How FynSight Works

Every market day, a fully automated pipeline fetches data, runs six intelligence engines, validates every output across four layers, and publishes the result. Zero human intervention. Zero opinions. Just data, engines, and validation.

The Data Foundation

FynSight's data pipeline collects market data from multiple sources every market day. NSE bhavcopy provides official Indian equity prices. Yahoo Finance provides global indices and supplementary data. RBI databases provide macro indicators. NSE provides FII and DII flow data.

The pipeline has accumulated 18 years of daily price data across 120 instruments, 600+ quarterly fundamental snapshots across 55 companies, and thousands of data points on institutional flows, sector performance, and macro indicators. Every data point is timestamped, source-tagged, and traceable back to its origin.

Price History

18 years

Instruments

120+

Fundamental Snapshots

600+

Companies

The 6 Intelligence Engines

Once data is fetched and validated, six intelligence engines run in parallel. Each engine is a self-contained analysis system. No engine imports from another, ensuring independent validation of every output.

Market Regime Classifier

Analyzes breadth, volatility, FII/DII flows, sector concentration, and market-cap participation to classify the market into one of 7 regimes. Produces a confidence score from three dimensions: data quality, signal alignment, and cross-asset consistency.

Sector Intelligence Engine

Ranks all 8 Nifty sectors daily on returns, breadth, momentum, and relative strength. Each sector gets a score from 0 to 100 with classification: LEADER, STRONG, NEUTRAL, WEAK, or LAGGARD.

Stock Intelligence Engine

Generates multi-layer analysis for 50 Nifty stocks. DuPont decomposition of ROE, margin structure analysis, earnings quality assessment, debt sustainability checks, and fundamental scoring with a 0-100 rating.

Causal Chain Detector

Identifies active causal chains driving market movement. Maps the sequence from global triggers (US yields) through intermediate effects (DXY, FII flows) to Indian market impact.

Historical Replay Engine

Matches today's market setup against 11 years of historical data. Finds the closest historical episodes and shows what happened next. 4-dimension similarity scoring across Nifty trend, VIX level, sector signal, and drawdown.

Fundamental Intelligence Engine

9-layer analysis including DuPont decomposition, earnings quality checks, margin trend detection, cash flow analysis, and 26 risk flags across 5 categories. All ratios self-computed. No reliance on pre-computed values.

The 4-Layer Validation Pipeline

Before any intelligence reaches you, it passes through four validation layers. Each layer checks a different dimension of quality. The final score determines whether the analysis is published, published with a note, or held for review.

Layer 1: Data Quality (30% weight)

Every instrument is checked for staleness, impossible values (close = 0, high < low), abnormal single-day spikes (>15% for stocks), duplicate rows, and NaN values. Instruments that fail are flagged and their data is excluded from engine computation.

Layer 2: Cross-Source Verification (25%)

For instruments available from both NSE bhavcopy and Yahoo Finance, the two sources are compared. If the difference exceeds 2%, the instrument is flagged. This catches data feed errors before they affect analysis.

Layer 3: Engine Sanity Rules (25%)

50+ hard rules verify logical consistency. Does a RISK-OFF regime correspond to breadth below 50%? Does HIGH confidence have 3 or more corroborating signals? Does VIX above 25 match 'elevated' volatility classification? If rules fail, the analysis is flagged and the score is reduced.

Layer 4: AI Coherence Review (20%)

An AI model reads the full output and checks for logical contradictions. If the model flags issues, they are surfaced as quality notes. This layer is optional and gracefully skipped if the AI service is unavailable.

Publish Decision

Final Score = L1 × 0.30 + L2 × 0.25 + L3 × 0.25 + L4 × 0.20. Score above 90: auto-publish. Score 80 to 89: publish with quality note. Score 60 to 79: hold, show previous data. Score below 60: critical hold, alert sent for investigation.

Internal Testing

FynSight's engines are tested internally against historical market data spanning 11 years. The regime classifier has been evaluated across 9 major market events ranging from the COVID crash of March 2020 to the narrow leadership period of 2025-2026.

Internal testing is conducted to verify that engine outputs are logically consistent across different market conditions. This is a quality assurance process, not a performance claim. Past internal test results do not guarantee or predict future engine outputs or market outcomes.

Important

FynSight is a market intelligence platform, not a SEBI-registered research analyst or investment advisor. All content including regime classifications, confidence scores, sector rankings, and engine narratives is provided for informational purposes only. Internal testing results reflect historical engine performance during development and quality assurance. They do not constitute performance claims, recommendations, or forecasts of any kind. All investment decisions and their consequences are solely your responsibility.

See today's intelligence

Every section above is running right now. Data is live. Engines are live. Validation is live.

Market GPS →Daily Brief →