Model Portfolios and Backtest Assumptions

Here are the Portfolio Wisdom theoretical backtest results and the underlying assumptions for a six ETF portfolio including SPY(S&P500 stock index), TLT(Long term US Treasury Index), IEF(Medium Term US Treasury Index), GDX(Gold Mining Stock Sector), IYR(US Real Estate Index), and EEM(Emerging Markets Index).

The interactive chart below shows the hypothetical (theoretical) relative performance of three different Portfolio Wisdom Models if each model had started with $100,000 on 3/12/2007 and was held with dividends reinvested until 7/7/2015.To reduce skepticism people might feel toward theoretical backtests, chart includes non-correlated asset classes going through bull and bear markets (including the crash of 2008). Like any other good methodology, there are times when the portfolio models falls in value, but over time, our methodology is designed to reduce the odds of a big drop in account value, thus producing better risk/adjusted returns.

The sacrifice we make is that when the stock market is shooting upward, our portfolio will probably not keep pace. However, most people are willing to make that tradeoff.

Please note the following caveats and explanations regarding the Theoretical Backtest Results for Portfolio Wisdom Model Portfolios These theoretical backtests are not real investment results in real accounts. Our clients did not experience these results primarily because the model portfolio algorithms were not completed until February 2011. Additional refinements reflected in these theoretical backtests were made as recently as Q2 2014. Further, individual accounts may be customized to specific client requirements, and I do make discretionary adjustments to real client accounts from time to time. Do not expect that actual investment results will always be as good as these theoretical backtests of the Investment Model Portfolios because:

  1. Backtests often do not include the effect of commissions, taxes, slippage, advisory fees, and other trading or account costs that occur in actual practice. In this backtest, I assumed a 0.1% slippage for every purchase and sale, and a commission rate of $.01/share (consistent with commissions at Interactive Brokers where many of my clients have accounts.)
  2. Backtests, by their nature take advantage of hindsight. These models may not do as well in the future as they have using historical data. Markets change.
  3. Portfolio Wisdom may, in the future, change or override the recommendations of the model portfolios or choose to use different asset class ETF’s or other investment vehicles.
  4. This backtest assumed rebalancing of the portfolio when any asset class differed by 5% or more from the target calculated by the algorithm for that specific asset class. For example, if the target was for 10% of the portfolio to be invested in the SPY (S&P500), a rebalance would be triggered if the actual SPY allocation fell to 9.5% or rose to 10.5% at the end of any given day. In reality, future rebalancing may occur more frequently or less frequently.
  5. Diversification and the approach we use does NOT guarantee against loss of principal. In fact, as the performance graphs show, there were periods of time when the backtested models lost money.
    Other information you should know:
  6. The backtests results shown above did account for dividends, which helped total return.
  7. The underlying algorithms for the model portfolios are proprietary to Portfolio Wisdom and were NOT optimized for performance by swapping out one asset class for another.
  8. The Exchange Traded Funds (ETF’s) shown in the models are the actual ones used in client portfolios during that period of time.
  9. Historical daily market data provided by CSI data was used for each model portfolio backtesting. This included raw stock data, split adjusted data, and dividend adjusted data. The TraderStudio software package by Murray Ruggiero was used to run the backtests.

While some of our actual client portfolios now follow the models shown below, other client account portfolios have been customized and will not follow the model completely.