Hi Guys,
A eureka moment for me that I wanted to immediately share with you. I just discovered a website that easily and flexibly allows a user to Backtest the effectiveness of his portfolio assembled from an endless array of investment categories.
This site is a warehouse of actual real world returns. It uses historical investment data as its primary source. It is definitely not (I repeat, NOT) a Monte Carlo simulator. The Link to this interactive resource follows immediately:
http://www.portfoliovisualizer.com/ViewHistoricalReturns The tool has been assembled by Portfolio Visualizer. The title of the referenced section is “Backtest Portfolio Asset Class Allocation”. It is one of several tools that Portfolio Visualizer offers. They also feature a Monte Carlo option that I have not examined yet.
Input is conveniently entered as a portfolio holding percentage. Input options exist. One such option is to define a starting portfolio value without change, or to define either an annual withdrawal or contribution schedule.
Output includes cumulative end wealth, annual compound return, standard deviation, Sharpe Ratio, and Sortino Ratio. With simple clicks the output can be changed to display portfolio return, annual return, and rolling returns. It is a goldmine of historical performance that can be exercised to explore past performance of various portfolio constructions.
This is super stuff. Please visit the site and give it a test run; no, give it a number of tries to challenge all its numerous Input and Output options. I have not done that, but I am excited and optimistic over this discovery. It significantly adds to my portfolio examination toolkit. Consider adding it to your bench-tools too.
The site provides an option to test portfolios with specific mutual fund and ETF inputs. I have not even accessed these screens. You might want to do so. Additionally, the site provides Monte Carlo and Asset Correlation sections. I have not examined these at this juncture. It’s play time.
Please give it a try. At this moment it is all free. Let us all know what works, and especially if anything is suspect on this resource. Since I have attempted only a few exploratory cases, independent feedback is essential. Good luck and good investment planning using this tool.
As an aside, I’ll be leaving on a cruise within the next few days, so I would appreciate a quick feedback even if it is incomplete.
Best Regards.
Comments
Allow me a quick update. I just accessed the Portfolio Visualizer version of a Monte Carlo simulator. It has some attractive features and is extremely fast. I'm sure I will explore its full capability when I have more time.
The code permits inflation considerations. More importantly, it features a Fat-Tail statistical distribution. The code runs 5000 or more random trials per case. It appears to be great for parametric studies.
It's output is simple, informative, and clear.
You might want to give this functional tool a test run or two.
Best wishes.
Hi David,
It appears to be total return data.
I appreciate your find.
prinxx
There is no value to doing performance analysis for the future whether via back testing or not. What exactly do you intend to do based on that result? That is the problem.
Beta exposure and asset allocation along with what the markets do in the future determine what your portfolio does. There is no trend or momentum persistence over long periods of time to exploit that or give the future expectation any validity. The best you can do is to pick a rough amount of beta exposure with hopes of making historical returns for that exposure (in broad terms for the asset class not specific funds) and use some stress testing tools to understand what some historical events IF it happened again (with no assumption they will) would imply for your portfolio to see if you can withstand that shock financially and psychologically or be better prepared to ride out such shocks.
“History doesn't repeat itself, but it does rhyme.”
― Mark Twain
For example, you can stress test a portfolio to see what would happen under a similar specific circumstance that happened in the past. This makes no prediction that such an event would happen or not happen. It simply says, if the same specific situation were to happen again (as modeled), what might happen to your portfolio. This allows you to understand the impact of such events and guage whether you will be OK with such an impact especially worst case scenarios. But it gives no guarantees on whether such an event will happen or not. The more specific the event being modeled, more valid the model for that event. Things get complicated if you start combining different events because the number of ways you can combine them starts to grow non-linearly. So the validity of the result of such a combination becomes less and less.
Performance testing over some arbitrary time period in the past which is quite different from the above is testing for an arbitrary sequence of such events as happened in that time period and is valid only if a sequence of such events happened the same way. This is where the validity breaks down. Even Monte Carlo simulations don't begin to cover the huge set of possibilities let alone back testing for the same set of events assuming similar sequence of events. So what are you going to do with that result?
Both are use of historical data but it doesn't mean they have same validity or practical value.
Physics has the same problem when you consider multi-body problem for interactions. The problem isn't how the model for interactions between two bodies was arrived at but of modeling simultaneous interactions between a large set of bodies. Schrödinger's equation is simple for Hydrogen atom but explodes for anything more complicated.
"A gold mine is a hole in the ground with a liar standing next to it"
-Sam Clemens
FundAdvice had the highest CAGR at 7.54% with SD of 12.41
Harry Browne's Permanent Portfolio came in at 6.69% with SD of 5.53
60/40 stock/bond came in at 5.32% and SD of 11.6
In a shorter time set from 2008 - 2013 the standings changed
Harry Browne's portfolio came in with a CAGR of 5.95% with a 6.66 SD
FundAdvice ultimate buy and hold came in at 5.15 CAGR and a SD of 15.65 (I think this portfolio benefitted from the 2000-2003 period when REITS and Small Caps performed better than the overall market).
60/40 came in at 6.75% with a SD of 14.88
From 1985 to 2013
Harry Browne CAGR 7.69%
FundAdvice 10.9%
60/40 9.83%
I agree with CMAN's post that its good for entertainment value.
Various time periods encompassing various markets would change the numbers and each market drawdown isn't the same. Maybe could be a tool to help match one's investment horizon with one's time horizon.
thanks MJG for posting, I will prolly play around with it some more
The Portfolio Visualizer website truly offers opportunity. Opportunity to learn.
There are so many potential and practical uses for this fine website that it is hard to characterize its full utility and scope in a few words. So I’ll settle for a lesser goal and suggest just a few specific applications.
Any such usage must always be accompanied with the standard cautionary warning that in a dynamic, nonlinear system, specific outcomes are never totally predictable. The Chaos whiz-kids could endlessly pontificate on this matter. But Chaos is not randomness. It borders on the threshold of randomness and it develops in a semi-controlled, non-arbitrary manner.
A famous saying, often but not always attributed to Mark Twain, captures the spirit of the uncertain future: “History does not repeat itself, but it Rhymes”.
Given the unknowable future, an investor must exercise judgment in recognizing that rhyming proclivity and a certain market rhythm. A knowledge of market history is necessary to achieve this level of understanding.
For the global climate change debate, it’s critical to know that the sun-spot activity level has a periodicity of about 22 years. For investors, it is important to know that 22 economic recoveries have occurred since 1904 with an average and median life span of 3.8 and 3.1 years, respectively. This type of knowledge allows an investor to develop a feel for market risk.
In general, the historical returns for the top tier of investment classes and for the next lower order investment categories serve as a guideline for potential, but never guaranteed, future market rewards. The category data provide guidelines for a logical, long-term expectation level from these various groupings, which is especially useful when assembling a portfolio asset allocation plan. Surely if you seek a 6 % to 8% portfolio annual return, a 100% bond portfolio simply will not suffice.
Both baseball and investing are awash with data. Properly interpreting this data, and that also means respecting its limitations, will make anyone a better informed investor when making investment decisions.
In baseball, the Moneyball that Billy Beane deployed was initially rejected by the baseball establishment. It is now the operational rule for all baseball. In the financial world, very little statistical awareness was applied in the mid-1950s. Today, the reverse is true, except for a few old diehards. You get to choose your own pathway.
To develop a feel for the sensitivity of an equity/bond return/risk tradeoff, just play a few what-if games with the referenced asset class allocation tool. Vary the percentages and see how sensitive the overall returns and the standard deviations are to the mix. There are never any free lunches. Vary the study timeframe to isolate sensitivity to that parameter. If you don’t like the historical database, invent your own preferences and use them as input. Experiment liberally!
These types of parametric explorations will shorten your learning period. You will develop a sense of what factors are influential and what are noise. The what-if game scenarios will likely make you a better investor whatever your goals.
You can use the Monte Carlo simulator that is also available on the website to project retirement portfolio survival rates for a variety of market circumstances. These type of studies can be used as positive feedback loops to adjust savings plans and investment risk requirements if you are in your accumulation phase (before retirement).
You can challenge the robustness of your candidate portfolios by checking survival rates for a normal Bell curve distribution and next stress testing it against a Fat Tail model. A version of a Fat Tail model is incorporated within the referenced Monte Carlo code as a user option.
I find it somewhat amusing that some members of the MFO family elect to discourage statistical applications, yet in the same posting endorse stress testing. Stress test against what, if not against some target specification gleaned from history. Statistical history serves to guide any meaningful stress test.
I encourage you to fully exploit the capabilities of this attractive website with its comprehensive set of investment tools. These types of analysis will help you formulate a portfolio that satisfies your return requirements while also serving to measure its risk profile. Of course, there can never be an absolute guarantee given the uncertainty of future exogenous events.
I wish you more informed and more confident investment decision making. You need confidence to stay the course when the potholes appear. And they will most certainly appear.
Best wishes to all and thank you for your participation.
Investors would be better off worrying about what is working in the here and now than the distant past.
Regards,
Ted
Mark Twain Bio
Thanks for the referenced video.
If you ever get the chance, visit his river boat style home docked at Nook Farm.
marktwainhouse.org/
I can add to your point regarding validity of performance testing. The
results of such test are usually biased.
The issue is that we usually test a portfolio selected based on historical returns. That means we already selected MF that performed well in the past and passed our mental stress test. (It is very unlikely our portfolio has funds that crashed in the past. )
For that reason back testing simulation of such portfolio will produce the results that are better than actual ones we can face in the future for that portfolio.
Our portfolio may have funds that will crash in the future and back testing practically excludes that possibility as we test only good funds from the past.