OmniVest User Forums
User-Created Portfolios
The Learning Curve
8 replies, 1222 viewings
|
OmniVest User Forums
User-Created Portfolios
The Learning Curve
|
Last Activity 7/22/2015 8:11 AM 8 replies, 1222 viewings |
|
|
|
| Printer friendly version |
Mark Holstius Elite    Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL  User Profile |
People have discussed problems with naming and difficulty using Conditions, but I’m just waiting for the finished products and “playing” with the “parts” as they come available. With that in mind, I don’t want to spend a lot of time testing / explaining anything yet. I’m trying to be patient, looking forward to the way these pieces will work together in the near future, and just posted what I’ve found with limited experimentation because I’m excited about what can already be done, trusting that it will only get better. With that in mind, I’ve come across a few things that might be of interest… I’ve built a few simple portfolios based on the strategies after applying Conditions – both the canned ones and a couple based on equity slope. Nothing fancy, just seeing what can be done. I’ve used the “eyeball method” to combine strategies in Strategy Lab after sorting them on MDD, CAR, or % Wins to see the differences. The following combination of strategies points out something Steve’s mentioned before – that one strategy can “hijack” the equity for an account… Here’s the result using 5 strategies from a run using Classic Trend and ELS – nice return, DD, and high percent invested (197.3%) – one of the things we’re looking for; But - what’s actually happened is that the Trending Strat (T1) has taken over (using a lot of the equity) – and here’s what happens if you drop it out of the portfolio; Notice that the Grey area disappears (it marked the date the T1 “went active”), that the equity is better, AND the MDD is half what is used to be (7.1% vs 14.0%) - giving a much better CALMAR (14.8 vs 7.2). So, everything is still inter-related - and focusing solely on a single “goal” (whether CALMAR, CAR, or % Invested) will get you great portfolios, but maybe not the “outstanding” ones without some extra work. The good news is that in the near future it’ll take MUCH less time to build those portfolios, either targeted to different situations or good over multiple markets – and then switch them based on various statistics. Also, I’m not sure any of us fully realize the potential of the ability to add Conditions and then do a “Hard” stop and start to trades… that should be a whole new (& very enjoyable) ballgame. BTW - here’s a comparison of that simple 5 strategy portfolio with and without the Classic Trend applied. (I had to keep the R17-B-ELS in both since that’s only “created” using the Strat Lab); I like that extra 40% / year… Mark [Edited by Mark Holstius on 5/7/2014 6:32 AM]  Attached file : Conditions Compared.jpg (595KB - 687 downloads) Attached file : Conditions 5 Strategies.jpg (619KB - 676 downloads) Attached file : Conditions 4 Strategies.jpg (636KB - 696 downloads) | ||
| Jim Dean Elite  Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
This points out a variety of things to keep in mind. One technique that you used which seems sooo obvious in retrospect, but that I've failed to employ in my searches, is that of "backing out" a strat to measure its impact, AFTER building a group of them - ie a "sensitivity check". I've tended to work more on the "should I add this new one" question, and haven't gone back through them afterwards, to REconsider which of the earlier-added ones might not have ended up meshing well with the whole. That one methodology you pointed out will help me a lot! You've put a lot of attention on maximizing percent invested ... obviously this is a core tenet behind OV's potential ... Would you care to elaborate more on any (manual) methods you've used to help find good sets of strat's that "mesh well" to create higher avg %invested? I've been requesting a "correlation of win-trades" metric output column for that purpose, but without it, I'm at a loss as to how to manually find an optimal mix. Also ... you've focused on one year ... do you do that to save recalc time, or to simplify the problem, or for some other reason? Thanks again! [Edited by Jim Dean on 5/7/2014 7:55 AM] | ||
| JimB Member Posts: 36 Joined: 3/7/2014 Location: Rogers, MN User Profile |
| ||
| Steve Mayo Legend Posts: 414 Joined: 10/11/2012 Location: Austin, TX User Profile |
With Condition manager, a wider range of trading strategies, and the ability now to use alternative asset classes (gold, real estate, bond, commodity, foreign ETFs, etc.) via the custom lists, I think the concept is worth revisiting. Hopefully, we can get that functionality (incrementally adding the next strat based on non-correlation) into Port Builder. Jim, your comment about a post-build "rejiggering" of the strat mix after an iterative successive build is a good one -- the problem is that it's difficult, with any programmatic accuracy, to know WHICH strat to switch out. And with high-TTM systems like the RTMs, every little change can make a big difference. Even when the strat you are swapping out doesn't seem to be making a bunch of trades, what you don't know is what will happen when the other strats step up to fill that newly available daily equity. I think the better approach might be to come up with better tools and metrics for evaluating the robustness of our ports -- things like Monte Carlo and non-parametric statistics that are, unfortunately, hard to calculate and even harder to interpret. Yes, we can rejigger to get a better equity curve on a simulation, but are we in-effect just"optimizing the simulation" to select out/in certain trades that are not representative of what the system will do in real-world trading? It's that big standard deviation/non-Gaussian distribution (bigger tails?) problem again! Is it better to have a system that can generate 150% in a year when the planets all align, or one that has a much lower potential but is more predictive? I think we'll need to look less at equity curves and start looking more at distribution curves, albeit not as much fun as seeing your simulated account grow to $1billion in 14 years! :-) | ||
| kmcintyre Elite Posts: 890 Joined: 10/11/2012 Location: Portland, OR User Profile |
I did a lot of work with v1 when all we had were canned RTM strategies. It was valuable and definitely sold me on the OV concept. But it was simply too time consuming (for me anyway) to keep turning the crank when the product was in such a state of flux. Besides, the work I had done offered enough promise that it was worthwhile (IMO) to watch the portfolios work for a while, while monitoring new OV features at a "survey" level. The big piece I'm waiting for is Portfolio Wizard. I particularly hope it includes the ability to solve for user-defined OScript formulas. My sense is that OV is very close to another quantum step in performance. I'm very excited at the prospects. I'm also very thankful to Ed, Steve, Mark, Jim, and others who keep posting such valuable and informative posts re. their experiments. Very encouraging... My only apprehension to date is that the real world has not rewarded me as well as my v1 testing portended. Steve's last post rings so true. The market is not based on scientific laws and certainly is not as predicatble. For me, robustness is very high on the list of metrics I care about. Unbelievably good CAR metrics that indeed prove unbelievable with the passage of time make for an entertaining journey, but not necessarily a highly profitable one. So Cheers! to a new round of tools, another hunt for the perfect compliment of tuning parameters, and another year of monitoring the results! Let the grail chase continue! Keith | ||
| Mark Holstius Elite Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL User Profile |
Focusing on one year was just a way to limit the variables while I was experimenting. I like to isolate one input and then run it through a bunch of iterations to see which things are affected & how (brute force vs Steve's finesse). It's actually nice we have so many different types of markets the past 8-10 years to test against (or over). Mark | ||
| Jim Dean Elite Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
| ||
| John W Elite Posts: 654 Joined: 10/11/2012 Location: Sydney, NSW, Australia User Profile |
I've often found that REMOVING strategies after you've built a portfolio makes a significant difference. My process is to build (say) 5 good strategies then TAKE OUT each of the 5 sequentially, record the portfolio performance difference, add the strategy back again, take out the next strategy, record, add it back and so on. Once this task is complete its often the case that one of the strategies materially deteriorates the performance the portfolio, and this strategy is removed for an immediate portfolio performance lift. This then makes way for (say) one or two more strategies that were discarded in the first build so that the final portfolio of 6 or 7 is superior to the original portfolio of six. The process I manually use is: 1. Decide how many strats to be built and the metric of choice for the initial portfolio (e.g 5) 2. Decide a cutoff level of significance for the metric to decide if a strategy from the first build should be removed e.g. portfolio CAR/MDD improves by 15%. 3. Decide how much any additional strategies to be added should increase portfolio performance (e.g. any additional strategy must increase portfolio performance by 10%) 4. Decide the max number of strategies allowed in the second build (e.g. 7) 5. Build the first portfolio (e.g.5) successively with each next best strategy 6. Take out the significantly worst performer (e.g. take out 1 to reduce portfolio to 4) 7. Rebuild by adding one or more that significantly improve the metric of choice (e.g. two more strategies each improved the portfolio CAR/MDD by 10%, portfolio is now finalised with 6 strategies) It takes time but it does result in the final portfolio having significantly improved performance compared to the initial portfolio. Many of these steps appear to be already in the process under discussion for automation, perhaps a final cut removal of the worst performer and addition of one or two better strategies could be considered. | ||
| BrianD Legend Posts: 302 Joined: 2/23/2013 Location: Grand Rapids, MI User Profile |
Very important concepts for those who have not invested thousands of hours studying strengths and pitfalls. And thanks to those who have! |
|
|
|
|
Administrator
Forum Moderator |
Registered User
Unregistered User |
 E-Mail this thread to a friend |
Toggle e-mail notification |
