Transparency
Our Commitment to Honest AI Analytics
finsail is built on the principle that users deserve to understand the tools they are using. This page documents our methodology changes, model updates, known limitations, and our ongoing commitment to clear, honest communication about how the platform works.
Our Transparency Commitments
No silent model changes
Every material change to a model, scoring methodology, or data source is logged publicly in this document with a version number, date, and plain-language description.
Explicit confidence levels
finsail surfaces model confidence and health indicators directly in the UI. When confidence is low or data is stale, the platform says so — it does not serve degraded outputs silently.
No investment advice
Nothing on finsail constitutes investment advice, financial advice, or a recommendation to buy or sell any asset. The platform is a read-only market analytics workspace.
Known limitations documented
We document the known limitations of our models, data coverage, and indicator framework below. We update this list as new limitations are identified.
Known Limitations
- Coverage is limited to assets with sufficient liquidity and trading history. Emerging or low-liquidity assets are not covered.
- All market indicators are derived from price and volume data. On-chain metrics, macro data, and sentiment are not currently integrated.
- The regime detection model is calibrated on historical crypto market cycles. Unprecedented market conditions may produce lower-confidence regime reads.
- Multi-timeframe confluence works best in trending conditions. Ranging or choppy conditions can produce conflicting indicator readings that the model may underweight.
- News flow is curated and filtered but may not capture all relevant market-moving developments, particularly from non-English language sources.
Methodology Changelog
Multi-timeframe confluence model recalibration
Recalibrated timeframe weighting in the confluence model. The weekly timeframe weight increased from 0.25 to 0.30 following backtested performance review. Daily weight reduced proportionally. No change to the scoring scale.
Regime detection model update
Introduced a four-state regime model (expansion, distribution, contraction, recovery) replacing the previous three-state model. Regime probability distribution is now surfaced in the Overview and Intelligence pages.
Key levels methodology revision
Updated the minimum validation requirement for key levels from one to two historical reactions. Reduced the total number of levels displayed per asset to improve readability. Added bi-directional pivot classification.
finsail public launch
Initial public release of the finsail analytics platform. Markets, Intelligence, News, and Reports surfaces live. Methodology published. Coverage limited to top 30 assets by liquidity.