POP Portfolio Optimization Description


	POP Portfolio Construction Suite POP is more than just a portfolio optimizer. While optimization is a central part of the program's functionality, it goes beyond the typical efficient frontier approach of other financial software. The genetic algorithm that powers the portfolio selection and asset allocation allows the solution of relevant problems, not just what is mathematically tractable. In particular it allows the generation of random portfolios -- these are extremely useful and will dramatically change the practice of fund management. POP is designed to be as flexible and far-reaching as possible while maintaining ease of use. The POP User's Manual contains a number of examples, one of which should be close to what you want to do. The manual also contains practical advice on performing optimizations regardless of the software used. POP's capabilities can be divided into four major areas: Generate Random Portfolios Obeying Constraints Optimally Select a Portfolio Optimal Asset Allocation Statistical Factor Model This functionality makes POP useful to a number of types of organizations: Traditional Fund Managers Hedge Funds Funds of Funds Consultants Plan Sponsors Brokers Key Functionality Generate Random Portfolios Obeying Constraints Generate a list of portfolios that obey portfolio selection constraints, but which are indifferent to the utility. Some uses of random portfolios are: • Measure the performance of a fund. • Test your investment process. • Evaluate the effect of constraints -- are they too loose or too tight? • Assess the quality of a risk model. • Bid on unknown portfolios with known characteristics. See Random Portfolios in Finance. You may also want to see Section 8.1 of the POP User's Manual for some more details on the uses of random portfolios. Optimally Select a Portfolio Create a portfolio of assets out of a larger universe of assets, or update such a portfolio. Features include: • Inputs and outputs are in meaningful units. The results need not be post-processed which can save significant labor, and makes automation feasible. • Specially designed for long-short portfolios. • Bounds on the number of assets to trade and the number of assets in the portfolio. • Ability to limit turnover, tracking error, volatility, expected return. • Mean-variance optimization, maximize the information ratio, minimize variance, or minimize tracking error. • Different trading costs allowed for long assets, short assets, buys and sells. • Non-linear trading costs allowed (polynomials of arbitrary order, and more generally arbitrary exponents on the number of units traded). • Linear constraints on the portfolio and/or the trade. • These linear constraints may be on either the gross or the net. • Constraints on sums of the largest weights. • Threshold constraints -- minimal amount of an asset traded, if traded at all. Likewise, portfolio thresholds impose a minimum amount in the portfolio, if present. • Forced trades may be specified. • Multiple benchmarks allowed. • Multiple variances allowed. For example, both a statistical factor model and a fundamental factor model could be used. • Round lotting within the optimization -- trade only at a round lot except to close out a non-round position. Optimal Asset Allocation Allocate among assets or asset classes. This is very similar to portfolio selection, but there are two major differences. There is no constraint on the number of assets that appear in the portfolio. The portfolio is constrained to be long-only. • Find and plot efficient frontiers. • Different trading costs allowed for buys and sells. • Non-linear trading costs allowed (polynomials of arbitrary order, and more generally arbitrary exponents on the weight traded). • Linear constraints on the allocation and/or the trade. • Constraints on sums of the largest weights. • Multiple variances allowed. • Simultaneous scenario analysis. Statistical Factor Model Calculate a statistical factor model. An essential input to an optimization is a variance matrix. The easiest practical method of creating a variance matrix for an optimization is to estimate a statistical factor model. The only input needed is a matrix of historical returns of the assets. This function allows: • Iterated (or not) principal factors. • Flexible missing data handling, or do your own. Additional functionality regarding variance matrices: • Add an index to a variance matrix using its constituent weights. • Transform a variance matrix into one that is relative to one of its assets. Types of Organizations Traditional Fund Managers • Portfolio optimization, generally long-only. • Asset allocation. • Random portfolios for testing prospective trading strategies, for deciding on the most useful constraints, for evaluating performance -- both for internal and external consumption. See Random Portfolios in Finance. Hedge Funds • Portfolio optimization, generally long-short. • Random portfolios for testing prospective trading strategies, for deciding on the most useful constraints, for evaluating performance -- both for internal and external consumption. See Random Portfolios in Finance. Funds of Funds • Random portfolios for evaluating the performance of funds. See Random Portfolios in Finance. Consultants • Asset allocation, including multiple scenario analysis. • Random portfolios for evaluating the performance of funds, and for creating mandates free of a traditional benchmark. See Random Portfolios in Finance. Plan Sponsors • Asset allocation, including multiple scenario analysis. • Random portfolios for evaluating the performance of funds, and for creating mandates free of a traditional benchmark. See Random Portfolios in Finance. Brokers • Portfolio optimization for trading and for research. • Random portfolios for research and for bidding on anonymous portfolios. See Random Portfolios in Finance.