Implication of Varying Estimation Period of Technical Indicators on Trading Performance Using Traditional ML Models
摘要
Recent studies of Security Exchange Board of India (SEBI) underlines the precarious condition of the retail traders of India. Keeping the retail traders in mind, the present study explored how retail traders’ performance can be improved with minimal difficulty in terms of complexity of computation techniques and requirement of computation power. The study clearly underlines that for short-term trading, reduction of evaluation period of technical indicators and selection of suitable strategy indices considerably improves trading performance. For 1-day and 3-day ahead trading, forecasting of direction of all indices can be best performed with a 3-day evaluation period. For 5-day ahead trading, 5-day based evaluation suits the best for Nifty 200 Quality30 and Nifty Alpha50. For Nifty50, 7-days of evaluation suits the best. The maximum return of Buy & Hold (B&H) strategy, from Nifty50 stands at 205.5%. But with customized evaluation period, the maximum return generated by Nifty50, Nifty Alpha50 and Nifty 200 Quality30 are 289.3%, 647% and 349% respectively. Such performance also outperforms the return from B&H strategy of respective indices. The study also identified that William’s R, Relative Strength Index (RSI), Fast-K are important predictors of periodic return of all indices under Ridge and LASSO. Similarly, volatility indicators such as VIX close, normalized average true range (NATR), T3 moving average are important predictors of periodic return of Nifty50 under spline. Most importantly, the study shows if trades are executed in 5 days, based on the trading signal generated with 7 days of data, result in highest return for all indices.