The buyer's problem
When a data vendor shows a signal, the question that matters is what else was tried. Without the failed versions and the out-of-sample checks, there is no way to tell a real effect from the best of many tries.
SageFusion's research data is built the other way round: every result arrives with its label and its history, and the failures are in the log next to the successes.
What firms can license
Coverage spans stocks, ETFs, options, futures, crypto and spot FX. The licensable outputs include:
- Evening shortlist: produced after each close, every evening, from a nightly scan of nearly 4,000 US-listed stocks and ETFs.
- Result labels: how each result held up.
- Research log: every tested idea, including the ones that failed, and why.
- Test history: what was tested, the dates, and how it held up on unseen data.
How each label is earned
- Score studies are built on 2005 to 2015 data and must hold up on 2016 to 2026, after trading costs.
- Nightly picks and breakout setups are each compared with a random stock on the same night.
- When many versions of an idea are tried, the best one is discounted for the number of tries.
- Tests use the date data was actually recorded and available.
The method is explained in plain words in our guide to testing a signal against random.
Known limits, stated up front
Most studies use today's listed stocks, so companies that disappeared are missing. That is survivorship bias, and each affected study says so. The research log also shows retractions, such as a machine-learning study withdrawn for a look-ahead bug. A buyer sees these before licensing, not after. See the trading ideas that failed.
Who it is for
The data is built for funds, quant and research teams, data vendors and other data buyers who need to know how a result was tested before it reaches their own process. It suits teams that want to check the work themselves: compare the outputs with their own universe, rerun their own baselines, and read the failures alongside the results.
It is research data for evaluation and research. It is not investment advice and not a managed product.
Delivery and terms
Format, frequency and coverage are agreed per licence, for example file delivery or API. Samples are shared after a short review of each request. See what a research data sample contains.
What is never shared: the rules, weights, thresholds and formulas behind the research. Firms license the outputs, labels and test history.