Seminar paper · Reconstructed revision · 14 October 2026
Bitcoin Bubble Detection with GSADF
When does Bitcoin price behavior exhibit statistical explosiveness? This seminar study explores recursive right-tailed tests and the distinction between sample-wide evidence and episode dating.
Research framing
Testing a market narrative against price data
The paper examines whether recursive right-tailed unit-root tests can distinguish explosive Bitcoin price behavior from a unit-root null. A global GSADF test and a BSADF dating sequence answer related but different questions; both depend on their sample and specification.
The revised paper was reconstructed from the original PDF because an editable original was unavailable. The original figure conflicts with its reported scalar statistic. Test settings, critical values and exact episode dates require replication; the revision does not supply a new empirical run or verified return and volatility comparisons.
Method
Recursive testing with bounded interpretation
- 01
Frame the series
Specify the Bitcoin price source, sample, transformation and frequency before interpreting explosive behavior.
- 02
Apply right-tailed tests
Use recursive right-tailed regressions and the sample-wide GSADF statistic, with explicit lags and minimum windows.
- 03
Specify episode dating
Compare the endpoint BSADF sequence with matching critical values and a stated minimum episode duration.
- 04
Constrain interpretation
Resolve the original figure–scalar conflict and reproduce the test before accepting exact episode dates or return comparisons.
Research record
The method and the replication boundary
- Right-tailed explosive-root tests
- GSADF across recursive windows
- BSADF with matching critical values
- Exact episodes and numerical results unverified
Lessons
The test is only as strong as its framing
- Specification matters
- Sample, transformation, lags, minimum windows and critical-value simulation must be recovered to reproduce the test.
- Dating is a separate step
- GSADF is the global statistic; BSADF supplies the endpoint sequence. Exact episodes in the original remain unverified.
- Inference needs context
- Changing volatility and repeated testing affect inference. Explosiveness alone does not establish fundamental mispricing or an actionable trading rule.
Research boundary
Academic interpretation only
This seminar paper is academic research. It does not provide trading signals, recommend buying or selling Bitcoin, or offer investment advice.