Questions
About the project, the research, and using the data.
01 What is this project?
A free, independent archive of trading experiments that failed their own criteria. An autonomous AI research system comes up with the ideas, tests them on public market data and writes up what happened.
02 Why publish failed trading strategies?
Because a negative result is still a result. Knowing that an idea didn’t hold up, and why, is useful information. It just rarely gets published.
03 Is this research generated by AI?
Yes, from the first hypothesis to the final write-up. AI can make mistakes, so every record shows its data, method and numbers for you to check.
04 Is AI research any good?
Judge for yourself: every record shows exactly how the idea was tested. The AI works the way a careful analyst would. It writes the rules down before looking at the data, counts costs, compares against a control and checks the result on data it hasn’t seen.
05 How are the results checked?
Each record is checked three times, independently and automatically. The test runs its own controls, a validator checks every number in the text against the evidence, and a separate AI fact-checker recomputes the figures from the source results.
06 How reliable are the results?
The numbers are checked, but this is not peer-reviewed research. Errors are still possible, and every result holds only for the period and conditions tested.
07 Does a failed experiment prove a strategy can never work?
No. It failed its criteria under the specific conditions, assumptions and period tested. Different conditions can produce different results.
08 Where does the data come from?
Public market data and other public sources. The dataset contains research outputs, not copies of third-party raw data. Original provider terms apply.
09 Can I reproduce the experiments?
Each record lists its data, rules, costs and steps, so you can redo the test. The code is not necessarily public.
10 How can I access the dataset?
Browse the archive, or accept the Dataset Terms of Use to download JSON or CSV. Agents can request a key and present the terms for human approval.
11 Why is the project free?
We want to share negative findings and see whether AI-generated research can be useful and easy to find. It is independent and non-commercial.
12 Is this investment advice?
No. This is historical experimental research, not financial advice, a recommendation or a trading signal.
13 How often is the archive updated?
New experiments are added in batches, with no fixed schedule.