Archive
Quant strategies
Backtesting methodology, portfolio optimization, and risk management strategies
Read articleQuantHow revised economic data mislead a backtest: two dates for every value
Separate observation periods from information availability, retain data vintages, and replay decisions using only the versions that were knowable at the time.
Read articleQuantUsing ATR to set stop distance and position size
How Wilder's True Range and smoothing produce ATR, and how to tie a volatility-based stop distance to a fixed risk budget per trade in a form you can test.
Read articleQuantReading Bollinger Bands: what band touches and squeezes do and do not tell you
How Bollinger Bands are calculated, the default settings, %B and BandWidth, and what band touches, walking the bands, and the squeeze mean, based on John Bollinger's rules and runnable code.
Read articleQuantBefore using Polymarket prices as event-probability data: public APIs, identifiers, and liquidity filters
How Polymarket's public APIs expose markets, order books, and price history, why market IDs and outcome token IDs must be kept apart, and how quote-quality filters and time alignment turn event probabilities into testable data.
Read articleQuantChecking backtest overfitting: walk-forward and purged k-fold answer different questions
A strategy picked after many trials can look good by chance. This note separates what chronological walk-forward, purged k-fold with label-overlap removal, and trial logging each do.
Read articleQuantDon't judge a factor combination by IC alone: consistency, turnover, overlap, and horizon
IC is a correlation between forecasts and outcomes, not a net-of-cost return. How to check ICIR, rank autocorrelation and turnover cost, overlap between factors, and IC by forecast horizon.
Read articleQuantTurning a Pine Script v6 strategy idea into something you can validate
When prototyping indicators and strategies in TradingView Pine Script v6, check fill timing, lookahead in other-symbol data, and cost settings, then port the same rule to Python and validate it chronologically.
Read articleQuantRSI at 70 is not a sell order: calculation and the conditions for reading it
How Wilder's RSI is calculated and smoothed, what the 70/30 zones, divergences, and failure swings describe, and why their meaning changes in trending markets, with sources.
Read articleQuantHow to read MACD: why crossovers arrive late and what to check
How the MACD line, signal line, and histogram are calculated, why signal-line and zero-line crossovers lag the trend, and the limits of reading divergences, with sources and runnable code.