PortfolioB passed ADF test, implying stationarity, but has poor backtest returns. Conversely, PortfolioA, non-stationary with poor ADF results, offers better returns and a Sharpe ratio of 1.8. Why is this so?
Test Stats PortfolioB: -2.98
ADF Critical Values {'1%': -3.45, '5%': -2.87, '10%': -2.57}
Test Stats PortfolioA: -1.19
ADF Critical Values {'1%': -3.45, '5%': -2.87, '10%': -2.57}
What's wrong?
Hi Amit,
This is an interesting question. There are few traders who get to see the limitations of the methods they implement. ADF test is a good objective method to test for stationarity but as you have found out, it has its own limitations.
Can you give us more details about Portfolio A and B. How long was the backtest period and the period used for the ADF test?
There can be various reasons for a pair which is co-integrated can fail. One is that the co-integrated pair will not always be co-integrated. You can check the portfolio performance for a longer backtest period and see if the performance remains the same.
Another reason is the volatility of the stocks in question. This could also affect the stock returns.
Hope this helps.