Articles and notes on quantitative trading, statistical methods, backtesting methodology, and Indian market insights.
Most retail backtests use 1 year of price history. In practice, GARCH(1,1) parameter estimates are highly unstable with fewer than 500 observations. I explain why extending to 3 years dramatically improves alpha/beta stability, especially in Indian equities where volatility clustering is stronger.
RSI is one of the most cited indicators in retail trading. But does the classic oversold=30 / overbought=70 rule actually generate alpha on NSE? I tested it across 30 large-cap Indian stocks from 2021 to 2025 and the results are more nuanced than the textbooks suggest.
How I built a weighted multi-factor model that combines fundamental quality, technical momentum, macro conditions, and governance into a single investable score for each NSE stock. Covers weighting philosophy, normalisation techniques, and how to avoid lookahead bias.
Value-at-Risk gives you a number, but Max Drawdown shows you the experience. I compare VaR, CVaR, and maximum drawdown as risk measures for individual Indian equity positions, and explain why a 25% drawdown on a position can be psychologically dangerous even with positive expectation.
FinBERT, trained on financial news, can classify news headlines as positive, negative, or neutral. I integrated it into a loss-stock decision engine using NewsAPI. This post covers the implementation, limitations on Indian market news, and how I weight sentiment alongside quantitative signals.
Lookahead bias, survivorship bias, overfitting to in-sample data — backtesting mistakes that make a strategy look far better than it really is. A practical guide to building honest backtests, with examples from my own early mistakes and how I fixed them.