Loading the live record…
This page reads signals/*.json from the repository.
Serve it from the repository root with python -m http.server 8000
and open http://localhost:8000/site/.
Five-market hive · synchronisation layer · research instrument
Market Fragility
This is not a crash predictor. It is a market-fragility observatory — five market cells (US, UK, Germany, Japan, China) and a synchronisation layer over them — that publishes its entire track record, including its failures, openly.
A witness, not a predictor · each cell and the sync layer is a permanently-unvalidated record
What this is
An observation instrument that measures how much structural fragility has accumulated across five markets and how aligned they are — not a crash predictor, and never a forecast of when a market will fall. Every verdict and the entire track record, failures included, is published openly. Nothing here is a validated signal: each cell and the synchronisation layer is a permanently-unvalidated record.
Background
What this is, and how it works
The question it asks
Not “when will the market fall?” but “how much fragility has accumulated in the system right now?” Eleven public indicators are converted into their own historical percentiles, grouped into five blocks, and averaged into a single 0–100 fragility index.
A regime label — Calm, Accumulation, Fragile, Alarm — is derived from that index with deliberately asymmetric confirmation: it worsens quickly and improves slowly.
Point-in-time discipline
Every observation carries two dates: the period it refers to, and the date it actually became knowable. The engine can only ever see what had been published at the time it is scoring, so a historical test cannot quietly borrow the future.
Leak traps in the test suite tamper with after-the-fact data and assert that the score does not move.
Pre-registration
Indicators, thresholds and success criteria were frozen in writing before the historical test was run. When the instrument failed those criteria, the criteria were not moved — the failure was published instead, and the instrument was left unvalidated.
What it is not
Not a forecast. Not a signal service. Not a timing tool. It does not estimate probabilities of a crash on any date, and it must not be used to make a financial decision.
Source documents
Full methodology
The methodology documents are frozen, pre-registered research records. They are written in Turkish — this page summarises them in English and links to the originals rather than replacing them. The Turkish documents are the authoritative record; where this summary and a frozen document disagree, the document governs.
Core methodology, v1.5 (frozen)in Turkish
v2 fuel × catalyst methodology, v2.1 (frozen)in Turkish
Full backtest report — criteria, sensitivity, ablationin Turkish
v2 refutation run — binding verdictin Turkish
How to cite
Methodology, reports and the signal record are licensed CC BY 4.0: share and build upon them, with attribution and an indication of any changes. Cite the report that contains the finding, and note the access date — the record grows append-only.
Gunaydin, M. A. (2026). Market Fragility: a pre-registered market-fragility observatory (Version v1.5 / v2.1) [Research instrument and dataset]. https://marketfragility.org (accessed YYYY-MM-DD).