OmniVest User Forums
User-Created Portfolios
Account Settings and Changes
25 replies, 2579 viewings
|
OmniVest User Forums
User-Created Portfolios
Account Settings and Changes
|
Last Activity 7/22/2015 8:11 AM 25 replies, 2579 viewings |
|
|
| Printer friendly version |
Mark Holstius Elite    Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL  User Profile |
When I made that most recent portfolio, I "fell back on" account settings that Steve & I have used previously by using "Alphabetic" trade selection when I ran all the Condition Testing, Strategy Selection, etc. I plan to do more testing using Price*Avg Volume Descending, etc. - but not until we have the automatic tools, as it takes quite a while. What I'd like to show is the effect of changing those settings once the portfolio is constructed, so I've run that same portfolio using all 7 Trade Selection Order criteria with no other changes to the account settings. (below - sorry, it's a long graphic) I've copied / pasted a "table" of the results at the top & all the complete equity graphs below. Notice the wide variation in results. This is why Steve & I have tried to stay with one set of account settings all through the process of building portfolios. It should be interesting to do more research once it's easier to accomplish... Mark [Edited by Mark Holstius on 5/11/2014 8:06 PM]  Attached file : Changing Settings 2.jpg (3379KB - 1193 downloads) | ||
| JimB Member Posts: 36 Joined: 3/7/2014 Location: Rogers, MN User Profile |
| ||
| Jim Dean Elite  Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
What do y'all think of an "Account Lab"? That is, a way of setting up several different account setting patterns then pushing a button, running the same Strat on all those AS permutations at once, then seeing the summary results like Marks table (click on any of them to drill down). | ||
| gbarber Veteran Posts: 282 Joined: 12/30/2012 Location: Pearland, TX User Profile |
| ||
| Steve Mayo Legend Posts: 414 Joined: 10/11/2012 Location: Austin, TX User Profile |
We all tend to "jigger the controls" looking for that little bit of extra performance in our strats/ports. In general, its great that we have this wonderful capability in OV to quickly try these changes, but the downside is that we may simply be over-optimizing. A relatively small change (filtering in/out only a few trades) can have a big compounded difference...but it's meaningless if that change is not something that will carry into future trading. How do we know if our systems are predictive (robust)? We don't have good tools for assessing robustness of our "systems" yet, but here's a simple technique that makes for a good stat. Just run your strat/port/account under each of the various sorting methods and average the return/drawdown across them (calculating a standard deviation and confidence interval would be even better). That helps eliminate some of the filtering error. A second technique, thoroughly discussed elsewhere, is to average returns over different time-periods/market modes. | ||
| kmcintyre Elite Posts: 890 Joined: 10/11/2012 Location: Portland, OR User Profile |
I like the idea of looking at the mean return of all sorting methods, as well as the min and max returns to get a sense of the robustness of an account (and it's collective portfolios/strategies). Great stuff. Also, tweaking other account settings a bit to ensure an account isn't highly sensitive to a small change in any particular setting prior to going live. A great smell test... Keith | ||
| Steve2 Elite Posts: 750 Joined: 10/11/2012 Location: Annapolis, MD User Profile |
I think this would be a real time saver. What I'd like to be able to do is have OV perform a sensitivity analysis on the account settings I'm interested in. For example, vary Max Strategies Trading the Same Symbol between 1 and 4, vary Min Average Daily Volume between 500K and 1,500K in 25K increments, vary Maximum Exposure % per Symbol between 10% and 35% in 5% increments, vary Trade Selection Order across all options, etc. Unfortunately, this could get pretty resource intensive requiring OV to execute tens of thousands of simulations. Steve | ||
| Jim Dean Elite Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
| ||
| Lompocian New User Posts: 3 Joined: 3/24/2014 Location: Lompoc, CA User Profile |
| ||
| John W Elite Posts: 654 Joined: 10/11/2012 Location: Sydney, NSW, Australia User Profile |
You've reminded me that the 'Alphabetical' setting is the only one without an up down direction, having Alphabetical Up and Alphabetical Down would complete the set. My real world OV trading typically buys the majority of symbols in the A to M range! Re settings - perhaps a user configurable set of (say) 4 settings pages that could be switched on and off when desired for more robust testing? | ||
| Bruce Britt Regular Posts: 81 Joined: 10/11/2012 Location: Louisiana User Profile |
As John W suggest, if the alphabetic-descending option was also available it would "complete the set", and complete your great experiment more fully. This is important work to help insure that the multiple strategy/list combinations we are now able to create remain robust. Hopefully the alphabetic sort option will be revised to include both ascending AND descending order in the near future. | ||
| tgrafa Regular Posts: 63 Joined: 10/11/2012 Location: Midland, Texas User Profile |
| ||
| TomTwiford Member Posts: 40 Joined: 12/20/2012 Location: Houston, Tx User Profile |
Best choices should include: 1. Ascending 2. Descending 3. Ascend or descend from a particular start letter, rotating through the alphabet. 4. Random order letter selection (most important addition) These choices, particularly #4, would insure great variation. | ||
| Jim Dean Elite Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
The random list should use the timestamp as a "seed" so the random series varies. | ||
| John W Elite Posts: 654 Joined: 10/11/2012 Location: Sydney, NSW, Australia User Profile |
[Edited by John W on 9/3/2014 8:42 AM] | ||
| TomTwiford Member Posts: 40 Joined: 12/20/2012 Location: Houston, Tx User Profile |
I fully agree with your recommendation. | ||
| kmcintyre Elite Posts: 890 Joined: 10/11/2012 Location: Portland, OR User Profile |
The downside I see for random ordering is that of repeatability. Each time the simulation is regenerated, different results will be displayed. I think that will confuse some people. Personally, I like price * volume descending for live trading. And I like to use all available sort orders to check for robustness. But I think a Market Cap filter or a PE filter could be interesting. Keith PS - I like Juan's RSI sort order. And Shares Outstanding. I'm sure great minds will think of many others. So how about another scriptable sort order? :-) [Edited by kmcintyre on 9/3/2014 4:12 PM] | ||
| Steve Mayo Legend Posts: 414 Joined: 10/11/2012 Location: Austin, TX User Profile |
That randomness is the whole point....to try to average out the stock selection bias. But, I get your point. It would have to be CLEARLY LABELLED that it is only to be intended for testing. What you are point out, though, is that few people understand the real difference between simulation results and what you are actually likely to get in live trading. | ||
| John W Elite Posts: 654 Joined: 10/11/2012 Location: Sydney, NSW, Australia User Profile |
I've stopped using alphabetical because too often I was just trading the early letters in the alphabet. Having reverse alphabetical reduces the likelihood we all trade the same stocks at the same time but then the symbols in the middle of the alphabet are still likely to be more lightly traded than those at the start or end of the alphabet. I really like Tom's list of 4 alternatives, it covers all the bases for live trading! A secondary possibility would be if several random simulations could be used to give confidence (or not) that a portfolio of strategies perform well in simulation. Perhaps Nirvana could extend such a random capability for live trading to also allow us to run 'n' number of simulations randomly to validate the performance of our favourite portfolios. [Edited by John W on 9/3/2014 6:57 PM] | ||
| Mark Holstius Elite Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL User Profile |
When you have enough capital, it has no effect - all trades would be taken. Mark | ||
| John W Elite Posts: 654 Joined: 10/11/2012 Location: Sydney, NSW, Australia User Profile |
OV sure has a lot of knobs, levers and gotchas (sigh)! Hopefully the random approach makes sense for anyone whose percent invested breaks their BP % regularly. | ||
| Geoff Veteran Posts: 180 Joined: 12/4/2012 Location: Byron Bay NSW Australia User Profile |
| ||
| Mel Member Posts: 20 Joined: 3/4/2014 User Profile |
Mel | ||
| Mark Holstius Elite Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL User Profile |
You can do various things to hit the 'ceiling' (higher allocation %, combining multiple Portfolios, etc), but whatever you do to increase the amount of time your account is hitting the ceiling the more often the 'ordered by' will kick in and have an effect. Mark | ||
| Eric Severance Member Posts: 23 Joined: 10/11/2012 Location: Incline Viallge, NV User Profile |
|
|
|
|
|
Administrator
Forum Moderator |
Registered User
Unregistered User |
 E-Mail this thread to a friend |
Toggle e-mail notification |
