Research
Preregistrations
A preregistration is a dated document that sets out, before the confirmatory tests, the question, the data, the protocol and what would count as success. It keeps a result from being “found” by trying variants until one happens to work.
Registered
The documents are written in Spanish.
- Article 4: can you collect the decay of UVXY and SQQQ? (H1–H5)
- Paper-trading study: BTC and HYPE carry on Hyperliquid (26 weeks)
- Article 6: shorting the −3x after rallies (H1–H5) and its annex
- Article 7: an optimal-leverage indicator for 2x and 3x ETFs (H1–H8)
- Hyperliquid, part 2: how to capture funding
What each preregistration sets out
- What has already been seen
- If any figure was looked at before registering, it's declared (“declared contamination”) and not cited as evidence.
- Data and windows
- Sources, start and end dates, and the robustness windows.
- Protocol
- When the signal fires and when the order is executed, the costs, and which statistical tests correct for the search across each family of variants.
- Hypotheses
- Each one written so the data can refute it, with the threshold that would falsify it.
- What gets published no matter what
- The results as they come out, including null ones, with their limits.
- Deviations
- Any later change is noted at the end, with its date and its reason.