Options Menu

The Options Menu offers ways to adjust the optimization problem.  Many of these options are displayed on the information bars at the top and bottom of the each worksheet when they are enabled, and the Info Worksheet provides a summary of your current settings and those used for the most recent optimization.

Optimization Options

• Active Weight Optimization allows you to perform benchmark relative optimization.
• When Long Short Optimization is enabled, long and short positions are permitted during optimization and constraints can be applied to the long-only and short-only parts of the portfolio.
• Enter a Multi-Period Horizon if desired.
• Change the number of Portfolios on the resulting frontier. The number of frontier points affects the charts. Fewer points result in choppier charts.
• Enter a Risk-Free Rate if desired.
• Entering an inflation rate will not change the Optimizer case, but if you load the case into LifeCycle, the inflation rate carries over.  This can be useful if you are working with the expectation of a particular inflation rate in Estimator and Optimizer while creating the frontier.

Advanced Options Sub-Menu

• Seed sets a seed for all pseudorandom numbers in the Monte Carlo simulations used during optimization, fixing the results for all runs.   With no seed, the results vary between runs, subject to the Monte Carlo error introduced by the random sampling.  This error can be reduced by increasing simulations, but it is always part of the solution, whether a seed is set or not.  Because of the inevitability of sampling error, using a default seed is only recommended for testing purposes.  Default seeds will work whether multiple cores are enabled or not.

• The Optimization Algorithms offer two different ways to compute the efficient frontiers that comprise the Michaud efficient frontier when averaged together.  

  • The Interior Point Algorithm uses separate runs of a sophisticated interior-point quadratic programming algorithm to optimize each point on each frontier.  It provides extremely reliable answers and supports all types of constraints, but may be slower than the critical line algorithm.

  • The Critical Line Algorithm uses Markowitz's famous method for finding the pivot portfolios, which determine the mean-variance efficient frontier. These portfolios are then used to find the optimal portfolios requested by the user.  This algorithm tends to be faster than the interior point option and is the default computation method.  Both algorithms allow different spacing methods. See the Frontier Spacing section below for details.

• Frontier Averaging options determine how the points are distributed along the entire frontier.

  • Return Rank -- Points have equally spaced means up and down the frontier.  This method produces frontier points which tend to be more dense in the low end of the frontier.  

  • Arc Length Rank -- Points are separated by equal arc length segments on the frontier.  In this setting the arc length is taken along the frontier plotted with respect to the same time period as the data, i.e. not annualized for monthly data.  This method spaces the points more evenly across the frontier and will fill in the gap typically left at the top of the mean-spaced frontier.  New Frontier recommends this option, and it is the default.

  • Sigma Rank -- Points have equally-spaced standard deviations.  Conversely to return-rank averaging, this method will fill in the gap at the top of the frontier, but may create gaps at the low end of the frontier.

• Rejection Sampling enables and sets the threshold for rejection sampling. This can eliminate more perverse simulations that are more inconsistent with the input assumptions. 

• Filter Matching enables filter matching for the rebalance test.

• Resampled Returns Distribution

  • Normal distribution assumes that the risks and correlations are perfectly known and that the returns distribution is normal, though unknown.

  • By default, the Optimizer resamples the mean vector and covariance matrix from a multivariate t Distribution.  We recommend the t distribution, especially for low forecast confidence optimizations. The t distribution is the statistical "best-guess" distribution of the returns when the variance is unknown.  

  • When you enable four moment resampling, you may specify a different multivariate distribution for resampling by entering skew and kurtosis values for each asset.

• Enabling Ledoit Covariance Estimation means that the Optimizer will use a Ledoit covariance estimate for each simulation's optimization input. If you disable this option, the Optimizer uses an ordinary sample covariance matrix.  This option may not be appropriate when using derivative assets or assets with a deliberate correlation of 1. However, it is always appropriate for low forecast confidence cases, especially if the number of assets is large.

• The Options Strategy allows you to model the impact of put- or call-type derivative instruments on the efficient frontier.  When you enable the options strategy, puts and calls columns appear on the Inputs Worksheet.

• Enabling Optimization Tolerance replaces the Simulations Setting as your way to tune the accuracy of the optimization.

• Reset Advance Options returns all settings to the defaults.

Miscellaneous Options

• Percentile Bound displays the lower and upper percentiles shown on the Results Worksheet.  Toggle the drop down menu to change the percentile bounds.  The default range is 25%-75%.

• Benchmark Portfolio adds a Benchmark Portfolio Column to the Portfolios Worksheet.

• Initial Portfolio adds an Initial Portfolio Column to the Portfolios Worksheet.  As initial portfolios impact rebalancing and transaction costs among other things, the initial portfolio is displayed by default.

• Portfolio Total Value adds the Total Value Row to the Portfolios Worksheet.

You may also want to review the description of the display options:

• Display Menu (columns and worksheets shown)