Home

Welcome to the knockoff_index wiki!

What?

This project aims to generate positive alpha by using an econometric tool –knockoff filter as in Barber and Candes (2015). The top 100 US stocks by market cap are considered as the candidates to construct active portfolios. The portfolios are based on analysis of the past 10 years performance and rebalanced annually. The performance of portfolios are measured by excess return (relative to effective funds rate) and are tracked over 1996-2020.

How?

The knockoff filter develops a regression model by controlling for false discoveries. This project uses four alternative US indexes – DJIA, S&P 500, NASDAQ, and Russell 1000 – to pick an optimal subset of qualified stocks. It is up to the practitioner to define what defines as qualified stocks. For the moment the qualified pool of stocks is defined as 100 US stocks with the highest market capitalisation. The stocks picked by the knockoff filter are used to build optimised portfolios with minimum RMSE in forecasting the benchmark (DJ, NASDAQ, S&P 500, or Russell 1000).

Results

TBC Photo 1 description Photo 2 description