Backtesting portfolio optimization of Nordic stocks

In this post, we present several (out-of-sample) backtesting results for Nordic stocks using Renexe algorithms. The original list contains over 300 Nordic companies active as of the year 2013 because the algorithms require at least 5-year training return data for the testing. The backtesting period is the period of the brief market crash of late 2018, i.e., 2018.09.03 - 2019.06.19 and the training period - ~5 years before starting date.

Figure 1. List of Nordic stocks for backtesting

We build the portfolio ´s with optimal risk/return characteristics for the neutral and negative macro-scenario forecasting and 95% confidence interval for tail risk, i.e., we are minimizing the potential expected loss of 5% worst scenarios. Furthermore, we are building the portfolios maximizing tail risk and drawdowns and observe the behavior of such portfolios during the market crash.

Figure 2. Backtesting portfolio optimization of Nordic stocks for market crash of late 2018

Figure 3. Optimal portfolio of Nordic stocks, 2018 market crash, neutral macro-scenario

As we can see from Figure 2, the optimal portfolio avoids a large part of the drawdown and yields significantly higher returns and Sharpe ratio at the end of the backtesting period. In the next backtest we build the optimal portfolio specifically for falling markets, i.e., using negative macro-scenario modeling using SP&500 as a stock market proxy. The goal is to build a portfolio of Nordic which is insensitive to the stock market fluctuations.

Figure 4. Low beta portfolio of Nordic stocks. Test 1.

Figure 5. Low beta portfolio of Nordic stocks. Test 2.

Figure 6. Optimal portfolio of Nordic stocks, insensitive to market drawdowns.

We can see from Figures 4 and 5 that optimized portfolios have very low drawdowns and generally low volatility. However, the insensitive (low beta) portfolio may not benefit from upward market swings as well, thus underperforming during stock market booms. The results below are obtained with relaxed constraint for maximum stock weight in a portfolio.

Figure 7. Backtesting portfolio optimization of Nordic stocks for negative macro-scenarios

Figure 8. Optimal portfolio with relaxed size constraints

In the next part of this report, we try to find the portfolio of Nordic stock which is the most sensitive to the market fluctuations (high beta portfolio).

Figure 9. Backtesting portfolio of Nordic stocks with high beta

Figure 10. High beta portfolio of Nordic stocks

As we can observe in Figure 9 the high beta portfolio experiences a massive drawdown of 45% (against ~17% of the benchmark) but quickly rebounds of the next weeks to the return level close to the benchmark. Hereby, in this post we demonstrated several applications of Renexe algorithms that allow us to build highly customized stock portfolios for different market conditions, only using local stocks of Nordic countries.