OmniVest User Forums
User-Created Portfolios
Conditional Experiments
41 replies, 4027 viewings
|
OmniVest User Forums
User-Created Portfolios
Conditional Experiments
|
Last Activity 7/22/2015 8:11 AM 41 replies, 4027 viewings |
|
|
| Printer friendly version |
Jim Dean Elite   Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA  User Profile |
An OVest "hard entry" is equivalent, similarly, to an OT trade plan entry order that uses an OScript or OLang Condition. The strategy creates a an entry Signal, but the OScript condition forces that signal to be delayed (or entirely skipped) based on the conditions dictated by the OScript. So, these are "new" to OVest but not to N-land. They are very powerful and very important. They offer one further step towards providing the "power of trade plans" to the OVest user. [Edited by Jim Dean on 5/10/2014 6:56 AM] | ||
| Steve Mayo Legend  Posts: 414 Joined: 10/11/2012 Location: Austin, TX User Profile |
Hard Switching was the technique we used in our mock-up of PortSwitcher....basically to sell all open trades in PortA and buy all open trades in PortB as of a certain date. In writing the code to implement it, Nirvana split it up into a "hard sell" and a "hard buy." So, now you can decide to let your open trades in Strat/PortA continue to a normal close AND/OR decide whether you want to jump into the in-play trades for Strat/PortB or wait until that strat/port fires new trades. Neat! However, with the hard switching (e.g, both boxes checked) that we (Mark Holstius and Steve Mayo) used in our mock-up, we could know (almost) exactly what the result would be -- basically, we were just stitching together pieces of the equity curves. But now, with 'soft' switching (one or both boxes UNchecked -- or as the current system is implement pending those checkboxes -- it's impossible to predict what the result will be without actually running it through the simulator (i.e., recalculating trade by trade, day by day). OV is all about trying to stay fully invested and what happens at the 'equity ceiling' to accomplish that. By electing to NOT close open trades (the current system without the checkbox), you tie up equity in Strat/PortA that may or may-not have performed better if it could have been employed by Strat/PortB. Likewise, by NOT immediately entering the in-play trades of Strat/PortB, you miss out on those trades (which may or may-not have improved/worsened your return). In other words, you don't know what will happen until OV runs it through the simulator. We added 'soft switching' because there was concerns that hard switching could mean closing out trades at a loss that might have later turned around and because you might end up getting into trades after they had made their initial move. That's absolutely true -- but comes from thinking like a trader. If, instead, you think of OV as a mutual fund, then hard switching is more logical. You buy-in or sell-out of a mutual fund at the daily closing price with no regard to what stocks the manager bought or sold that day based on that fund's performance (read equity curve), based on what that fund has done in the past and what you expect it to do in the future. With soft switching, we introduce a new set of conditions that make the past and future unpredictable, at least over the time that it takes the soft changeover to equilibrate (and then compounding that unpredictable loss/gain over time). Granted, everything in technical analysis is based on backtested simulation -- soft switching is no different -- but as a scientist, I have to caution that it is introducing a new variable that can dramatically change the "characteristics" of the system. Hard switching does too, but it's more directly a function of the individual strats/ports used in the composite; soft switching is not the sum of the parts but rather a whole new creature, at least during the equilibration period. Using conditions and soft-switching at the strat level means the "switching" condition gets applied daily so the system may never really equilibrate, of course, depending on how frequently the condition takes effect. Bottom line, this is a brave new and exciting world. We need to do a lot of testing to better understand it. | ||
| Mark Holstius Elite  Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL User Profile |
And, I want to re-iterate something you posted in this thread a while ago. Consider the charts I’ve posted as a “proof of concept” for me of ways to set up and use Strategy Lab and Conditions - with the hope of utilizing some of those concepts when everything comes together with the Portfolio Switcher and Wizard. They’re a great example of “curve fitting” – but an even better example of what these tools allow us to create. I’ve never been able to get an equity curve like that before – especially with all the restrictions in my account settings. And, I think I can tweak the concepts / rules I used this week to create even better results (yes, there’s room for improvement from what I’ve seen). But – we’ll need to use the Switcher / Balancer to find statistically reliable ways to choose the correct Portfolios and Strategies to trade at the HRE through “walk forward” testing. Like Ed said – this is the only product “out there” with these capabilities. Like you said – it’s a brave new and exciting world (full of possibilities). Looking forward to it, Mark | ||
| Jim Dean Elite Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
There's one other consideration tho re hard Entries ... it's a nuance but sometimes might be an important one. I posed this as a question, but haven't heard back yet. The question is ... if we do a "hard entry" in the "middle" of an already-in-process trade, then how does that strategy's Trade Plan get applied, FOR that hard entry, both in simulation and in the Trade Processor? Two possibilities: 1. TradePlan acts like the trade started at it's theoretical point, not at the midpoint. 2. TradePlan is shifted to start at the hard-start "midpoint". Pros and cons here ... both re trading logic, and re difficulty to implement it in the TradeProc. Something to consider: Some Trade Plans (like for RTM strat's) are super-simple "get out after five days" ... for a plan like that, the hard-entry might occur on the last of those five days, and exit on the next bar. Good? Bad? hard to say ... but imho it's more likely to be a "noise" trade, not actually capturing most of the profit. Other Trade Plans, such as those used for Trendfollowing strat's, can be much more "robust", using rule of eighths, trailing profit with various cushions and thresholds, or any of an infinite number of custom TPlans once Elite Trader is active. Some of these, it might be smarter to "shift" it to use the #2 method. For others, maybe #1 is best ... but, again, my guess is that the #1 method will often result in very quick exits. Good? Bad? who knows? Ideally, we get an option for this. A user input that says a. Hard Entry with simulation Trade Plan b. Hard Entry with shifted Trade Plan For those of us who will be doing a lot of ET development, with a heavy focus on stops, that distinction is important. If you find out what the "plan about the Plans" is, please let us know. | ||
| Steve Mayo Legend Posts: 414 Joined: 10/11/2012 Location: Austin, TX User Profile |
Ed's goal was to replicate the switching approach Mark and I used at the portfolio level -- which would presume #1 in your list of possibilities. So, I'm thinking that is likely what the programmer did. I think it will just include/exclude that trade (or that portion of it) when generating the equity curve. However, Ed has noted that it takes much longer to run with hard entries than with hard exits turned-on. Seems to me that both would trigger a next-morning transaction affecting the available equity that would have to be processed so I'm not sure I understand why they are different. Maybe it does recalculate the stops -- treating them sort of like newly signaled trades-- as well, but I'm doubtful given how the trades are coming from a different server that is running all the strategies each evening. | ||
| Jim Dean Elite Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
Or, maybe there will just be a "rule by fiat" that all uploaded ET strat's MUST have only one entry and one exit, and very limited rules for virtual stops, etc. I hope that last one is not the case! I can see how scaling in/out might be outside the OVest paradigm, but I hope robust stops aren't. Please discuss this with Ed. There may need to be one solution for initial release, but firm plans for enhancement later. [Edited by Jim Dean on 5/10/2014 12:54 PM] | ||
| kmcintyre Elite Posts: 890 Joined: 10/11/2012 Location: Portland, OR User Profile |
Mark, I have always thought your posts to be valuable reads. Please "show off" (aka "share") frequently! Steve, I never miss one of your posts either. You guys have presented so much valuable information - it's SICK!!! (That means "good" these days...) All OV users owe you two for your contribution to current Nirvana development efforts. I have no idea what your business relationship is with Nirvana, but I hope you guys are being compensated. You have my admiration. As soon as I make 10% on my live account I'll buy you two (and Ed) dinner... Cheers! Keith | ||
| Ed Downs Elite Posts: 645 Joined: 2/7/2007 Location: Austin, Texas User Profile |
We are testing Dynamic Entries and Exits, which incorporate a "Hard Switch" that Mark is referring to. Attached is my PDF that shows a specific Strategy before and after employing Dynamic Entries/Exits. The effect of "chopping off" trades when the curve goes against you and "re-entering" trades when it goes in your favor can be quite dramatic. The experiment I ran shows Draw Down reduced by half and CAR nearly doubled. See attached PDF for all the data. One last thing about this somewhat simple example - I used the very basic "Continuous Trend" filter on SPY to specify switching, NOT the Equity Curve. So, if Continuous Trend on SPY shows that the ETF is going down, it closes all trades and when it's going up, enters all trades that would be open. Note tha this has NOT yet been released. All of Mark's experiments have been done without the benefit of this little jewel. I can't wait until he and others get their hands on it. We are running weekend trading tests with this on our test server. Fingers are crossed that we can release it Monday.  Attached file : e1.jpg (43KB - 558 downloads) Attached file : e2.jpg (47KB - 569 downloads) Attached file : dynamic3.jpg (81KB - 557 downloads) | ||
| Jim Dean Elite Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
[Edited by Jim Dean on 5/10/2014 1:19 PM] | ||
| Mark Holstius Elite Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL User Profile |
So sad - I'll be out of town for a week starting Thursday. ;-) Mark | ||
| tgrafa Regular Posts: 63 Joined: 10/11/2012 Location: Midland, Texas User Profile |
http://videos.nirvanasystems.com/OmniVest/4-30-14/breaking_news/video.html | ||
| kmcintyre Elite Posts: 890 Joined: 10/11/2012 Location: Portland, OR User Profile |
Keith | ||
| Mark Holstius Elite Posts: 744 Joined: 10/11/2012 Location: Sleepy Hollow, IL User Profile |
So you don't have to "page back" I'll repeat the original chart comparing it to Ed's ARM4 (my "personal benchmark"): And here's the "out of sample" performance so far in 2014... Mark [Edited by Mark Holstius on 5/11/2014 6:34 AM] Attached file : Port v ARM4 Developing.jpg (589KB - 474 downloads) Attached file : Out of sample performance 2014.jpg (698KB - 493 downloads) | ||
| Jim Dean Elite Posts: 1059 Joined: 10/11/2012 Location: L'ville, GA User Profile |
That "completes the story". IMHO, all samples of portfolio results that are posted should show the "trained" region, as well as one or preferably more "out of sample" regions. Gives a proper perspective. | ||
| Geoff Veteran Posts: 180 Joined: 12/4/2012 Location: Byron Bay NSW Australia User Profile |
FWIW my own observations (and others I'm sure) from analysing strat performance, at the point where market conditions change (eg, from bull to bear) it seems that often, once the market starts to pullback it takes about 4 to 5 days for the strats to fully react at which time they either start firing short orders or stop issuing long orders. During this 'Market' transition time the strat indicators are obviously re-adjusting and the strats adapt to the new market conditions which is why I think the OV Ports do often rebound sharply. This rebound often starts BEFORE the market has finished pulling back. Conclusion: 1. For a Port to achieve its potential max performance, it needs to capture the rebounds (obvious). 2. If you set a 'Soft Sell' could you run the risk in some cases, of locking up 'equity' in trades that are likely non-performers, and risk missing out on taking up new orders from strats that have already adapted to the changed conditions? I realise this is all hyperthetical and there is probably no definitive answer but, it's the reason I asked the question about 'Hard Switching', sounds to me like a great feature after my experiences with switching ports that have NOT issued an order because of old positions not closing. | ||
| John W Elite Posts: 654 Joined: 10/11/2012 Location: Sydney, NSW, Australia User Profile |
| ||
| Geoff Veteran Posts: 180 Joined: 12/4/2012 Location: Byron Bay NSW Australia User Profile |
|
|
|
|
|
Administrator
Forum Moderator |
Registered User
Unregistered User |
 E-Mail this thread to a friend |
Toggle e-mail notification |
