Geoff
 Veteran
    Posts: 180
Joined: 12/4/2012
Location: Byron Bay NSW Australia
User Profile |
I have been looking at possible ways to determine the current performance status of a portfolio to help with the process of determining which portfolio(s) to use for live trading. Yes I realise this idea will probably be made redundant with the release of Portfolio Balancer but, if it has merit then those who aren’t Pro users may find it useful.
What is available in OV now to evaluate current port performance status?
1. Portfolio page stats. These provide an instant snapshot of how ports are performing for a given sim range, all other settings remaining constant for the purposes of this exercise.
2. Equity Curve. This provides good information but is low resolution. For example, a port may be performing at a high level say CAR=70% but has had declining performance over the short term, say 1 to 2 weeks down to say CAR=30%. Still good BUT, this trend might not be readily detected by just observing the EC, IMHO.
Importantly neither of these readily shows the short term performance trend of an individual port.
So, what I have been looking at is the Moving Performance of a port for a given simulation date range. For example, set the sim range to 3 months, record the performance stats, then move the End and Start Dates back one day, and record the stats, then move the dates back another day and record the stats and so on until you have sufficient stats to analyse. When you chart the results you get a line that shows the 3 Month Moving Performance Indicator or 3M MPI as I call it.
Please NOTE, this is most likely not a new idea or a unique one, and may or may not have merit. I am providing the data that I have extracted, so that those who are interested can play around with it and hopefully provide feedback and ideas.
The data attached (see links below) are of 35 portfolios and their respective 2MMP, 3MMP, 6MMP and 12MMP results going back 270 working days, that’s about one year. Each MMP is in its own separate Excel File. At the top of each Sheet I have also set up Data filters so that individual Portfolios data can be viewed and charted.
.
I have set up charts to graph the data. The charts will only work if one, and only one portfolio is selected at a time in the ‘Portfolio’ drop-down filter list.

............
.
Here is an example of a chart. The charts may look complicated at first but, take your time and look at the legend, and you will see that they are straightforward line charts with some moving averages applied.
I like the 3MMP results myself; the charts do show that port performance does cycle with market conditions and IMHO they do reflect the port’s current performance trend. Following on from Steve and Mark’ great work, could it be possible to set some switching rules to select ports for trading using the MPI?
.

.
The original file was too big apparently, so I made separate files out of each of the Sheets in the original Excel Workbook and uploaded them. (See below).
[Edited by Geoff on 8/27/2014 5:52 PM]
Attached file : Port Sort #1.png (461KB - 444 downloads)
Attached file : Port Sort #2.JPG (117KB - 432 downloads)
Attached file : 2MMP Analysis - GJS Ver 1.04.xlsx (1593KB - 212 downloads)
Attached file : 3MMP Analysis - GJS Ver 1.04.xlsx (1595KB - 236 downloads)
Attached file : 6MMP Analysis - GJS Ver 1.04.xlsx (1724KB - 226 downloads)
Attached file : 12MMP Analysis - GJS Ver 1.04.xlsx (1624KB - 221 downloads)
Attached file : Beta 300 3MMP example.JPG (154KB - 349 downloads)
|
gbarber
 Veteran
     Posts: 282
Joined: 12/30/2012
Location: Pearland, TX
User Profile |
This looks very interesting. It gives a picture of stability and range of performance of a portfolio over a range of time. This could be boiled down to a set of stats such as mean, median, variance, range etc. That would be a valuable metric to use in picking portfolios. It also could be used to define a portfolios performance over a set of known market types and thereby categorize the portfolio. If one made use of stats regarding market characteristics (e.g., VIX, S&P 500 price rate of change, and others) one might be able to draw some conclusions related to correlation of market stats to predictive ability of the portfolio (I am not a statistics guy so this may just give irrelevant info (never repeats or only by chance)). This could be a very valuable addition to the set of tools available in Ovest pro. Programming probably not too difficult but use of servers would go up significantly. I suggest adding this to the growing list of valuable additions to Ovest. Great job Geoff.
|