Methodology

Data Sources

Ansaar aggregates data from NSE (National Stock Exchange of India), BSE, and cryptocurrency exchanges. Data includes daily OHLCV prices, F&O open interest, delivery volumes, institutional trade data (FII/DII), sector flows, and fundamental ratios such as P/E, P/B, ROE, and debt-to-equity.

Crypto data is sourced from major exchanges via aggregated APIs. Global macro data (VIX, DXY, oil, copper) is sourced from market data providers.

Equity Quantitative Research Model

The equity model (ADR-0061) is a LightGBM gradient-boosting model trained on ~150–180 features derived from price, volume, F&O, institutional flow, and fundamental data. It produces return estimates across 1-day, 5-day, and 21-day horizons.

Model outputs are scaled to a standardized range and combined using a Forecast Diversification Multiplier (FDM) following the framework in Robert Carver's “Systematic Trading”. The combined score represents the relative conviction of the model — descriptive, not an absolute return estimate.

• Training information coefficient (IC): 0.35–0.61
• Retrained bi-weekly on rolling historical data
• Position sizing via Half Kelly criterion with volatility targeting

Crypto Quantitative Research Model

The crypto model uses a relaxed LightGBM architecture trained on ~77 features including price-derived statistics, funding rates, global market metrics, and liquidity indicators. It targets 5-day risk-adjusted returns.

• Training IC: 0.08–0.30
• Retrained bi-weekly
• Stablecoin and low-liquidity pairs filtered out

ETF Carver Model

ETF quantitative research combines rule-based signals following Carver's methodology: EWMAC trend indicators (16/64 and 32/128 crossovers), Donchian channel breakout indicators, and NAV arbitrage signals. Signals are scaled to the same –20 to +20 range and blended using an FDM.

Macro Regime Classification

Market regime is classified using a Hidden Markov Model or threshold-based classifier applied to price momentum, volatility, and breadth indicators. The three classes are: Bull Trend, Bear Trend, and Sideways. Regime probability represents the model's confidence in the current classification.

Institutional Activity Indicators

The institutional activity view tracks delivery volume trends and F&O open interest patterns to surface potential accumulation or distribution at the institutional level. These are statistical indicators derived from public NSE data — not confirmed insider or institutional trade data, and not buy/sell calls.

Limitations

All models have limitations. Model outputs degrade during regime changes, low-liquidity periods, or when market structure shifts significantly. Backtested performance is calculated on historical data and may not reflect live trading outcomes due to transaction costs, slippage, and look-ahead bias.

This platform publishes research outputs as-is. See our full disclaimer.