to trust the author — you're given the means to check. |
| "Up 400% this year" | A single spectacular figure, usually unverifiable, usually cherry-picking its own start date, silent about risk, sample size, and what happens in a down market. | Per-pick, short-horizon numbers with denominators, shown as a distribution rather than a headline average. Small, real, and checkable beats large and unfalsifiable. |
Five questions that expose almost everything
- What's the denominator? Any statistic without its n is decoration. A profit factor on 16 trades is an anecdote. (We wrote up the best small-n number we ever refused to trade.)
- Was it tested out-of-sample? A strategy tuned and measured on the same data will always look brilliant. Ask where the holdout was.
- How often would random do this? If the seller has never run a permutation test, they don't know whether their edge exists. Most "edges" don't survive one.
- Where are the losers? A record with no losses visible is not a track record; it's an ad.
- What happened to the failures? Everyone who tests honestly kills most of what they build. If there's no graveyard, there was no testing.
Our answers, on the record
6 live plays with backtested, walk-forward out-of-sample alpha. Every published pick carries its play, its conviction score, its reasoning, and a cryptographic timestamp; the track record carries real entry and exit prices — for options picks, the exact contract and the premium at recommendation. Every play carries its validation evidence, and every kill carries its cause of death (24 so far, with far more ideas dead in research before ever becoming plays). Read the methodology, browse the research teardowns, or start with what a play is.
And the standing disclosure, always: c8alpha is a research publication, not investment advice — the full disclaimer.