reduce risk
What is RiskCog?
By Matthew MacClaryJanuary 1st, 2010
Executive Summary
RiskCog is a portfolio construction methodology that aims to reduce the risk of a financial portfolio given a target compound rate of return. The RiskCog approach measures return based on compound (geometric returns) and measures risk based on time domain concepts such as max-drawdown. A RiskCog optimal portfolio is the portfolio with the least risk chosen from the set of all portfolios that have acceptable compound returns. Standard deviation and other statistical quantities are not used in the RiskCog approach.
How to Measure Return
The RiskCog approach measures the return of financial assets using compound rates of return. This measure is chosen over alternatives such as "average annual return" because the compound return correctly predicts the terminal account value even for volatile assets.
How to Measure Risk
When implementing RiskCog, the risk measure should match the portfolio owner's definition of risk as closely as possible. Example measures that may be chosen include:
- Worst Year
- Worst 12 Month Return
- Worst 12 Month Benchmark Lag
- Maximum Open Draw-down
- Ulcer Index
Additional risk measures not included on the list may also be employed if desired. An example of an infective risk measure that would not be used with RiskCog is the variance or standard deviation of the average period returns for the portfolio.
How to Choose the Best Portfolio
RiskCog optimal portfolios have the lowest risk for a given return. A portfolio in this context means a weighted list of financial assets. Conceptually you can imagine searching the set of all portfolios with compound returns equal to or greater than a return target. The portfolio from this set with the lowest measure of historical risk is the RiskCog optimal portfolio.
Comparisons to Mean Variance Optimization
Mean-Variance Optimization (MVO) is a method rooted in "Modern Portfolio Theory". Like RiskCog MVO is intended to tell a portfolio owner how much of each type of financial asset they should own for the best risk-adjusted long term results. RiskCog was created to address the short comings of MVO.
The inputs to MVO include the mean returns and standard deviations for each asset that could be included in a portfolio. Also the correlation matrix for the assets is required for MVO. The inputs to the RiskCog optimizer are the historical returns of each asset, and a function to compute the risk and return of a portfolio made up of arbitrary weights of each asset.
| Mean Variance | RiskCog | |
| Risk Measure | Statistical variance of returns | Worst Drawdown (or related measure) |
| Return Measure | Mean Period Return | Equivalent Compound Return (CAGR) |
| Portfolio Selection | Maximum return given variance target | Minimum risk given return target |
| Optimization Algorithm | William Sharpe's MVO Algorithm | Stochastic Hill Climbing |
| Time for Optimization | Faster | Slower |
| Quality of Portfolio | Lower | Higher |
There is an article available on this website which explores the differences between portfolios created using the Minimum-Variance approach and the RiskCog approach: "RiskCog Beats Mean-Variance Portfolio in 2008"
Why Name this Simple Concept "RiskCog"?
I am trying to popularize a more coherent method for portfolio construction by naming it "RiskCog" and creating a useful website. The purpose in doing this is to give the investing public a superior alternative to MPT - an alternative designed to work more successfully in real life.