Methodology

How finsail's AI Analytics Engine Works

finsail surfaces structured crypto market intelligence across six connected layers. This document explains the methodology behind every analytical output, score, regime label, and confluence output that appears in the platform — so that users can evaluate and trust the tools they are using.

1. Core Philosophy: Context Over Prediction

finsail does not generate directional trade indicators, price predictions, or investment recommendations. Every AI model in the platform is designed to produce structured market context — the kind of information that helps a human analyst ask better questions, not the kind that removes judgement from the process.

The distinction matters. Prediction-based systems tend to overfit to recent conditions and fail opaquely. Context-based systems surface the underlying structure of the market so that users can reason about conditions themselves. finsail is built on the second model.

All outputs are derived from price and volume data. No macro narratives, no social sentiment proxies, no off-chain data inputs are allowed to overwrite the quantitative structure of the model. The market is the primary source of truth.

2. Data Sources and Coverage

finsail ingests market data across multiple timeframes — intraday, daily, weekly, and monthly — for Bitcoin, major altcoins, and a curated list of crypto assets with sufficient liquidity and trading history to produce reliable outputs.

Coverage is intentionally selective. Assets are included only when the data quality meets minimum standards for indicator stability. The platform displays coverage status in the Overview surface so users always know what is currently modelled and what is not.

Data freshness is shown on each surface. Staleness indicators appear automatically when model outputs are older than expected thresholds for a given timeframe.

3. Market Regime Detection

Regime detection is the foundation of the finsail analytics layer. Before any asset-level indicator is interpreted, the platform establishes the current macro-market regime: whether the broader crypto market is in an expansion, contraction, recovery, or distribution phase.

Regime labels are derived from a combination of trend structure, volatility state, breadth across covered assets, and cycle positioning. The model produces a probability distribution across regime states — not a hard binary label — which allows the interface to communicate confidence alongside the current regime read.

Regime context is displayed in the Overview surface and flows into asset-level intelligence. An asset that looks bullish in isolation may look very different in the context of a distribution regime — the platform surfaces this relationship explicitly.

4. Multi-Timeframe Confluence Model

The Intelligence surface is built around multi-timeframe confluence analysis — the systematic evaluation of whether market indicators across the 15-minute, hourly, 4-hour, daily, and weekly timeframes are aligned or in conflict.

Each timeframe produces an independent indicator reading derived from trend structure, momentum, and key level proximity. The confluence model then aggregates these indicators, weighted by timeframe hierarchy, to produce the overall score displayed in the Intelligence surface.

High confluence — when multiple timeframes agree — is surfaced differently from low confluence. A strong indicator reading on one timeframe with conflicted indicators on others is presented with explicit context about the disagreement, not as a clean directional call.

5. Scoring System

finsail scores are normalised on a 0–100 scale. Scores above 60 indicate broadly constructive conditions. Scores below 40 indicate broadly deteriorating conditions. The middle range (40–60) reflects neutral or mixed conditions where no dominant structure is present.

The overall score is a weighted composite of several sub-scores:

• Trend Score: Based on price structure, higher-high/higher-low sequences, and moving average relationships across timeframes.
• Momentum Score: Derived from rate-of-change, relative strength, and oscillator positioning — normalised to remove recency bias.
• Volatility Score: Measures current volatility against historical ranges and flags expansion or contraction regimes.
• Structure Score: Evaluates proximity to key support and resistance levels, pivot zones, and high-volume nodes.
• Regime Alignment Score: Measures how well asset-level market indicators align with the current macro-market regime read.

Score weights are recalibrated periodically using backtested performance against historical market conditions. Any significant methodology update is logged in the Transparency page.

6. Key Levels Detection

Key levels in finsail are derived from a combination of price structure pivots, high-volume zones, and historical reaction points. The model identifies levels that have produced measurable price reactions on at least two occasions across multiple timeframes, with preference given to levels validated at higher timeframes.

Levels are classified as support, resistance, or pivot (bi-directional) and displayed with relative strength indicators based on the number and quality of validations. Levels near the current price are highlighted, and the Intelligence surface surfaces proximity context explicitly.

7. Analytical Scenario Framework

The scenario framework output in finsail is a structured summary of likely follow-up paths given the current regime, confluence score, and key level context. It is explicitly not a recommendation to buy or sell.

A scenario framework includes: a primary directional bias (bullish, bearish, or neutral), the most important support and resistance levels to watch, a description of what conditions would confirm or invalidate the primary bias, and a risk note that flags the main contradicting market indicators.

Scenario frameworks are generated per-asset and updated on each model run. They are designed to be a starting point for analysis, not a final answer.

8. Model Health and Quality Indicators

finsail exposes model health metrics so that users can evaluate the reliability of outputs at any given time. Health indicators include:

• Data freshness: time since last successful model run for each asset.
• Coverage breadth: the percentage of tracked assets with current valid outputs.
• Indicator stability: whether the current market indicators are consistent with recent prior runs or have shifted materially.
• Regime confidence: the probability mass behind the current regime label.

When model health is degraded — due to data latency, insufficient coverage, or unusual market conditions — finsail displays explicit notices rather than silently serving stale or low-confidence outputs.