Methodology

Quotes and historical OHLCV are pulled from Yahoo Finance for NSE (suffix.NS) and BSE (.BO) tickers via server-side functions. Live quotes cache for 1 minute; daily candles cache for 6 hours so the dashboard refreshes through the trading day without hammering the upstream API. The NIFTY 50 universe is pre-curated for the dashboard, but you can search and analyze any listed ticker.

• Daily simple returns rₜ = (Pₜ − Pₜ₋₁) / Pₜ₋₁ over the last ~252 trading days.
• Annualized return = mean(r) × 252; annualized volatility = stdev(r) × √252.
• Sharpe ratio uses an Indian risk-free rate proxy of 6% annual.
• Max drawdown computed on close-to-trough basis from the running peak.
• Skewness and excess kurtosis use the standard sample estimators.
• Value-at-Risk 95% (1-day) is the historical 5th-percentile loss; CVaR averages the tail.

Two complementary models are run for every stock:

1. Linear regression forecast. Ordinary least squares fitted on the last 60 trading days, projected 30 sessions forward. The shaded band is the 95% prediction interval derived from the residual standard error and leverage of the future point.
2. Monte Carlo simulation (GBM). 1,000 paths under a Geometric Brownian Motion: Sₜ₊₁ = Sₜ · exp((μ − ½σ²) + σZ), with μ and σ estimated from log-returns. We report the 5/50/95th-percentile cone and the empirical probability of finishing above today's close after 30 trading days.

Technical signals layer on top: 20/50/200-day SMA crossovers, EMA-20, RSI-14, MACD (12, 26, 9), and Bollinger bands (20, 2σ). The composite score weights these inputs into a transparent BUY / HOLD / SELL recommendation.

The portfolio optimizer applies Markowitz mean-variance optimization to your basket. We sample 4,000 random long-only weight vectors that sum to 1, compute each portfolio's expected return wᵀμ and volatility √(wᵀΣw) using the annualized covariance matrix Σ from 2 years of daily returns, then identify:

• Max-Sharpe portfolio — the tangency portfolio maximizing (return − rf) / vol.
• Min-volatility portfolio — the lowest-risk allocation.
• Equal-weight benchmark — naive 1/N allocation, for comparison.

Random-search optimization is fast and approximates the efficient frontier well for baskets ≤ 10. For larger universes we'd swap in quadratic programming.

Clairvoyance is an educational analytics platform. Forecasts are statistical projections based on historical data and assume returns are stationary — they do not account for earnings, macro shocks, regulatory news, or corporate actions. Nothing on this site is investment advice. Always do your own research and consult a SEBI-registered advisor.