By Caroline Löbhard, Feb. 4, 2022
Testing the software
Backtests are a crucial tool to test the performance of newlz developed algorithms. We explain the basic technique and show some backtesting results.
For so-called backtests automated decisions are checked against historical data.
Since the portfolio decisions of the Goldmarie algorithm are based on historical data, the strict separation of the data for "learning", from the data for verification is essential for a fair and realistic testing procedure.
The Goldmarie portfolios evaluated in this article are each based on historical price trends (gray curves) over 3 years.
For example, stock market data from 2014 to 2016 is used to compile a stock portfolio (gray area).
Then real data was used to simulate how the value of the calculated portfolio (red curve) would have performed over 2017 (red area). The results of these simulations are plotted below as green bars in each case.
The tests use criteria from various sustainable funds (such as that of the GLS Bank equity fund or Ökovision Classic fund from Ökoworld) to create a stock universe of about 200 stocks.
In the current version of our software, a rating of imug rating is used.
Portfolio weights are then calculated using the classic Markowitz model.
The model uses historical stock prices, their statistical covariances (correlations) and stochastic future forecasts in a Monte Carlo approach.
In the future, larger investment universes and more complex models will also be used.
Comparison of Goldmarie with various indices and equity funds
In this graph you can see a performance comparison of the Goldmarie Portfolio "Balanced" with the popular index MSCI World, the equity fund of GLS Bank and the Ökovision Classic equity fund of Ökoworld.
The Goldmarie Portfolio used here corresponds to the risk-return balanced portfolio shown in green.
Our weights were updated for this calculation three times a year (each time based on the previous 3~years).
With an average return of about 22%, the Goldmarie portfolio performs above average compared to the other providers.
Test of the strategy on alternative stock universes
The Goldmarie portfolios are based on a specific, strictly sustainable stock universe.
To show that our strategy also performs well on other universes, we tested it on the universe of the well known US index NASDAQ100.
This graph shows the result.
Over the period of 10 years, our strategy is invested with three different focuses (NASDAQ MinRisk, NASDAQ Balanced and NASDAQ OptReturn) in a historical comparison.
These foci correspond to the portfolios illustrated in blue, green and red in Fig. markowitz.
For comparison, the performance of the NASDAQ100 Index (in purple) and a possible equal distribution across all stocks in the NASDAQ100 universe (NASDAQ Equal Weights in gray) is shown.
As can be easily seen, our strategies perform very well in comparison:
The NASDAQ Balanced strategy, which corresponds to the green "Risk-Return Balanced" strategy in the examples above, is about 19.5% in overall performance, while NASDAQ100 is about 15.2%.
Comparison of Goldmarie with various robo-advisors
Higher returns can be achieved with investments in individual stocks: Fig. performance_robo on the left shows a performance comparison of various robo-advisors for the years 2017 to 2020.
For each provider, the portfolio that comes closest to Goldmarie Finanzen in terms of sustainability and equity share has been selected here.
The costs of the ETFs offered are usually included in the performance, but the costs of the respective providers are not.
In the overall performance over the years 2017 to 2020 in the graph on the right, it can be seen that the single stock investment providers deliver very good performance: Goldmarie Finanzen with 22%, followed by Estably with 16% and Solidvest with 9.7%.
Article overview
