Current location | Thread information | |
![]() ![]() ![]() ![]() ![]() ![]() |
Last Activity 10/7/2016 12:00 PM 23 replies, 1456 viewings |
|
|
Printer friendly version |
^ Top | |||
Steve2![]() Elite ![]() ![]() ![]() ![]() ![]() Posts: 750 Joined: 10/11/2012 Location: Annapolis, MD ![]() |
While there are a few things that need to be cleaned up in Beta Release 1, I found the Portfolio Wizard to be quite usable. My first experiment was to take one of my V2 static portfolios and create a corresponding dynamic portfolio that provides better performance using the same strategies. The static portfolio I chose had 23 strategies (20 RTM and 3 Trending). I loaded the strategies into Portfolio Wizard and did an initial analyzer run using one month periods, selecting 10 of the 23 strategies each period with all 21 EF's enabled. The simulation date range was 6.5 years. This took 3-4 hours to run (wasn't at the computer when it finished so don't know exactly how long it took). The CAR EF produced the best stats so I then did a series of analyzer runs using that EF and varying the number of strategies selected between 10 and 17 and varying the dynamic portfolio allocation within my test account. Each of these Analyzer runs took about 11 minutes. The best results were obtained when selecting the best 15 strategies each period with a portfolio allocation of 200%. The dynamic portfolio improved CAR from 94.5% to 118.5% over the static portfolio. MDD was reduced from 37.6% to 34.2%. % Invested decreased from 167.4% to 155.5% and TPM was reduced from 106.9 to 68.5. So, I've convinced myself that using equity curve analysis to dynamically select strategies will definitely improve simulated results which should translate into improved real results. Thanks Nirvana for another great tool! Steve [Edited by Steve2 on 8/3/2014 3:35 PM] | ||
^ Top | |||
Steve Mayo![]() Legend ![]() ![]() ![]() ![]() Posts: 414 Joined: 10/11/2012 Location: Austin, TX ![]() |
Great post, Steve. Thanks for sharing that. I was trying to do something almost identical today for Ed's new user presentation but couldn't get it finished in time. Happy to see that it works. :-) A further idea to try now would be to save that portfolio, and then tweak the %Invested on the portfolio page. You might be able to drop that MDD a bit more while still getting the same original return. | ||
^ Top | |||
smartprofit![]() Member Posts: 14 Joined: 3/2/2014 ![]() |
Is omnivest pro a one time fee? or is it going to be a yearly fee. | ||
^ Top | |||
Steve2![]() Elite ![]() ![]() ![]() ![]() ![]() Posts: 750 Joined: 10/11/2012 Location: Annapolis, MD ![]() |
OV Pro is a one time fee and, in my opinion, well worth the money :) Steve, I try to optimize for max number of profitable quarters (i.e., consistency of returns) rather than min MDD but I suspect there is some correlation between the two. This is on my list to do this afternoon. Will let you know how it goes. Steve | ||
^ Top | |||
Mark Holstius![]() Elite ![]() ![]() ![]() ![]() Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL ![]() |
Steve(2)... Have you tried optimizing for max # of profitable months? That could be interesting with the ECA. Mark | ||
^ Top | |||
Steve Mayo![]() Legend ![]() ![]() ![]() ![]() Posts: 414 Joined: 10/11/2012 Location: Austin, TX ![]() |
Steve, Looks like the ECA data downloader on OmniVesting.com got broken when we went to the server. I'll try to get that working for you. That would allow you to more easily calculate performance numbers. I've been working on a set of performance metrics. Consistency of return and better analysis of drawdown are high on my list. :-) | ||
^ Top | |||
Steve2![]() Elite ![]() ![]() ![]() ![]() ![]() Posts: 750 Joined: 10/11/2012 Location: Annapolis, MD ![]() |
Finally getting back to this... Mark, sadly, I'm not setup to easily do this. The account analysis tool that rokfreek and I developed is setup to do quarterly analyses and not easily changed. Attached is the quarter by quarter performance comparison between the static and dynamic portfolios. This analysis is based on realized P/L and draw downs, not the MTM numbers you see from OV. The analysis is interesting. While the ending equity for the 6.5 year sim was much better for the dynamic portfolio, the quarterly analysis shows that this was due to the dynamic portfolio significantly outperforming the static portfolio is the first two years while not performing quite as well as the static portfolio in the last 4.5 years. I was hoping for consistently better performance throughout the simulation period. One factor may be the way trades were filtered out in the dynamic portfolio because it got to be significantly larger than the static portfolio ($1.1M vs $490K after year 2). So, I'll go back and do a set of one year sims through the date range which should provide a more realistic performance comparison. Stay tuned... Steve ![]() | ||
^ Top | |||
Steve Mayo![]() Legend ![]() ![]() ![]() ![]() Posts: 414 Joined: 10/11/2012 Location: Austin, TX ![]() |
Steve, Thanks for the analysis. Quick question: are you pulling the transaction data from the PW History page or from the regulator OV trades/portfolio page? If the later, any chance you are collecting data from a period that extends beyond what PW used? Steve M. | ||
^ Top | |||
Mark Holstius![]() Elite ![]() ![]() ![]() ![]() Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL ![]() |
Thanks Steve(2)... Steve & I can fully appreciate all the hard work in developing a spreadsheet like that! Nice of you to share the results. We've found the same thing to be true (early good results with some Ports), and that's why we're focusing on finding "straighter" curves and better CALMAR vs just high Equity, like the top chart vs the lower one in this snag; Mark [Edited by Mark Holstius on 8/6/2014 3:15 PM] ![]() | ||
^ Top | |||
Steve2![]() Elite ![]() ![]() ![]() ![]() ![]() Posts: 750 Joined: 10/11/2012 Location: Annapolis, MD ![]() |
Steve M, The data was pulled from the regular OV account page (simulated historical positions). One extra day of data was pulled beyond the end of the range covered in the PW run. Since strategy selection takes place on the first of every month, this shouldn't be an issue. Steve | ||
^ Top | |||
Steve2![]() Elite ![]() ![]() ![]() ![]() ![]() Posts: 750 Joined: 10/11/2012 Location: Annapolis, MD ![]() |
Attached are the one year sim results. They exhibit the same performance characteristics as the 6.5 year sim so trade filtering due to account size is not an issue. Note that these are MTM results, not realized results like the quarterly analysis spreadsheet. ![]() | ||
^ Top | |||
Mark Holstius![]() Elite ![]() ![]() ![]() ![]() Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL ![]() |
Steve(2) - that's the same file (attached) as the 1st one... Mark | ||
^ Top | |||
Steve2![]() Elite ![]() ![]() ![]() ![]() ![]() Posts: 750 Joined: 10/11/2012 Location: Annapolis, MD ![]() |
Oops, that post has been updated with the correct file. | ||
^ Top | |||
Mel![]() Member Posts: 20 Joined: 3/4/2014 ![]() |
It would be a big help to me if the output of each run was automatically put into a folder in HTML form, for me to rename, delelte, copy into a spreadsheet or whatever. Last night I set a up a long run, went to bed, and this morning was greeted with a login prompt. No output from the run. I would like runs to finish after a disconnect and save the data in the folder. Mel | ||
^ Top | |||
Geoff![]() Veteran ![]() ![]() ![]() ![]() Posts: 180 Joined: 12/4/2012 Location: Byron Bay NSW Australia ![]() |
Mel, that is a great idea. I am surprised though that your OUTPUT is not there. I setup an overnight run too, but once it got started, I logged off and shutdown my PC. The next day I logged back in and opened up the OUTPUT to see the results waiting for me. One thing worth considering though is checking that you have the same 'Settings' selected as was used in the 'run' in question. If you have different settings selected then there will be no OUTPUT until you change back to the 'Settings' used. | ||
^ Top | |||
Mel![]() Member Posts: 20 Joined: 3/4/2014 ![]() |
If there is going to be no output, I would appreciate a "nothing to do" message instead. When I went to bed, it was crunching away, with visible progress. My real problem is the extra paperwork trying to keep track of what I have done, and what the basis of the results were. Once you get beyond a few runs, it is confusing. I would like the output in my folder to have the results, with the strategy list enabled as a page, and with another page with setting of the account and portfolio, along with any other settings that were the basis of the report. This would significantly reduce the workload for me, and keep me from having reports that I can't later reproduce. Mel | ||
^ Top | |||
Steve Mayo![]() Legend ![]() ![]() ![]() ![]() Posts: 414 Joined: 10/11/2012 Location: Austin, TX ![]() |
ECA is supposed to save the outputs. Nirvana had some load-balancing issues when they moved it to the production servers -- the calculation server was getting overloaded and some outputs were getting lost or not run at all -- but I think that got fixed on Tuesday or Wednesday. If its not there yet, there will also soon be a means to specifically delete old outputs. The next step is to add the ability to parse EF parameters over specified ranges -- that will generate thousands of outputs, so we definitely will have to have auto-save and manual delete capability for that. :-) | ||
^ Top | |||
Mark Holstius![]() Elite ![]() ![]() ![]() ![]() Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL ![]() |
FWIW, I had the same problem this morning MEL & Geoff... I could see the Output data after a long run overnight, but as soon as I did something (trying to save a Portfolio) that forced me to re-sign in, then the Outputs went to "No Records Found". I documented it with screen shots to Ed @ Nirvana, so I hope they address / fix it today. Mark | ||
^ Top | |||
TimW![]() Member ![]() Posts: 38 Joined: 3/1/2014 ![]() |
I have never seen the ECA save the outputs, and I haven't changed any settings. For me, once I save a portfolio or do any account action in other OV pages the Outputs page becomes empty. It's especially maddening when you run multiple EFs at one time. All will be lost once you save one of them. This is my #1 item in this thread: http://www.omnitrader.com/currentclients/omnivestforum/thread-view.asp?threadid=7298&posts=2 | ||
^ Top | |||
Steve Mayo![]() Legend ![]() ![]() ![]() ![]() Posts: 414 Joined: 10/11/2012 Location: Austin, TX ![]() |
That's why it's just a beta release. This is extremely complex software and the environment now needed to support hundreds of users is even more complex. Nirvana will get it working correctly, but it just always takes longer to squash bugs that anyone ever expects. | ||
^ Top | |||
Mark Holstius![]() Elite ![]() ![]() ![]() ![]() Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL ![]() |
Tim... I saw your other post and was waiting to see if anyone at Nirvana replied. As you may have read, I had the same problem ("No Records Found") when I had to do something on another page or sign in again after doing a run overnight - but (so far) only if the run was done on a previous day. I did another run involving 6 Eval Functions this morning and have signed in a couple of times, saved Portfolios in PW, gone to other pages, etc - and my data and statistics has remained on the PW pages. They know it has to be fixed, but until they do I just wanted to let you know that it seems to work OK when everything is done on the same day...??? Thanks for the report, Mark | ||
^ Top | |||
TimW![]() Member ![]() Posts: 38 Joined: 3/1/2014 ![]() |
@Steve - Understood, that's why I was reporting the bugs here :) @Mark - Thanks for the info, I have not seen any response from Nirvana yet. I will try a couple of runs tonight to see if things might have changed. Frankly I'm mostly worried about the issue where the positions in the Trades page do not match the Accounts page (see my other thread and screenshots). I don't know about other users, but I do not spend a lot of time verifying that the site is displaying valid data for CAR, trades, draw down, and the myriad of other things that we all use in our decision making. | ||
^ Top | |||
gbarber![]() Veteran ![]() ![]() ![]() ![]() ![]() Posts: 282 Joined: 12/30/2012 Location: Pearland, TX ![]() |
I used strat lab to define a large set of of strategy/list combos that had very good results in period from 1/1/2014 to 8/21/2014. I gave those to PW and got some pretty amazing results as shown in output table. Also has a nice equity curve as shown in files attached. These numbers are pretty enticing. The trades it gave me today were 3 shorts and 1 long. Overall it was a small loss today with a flat market. Just posting to show another case of the high potential of this tool. Sure hope it has the predictive powers that ya'll worked hard to put in there. [Edited by gbarber on 8/22/2014 3:27 PM] ![]() ![]() ![]() | ||
^ Top | |||
gbarber![]() Veteran ![]() ![]() ![]() ![]() ![]() Posts: 282 Joined: 12/30/2012 Location: Pearland, TX ![]() |
To answer Sean's questions: what is your process for finding good strategies in strat lab? I have not added any of my own strategies but I did create a lot of custom strategy list combinations with strat lab. To do this I used the setup in the first snag in the attachment. I used test one list against all systems, hit recalculate, and then sorted by CAR. I am interested in a high CAR but most of all I want a low drawdown. So went down the list looking for that combination (i.e., a high CALMAR). I would forgo a high CAR if the drawdown was also high. I enabled each attractive one I found and then hit 'add selected'. I did this for each list I was interested in. I used primarily the dynamic lists supplied by Nirvana but also included Kipp and K510 and a couple other static lists I created (from Dow Theory and Personal Finance). In general the dynamic lists performed the best but the Kipp list strategies were also good performers. I suspect this is because my date range is rather short. I picked 01/01/2014 to current because I wanted to get good performers for current market. I have seen others are using long date ranges to include many market types including the big bear in 2008. I probably will do that later to get a 'set it and forget it' sort of strategy and portfolio selection but I wanted to start with something that works well in the current market expecting I will have to change it from time to time. I then went to PW and brought up the strategies tab. This showed me all the custom strategies I had created in strat lab. Again I sorted by CAR and again I went down the list enabling the high CAR and low MDD strategies and not enabling the high CAR and high MDD strategies. I made sure I had at least 30 strategies enabled to give PW plenty to work with. Then I enabled all of the EVs on the evaluation tab. I do not have any custom evaluation formulas. Then I hit 'analyze' on the dashboard tab. That gave me an output with all of the EV functions listed. The setup I used in dashboard is in the attachment. The output I got is in the attachment of the previous message. I had a very good selection to choose from. So I picked the best and started using it. what account settings did I use. This is in the attachment. what date range? I used 01/01/2014 to present ![]() |
|
|
Legend | Action | Notification | |||
Administrator
Forum Moderator |
Registered User
Unregistered User |
![]() (un)/Freeze thread | |
Toggle e-mail notification |