Approach
RVB builds diversified portfolios of systematic strategies. This page sets out the standards a result is held to, how an idea becomes a funded account, and which markets the firm trades today. Each section links to the page where the claim can be checked.
Beyond return
Six standards, applied at every stage.
The six below are applied throughout research, selection and execution rather than checked once at the end, and a failure against any one of them is enough to stop a result. A strategy that earns well on paper but does not survive the search that found it, the costs of trading it, or the execution it would require, is not carried into a portfolio.
Return
Measured net of costs and after a one-bar execution delay, on the weights a strategy actually held rather than the ones it would have wanted. The other five standards qualify this one.
How a return is defined →Risk
A return is quoted against what it cost to earn. Sharpe, Sortino and Calmar are published in excess of the risk-free rate, so a strategy that returns less than cash is recorded as doing so.
How the metrics are computed, and when they are published →Robustness
A result has to survive the search that found it. Every strategy is re-measured against the whole catalogue's recorded trials, and most results that clear the standard bar do not clear that one.
The whole-catalogue bar →Diversification
The unit the firm funds is a portfolio; a strategy is one component of it. Ideas are clustered into families by correlation and by shared code, so a sweep of forty variants counts as one family rather than as forty independent results.
The published portfolios →Execution
A strategy is measured on the costs and delays it would actually meet. The same logic that was measured in research runs on the account, through the same engine, and the difference between the two is measured.
What a fill costs →Evidence
Every published figure can be re-derived from public files, and what is presented as a result is limited to what can be checked.
How to check the record →
Why the search is published
The denominator behind every result.
Search enough and something will look significant by chance. That is the central problem of systematic research, and the only defence is to count every search and correct each result for it.
Every backtest, sweep and grid cell is recorded in an append-only ledger, and a strategy’s result is deflated by how much was searched before it. The size of the search is published alongside what survived it, so a reader can apply the same correction.
The search, as published
- Backtests recorded
- 528,527
- Presented as an edge
- 42
- Portfolios published
- 7
How a result becomes a portfolio
Six steps, and most ideas stop at the second.
- 01
Research
An idea is written as a strategy in a standard shape and backtested inside the firm's framework. Every run is recorded, including the ones that go nowhere.
- 02
Validation
The result is put through the anti-overfit battery, then re-measured against every trial the firm has ever recorded. Most results stop here.
- 03
Selection
What survives is graded and filed by verdict. Nothing the grading rejects is presented as an edge.
- 04
Portfolio
Surviving strategies are assembled into a fixed portfolio: a committed set of strategies and target weights, not re-chosen between sessions.
- 05
Live execution
The portfolio is funded on its own broker account and traded by the desk, which re-uses the research path rather than re-implementing it.
- 06
Monitoring
Each account is marked after its close, the session is hashed into a chain, and the live result is measured against the simulation behind it.
The first three steps end with less than they started with.The last three run whatever survived them.
Portfolio construction
The unit the firm builds and funds.
The unit RVB builds and funds is a portfolio: a committed set of strategies and target weights, assembled from the surviving catalogue and traded on its own account until the firm changes it. A single strategy is a component of a portfolio rather than something the firm runs on its own, which limits how much any one result can matter.
That is also why the research counts ideas in families rather than one by one. Forty variants of the same template swept across forty instruments are not forty independent edges, however uncorrelated their returns happen to look, and a portfolio built as though they were would be concentrated in exactly the way it claims not to be.
Markets
What is traded today, and what is not.
Current focus
- US marketsEquities and ETFs, on paper accounts at a US broker.
- CryptoIncluding the portfolio that trades the firm's own real capital.
Every account behind these is listed, with its kind and its record, under portfolios.
Areas of expansion
- Commodities
- Electricity and power markets
- Further geographiesIncluding emerging and African markets.
None of these is traded today.
The framework, the validation battery and the desk are not specific to an asset class, so extending the research platform to a new market is a matter of a data source and a cost model rather than a separate research process.
Long-term objective
Where the firm is heading.
RVB’s long-term objective is to build an institutional investment platform around a diversified portfolio of systematic strategies.