Rionda Capital runs one strategy in production, GC8, and one in research, R34. What follows is how a strategy gets here, how it is run once it does, and what stops it.
Most trading failures are not failures of insight. They are failures of adherence — a position held past its exit because the story still sounded good, a limit widened once for a special case that turned out not to be special. The advantage in a systematic approach is not that the rules are smarter than a person. It is that they are the same on the worst day as on the best one.
That places the burden on the research rather than the trading. A rule we cannot defend before it runs is a rule we will not be able to defend while it is losing money. So we spend the effort up front, in validation, and accept that most of what we test does not survive it.
None of which is an argument for standing still. We are not trying to build a large firm; we are trying to build an excellent one, and we treat the technology as the product rather than the plumbing. Better data, faster execution, and a tighter match between what the specification says and what the market actually returns — each of those compounds, and none of them is ever finished.
The specific mechanics of GC8 — how positions are selected, sized and timed — are proprietary and are not published. Prospective investors receive a fuller methodology discussion under confidentiality.
It comes from De La Rionda — of the river. We chose it deliberately. Investing is, in the end, a long-term exercise in judgment, stewardship and reputation, and a family name is not something you put on a door lightly.
The capital entrusted to us represents generations of work by the people who commit it. We approach it accordingly. That is also why the rules on this page are written down and enforced in code rather than held in someone's head: a standard you can be talked out of is not a standard.
Discretion is removed from the investment process, not managed within it. Every position is sized, entered and exited by a specification written in advance. Anything that has to be decided later is a gap in the specification, not flexibility — and a rule bent once becomes a rule that no longer exists.
Nothing trades until it has survived testing on periods the research never saw. A result that only holds on the data used to build it is a reason to stop, not a reason to launch. Most candidates end there, and that is the process working rather than failing.
An investment system is an engineered system, and engineered systems are judged by how they behave when something fails — a missed fill, a stale price, a broker that does not answer. We spend more time on the failure paths than on the narrative.
Position size, concentration and total exposure are limits enforced in code and checked on every order. They do not widen because an opportunity looks compelling. A constraint that can be argued with is not a constraint.
Established, widely-held US listed equities. We do not trade instruments whose exit depends on finding a willing counterparty in a stressed market.
Positions are taken on both sides of the market. Returns have not depended on the index rising, which is what a 0.34 correlation to the S&P 500 describes.
Per-position size and total gross exposure are fixed in code and enforced on every order. Neither is discretionary, and neither has an override path.
No illiquid or private holdings, no leverage beyond the stated exposure limit, and no discretionary positions outside the specification.
The five-year record rests on tens of thousands of individual positions. No single holding is material to the result, which is what makes the average meaningful.
Returns have been largely independent of market direction over the period — including 2022, when the index fell 19.9% and the strategy returned 102.1%.
Return per unit of volatility over five years, against a maximum peak-to-trough decline of 21.0%. The full risk profile is on the performance page.
All figures are simulated and derived from historical backtesting over the five years ended August 2026. Simulated performance carries inherent limitations described in full on the performance page and in the footer below.