Quant Engine

The Quant Engine is pure deterministic math — no AI involved. Every number here comes from a calculation you could reproduce yourself with the same price history.

The pipeline

1. Feature engine — computes roughly 120 technical and statistical features from price history. 2. Model layer — an ensemble of forecasting models (trend + seasonality, autoregression, and others) produces a 7-day price projection.

Reading the forecast

The forecast shows a central path plus a 95% confidence band — a range, not a promise. A wide band means the models are honestly uncertain; that's a feature of the output, not a bug. Every component model is scored against a random-walk baseline by walk-forward validation (fit on the past, tested on the next unseen step, repeated) — a model only earns weight in the ensemble by beating that baseline.

The diagnostics table

RMSE and MAPE measure average forecast error (lower is better); directional accuracy is how often the model got the sign of the move right. Don't expect high numbers here — for daily equity prices, directional accuracy modestly above 50% is a normal, honest result, not a weak one.

Regime and volatility

The engine also classifies the current regime (trending-bull, trending-bear, or mean-reverting) and estimates EWMA volatility — both describe recent behaviour and can change without warning.