@Epi @Krush82 @Simon_n How do you deal with correlation changes, sometimes there are strong rotations, how do you try to take out risks in the future? Since this has a massive influence on backtesting and can lead to errors in the future and overfitting? For example, the correlation between the S&P500 and the 10-year US bond has gone from 0.62 to -0.61 within 7 years since around 1993.
•
11
•
18H
@Aktienfox When designing the asset basket, we took care to represent all important asset classes with the most important and at the same time least correlated two assets, e.g. equities: QQQ and EU50 (instead of EM), bonds: TLT, Xeon.
It is clear that the correlations fluctuate, but they should average out across all asset classes, i.e. if two assets correlate more strongly with each other, they correlate less with others. In other words, the fact that the GTAA models map as many asset classes as possible means that the correlations between them rotate, so that the average correlation should remain relatively constant in the future. And that is what matters.
If there is another important, uncorrelated asset class in the future (e.g. token-based personalized carbon emission certificates or something crazy like that) and a gracious issuer makes it leveraged investable, then it will be included in the model. 🤷
It is clear that the correlations fluctuate, but they should average out across all asset classes, i.e. if two assets correlate more strongly with each other, they correlate less with others. In other words, the fact that the GTAA models map as many asset classes as possible means that the correlations between them rotate, so that the average correlation should remain relatively constant in the future. And that is what matters.
If there is another important, uncorrelated asset class in the future (e.g. token-based personalized carbon emission certificates or something crazy like that) and a gracious issuer makes it leveraged investable, then it will be included in the model. 🤷
•
11
•••
18H
@Aktienfox Exactly, first the token would have to have a significant market capitalization and it would have to have been tradable on the market for several years. Then you can start correlation analyses and backtests, possibly synthetically with the unleveraged instruments.
Actually, everything that is uncorrelated and trend-stable to the rest improves the performance values of GTAA.
Actually, everything that is uncorrelated and trend-stable to the rest improves the performance values of GTAA.
•
11
•@Epi Thank you for the input
•
11
•