Understanding why we use backtested data in our webApp
Our new ‘home’ is a place to evaluate current strategies and design new portfolios going forward. The data we use is backtested data based on the latest strategy parameters. The reason we do that is to be able to see how each strategy would have performed at current parameters so that we can then combine them into a portfolio. In essence, our new tools are forward looking tools that help build a portfolio. They are not tracking tools and should not be used to monitor your current strategy or portfolio performance.
A side effect of using backtested data is that when strategy parameters are updated, all backtested history (including past month allocations) are updated as well. To look at performances based on actual published past signals you can visit the historical data section.
Creating a portfolio from strategies
Let’s assume we want to construct a portfolio that includes amongst others, the Universal Investment Strategy (UIS) and the Enhanced Permanent Portfolio (EPP). The first step would be to analyse the performance of UIS vs the performance of the EPP strategy.
A little background on UIS & EPP: Our first implementation of the Universal Investment strategy, used only 2 ETFs: SPY and TLT (the SP500 and Treasury ETFs). As market changed UIS evolved. Starting January 2018 UIS was tweaked to allow allocations to GLD (the gold ETF). The EPP strategy is an ‘all weather’ strategy that can allocate to SPY and TLT (like UIS) but can also include GLD in inflationary environments.
Again, 2 sets of performance data to consider:
a. The performance data from the ‘actual’ historical allocations. This would be the data based on what an investor could have traded, on a real account, based on allocations published at the time. This data would not have included allocations to GLD before January 2018.
b. Backtested data based on the latest parameters and asset universe. This data is produced from backtests performed today taking into account current strategy settings and asset universe, i.e. the possibility of UIS having invested in GLD. This backtested performance data is not real as it could never have been traded at the time. It is ‘virtual’ performance not appropriate for tracking account performance.
If we use historical performance data and analyze say 5 years of history, the data would include a long period where UIS did not allocate to GLD. This would result in a (false) lower correlation between the strategies. In other words, UIS and EPP used to be more different in the past. They are now more similar. Using historical data would be misleading and would result in a portfolio with higher than optimal portfolio exposure to GLD.
So in order to run correct risk/reward/correlation analysis of strategies in a forward looking portfolio construction process we need to use backtested data. Of course we also track ‘real’ (out-of-sample) performances for obvious reasons. This may create some confusion but we do need both sets of data to cover our historical evaluation and portfolio construction tools.