OmniTrader Upgrade Forums
Trade Plans
Two "Exponential" Moving Average Crossover
113 replies, 12135 viewings
|
OmniTrader Upgrade Forums
Trade Plans
Two "Exponential" Moving Average Crossover
|
Last Activity 1/2/2025 4:50 PM 113 replies, 12135 viewings |
|
| Printer friendly version |
Jim Dean Sage   Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA  User Profile |
| ||
This accout has been deleted |
| ||
This accout has been deleted |
Now, you told me that is for optimization purposed... and we are not using any optimization whatsoever... so how come it impacts the signal? No TV?! DVD is higher quality entertainment. I have to cut down on my TV and more into Omnitrader... LOL. | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
Now, look at the expanded votelines and see if you can figure out why the 20 matters. Look at specifics. How and what changes if you change it to 19 or 15 etc. It's primary purpose IS optimization - but there's a secondary impact in play, too. Labels rarely tell the full story. | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
Goal: to learn how those inputs impact how these systems actually fire (or don't fire) signals. After you've figured that out (one thing at a time) -see if that reversing-signals checkbox makes a difference - and if so, how. Then after that, check out the dropdown box that currently is set to NBars. This is the way that many of us have learned stuff that the manuals are silent about. It's sorta fun - as long as you learn it BEFORE you've got money on it! [Edited by Jim Dean on 7/9/2010 9:23 PM] | ||
This accout has been deleted |
If we set to 5. it will stop at the 5th bar (for heaven's sake of simplicity) | ||
This accout has been deleted |
With the reversing signal, the system will fire at every crossovers it makes. With this feature checked,the system voting line is different as there may be a short one bar sell crossover and then trigger a buy again. Still with the reversing signal feature, there is still the 20 N-bar stop that will stop out of the trade when it hits the 20th bar (day). Wow... all these years, I never bother to zoom in so closely... now I even see another feature but this is another rabbit trail. | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
I trust that by now you have a better feel for how the inputs on the bottom right of the System and Confirm blocks work, with Optimization turned OFF. Here are some summary notes that are worth saving ... if any of them are "news" to you at this point, I suggest you take a moment to TEST them ... maybe you will notice something that calls for a revision. This info is NOT, for the most part, found in a manual anywhere: 1. SB & CB: DblClick Enable on top left to see only checked systems 2. SB & CB: To uncheck all, click top sys, shftclk bot sys, and set/clear Enable at bot left 3. SB & CB: Can use Duplicate button to make extra copy of a sys, with diff param's ... that copy ONLY exists in THAT BLOCK ... does not create a new sys file on HD ... and for custom sys, abbrev-name does not change 4. SB & CB (not FB): Performance > Metric=N-Bar + Number of Bars does affect system operation even if no Optimization active ... makes SysVote persist longer ... prevents sys from firing in the other direction during that period (if no Reversing active) ... DOES allow add'l sig's in SAME direction (look like voteline "reinforced", but not affected by Chart Options > Trades) 5. SB & CB (not FB): Performance > Use Reversing Signals does affect sys operation even if no Optimization active ... signals can flip directions during the Number of Bars period - in this case, AND if no optimization is active, NofBars input does not matter 6. SB & CB (not FB): Performance > Metric=NextPivot + S/M/L does affect sys operation even if no Optimization active ... trade persists to next Pivot ... a Pivot cannot be identified at the HRE, so signals might change in "real" life - UNLESS Reversing is also used. In that case, the "late" appearance of the NPP will not prevent any system signals from being fired, so although some nuance-details of the expanded-voteline might change, the final vote will not (in theory, at least :~) Okaaaay ... that last one is sort of difficult to follow, maybe ... but I suggest that you play with the inputs ... set up NPP (Long), without reversing signals, and compare to NPP (Long) WITH reversing activated. To make these comparisons simple, you might want to SaveAs a new Strategy for that purpose. Please let me know if you "get it" ... if not, please pose some questions that will help me explain it more effectively. [Edited by Jim Dean on 7/10/2010 9:12 AM] | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
Since the "Number of Bars" is apparently having an effect on our strategy's operation, we need to ask ourselves, "is 20 the RIGHT value"? Our thots immediately turn to optimization runs, etc. BUT ... FIRST ... rather than beating it to death, let's use our BRAINS and think about WHY. In order to answer that Q, we need to ask "how long SHOULD a given signal's DIRECTION persist"? This latter Q might confuse you at first ... what does N=20 have to do with long vs short direction. Answer ... our experiments make it clear that DURING the N=20 window after a Long system-signal fires, although OT permits "reinforcing" (hollow-triangle) signals to fire in the SAME direction, it does NOT permit that 20-bar window to be "interrupted" by a change in direction ... that is, if the EMA's cross back in the opposite direction during those 20 days, the resulting SHORT signal is not allowed to appear. So ... do we WANT to allow a short signal to appear within that 20 days or not? And if we do, what's the best way to go about it? Second question first ... there are TWO WAYS to permit a short signal to appear after a long one, within 20 days after the long one has fired. Please figure out the two methods and report them back. This only affects the System block, btw. OK, to decide IF we want this to be the case ... maybe the best way is to set up two Strategies that differ only in that way, and look at some charts and see how things work. Please modify the "2in1sys" Strategy to implement one of your solutions. Leave the original DoubleDualEMA strat as is. Now work through the DJ30, using 6-mo window, expanding the votelines when you need to so that you understand the differences. LOOK AT WHAT'S GOING ON WITH THE CHART ... for the first time in our journey, we're going to evaluate which setting gives us "better trades" ... but let's not get excited about what these measely 6-mo's and 30 syms tell us specifically ... let's USE them to THINK about how this strategy will work IN GENERAL. This will occasion a number of discussions ... but at junctures like this, we should try to pick one of the two paths early on, so that our ongoing work will not develop too many persistent branches. Have fun! | ||
This accout has been deleted |
Now maybe 'what weekend?' may pop up in some minds... I really admire it... even for me, there is no 'weekend' but weekend is a tad busier of running in and out of the house. Yes, that crossed my mind all along of what can we do to improve and how to do so to achieve a good performance. | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
| ||
| SteveL Veteran   Posts: 262 Joined: 8/19/2005 Location: Boulder, CO User Profile |
I'll step out of the shadows and join the class. Many thanks for the tutorial. It is excellent. I believe the question you have posed is: What ".. are TWO WAYS to permit a short signal to appear after a long one, within 20 days after the long one has fired" by modifying the system block (not the confirming block): 1. "Use Reversing Signals" 2. Use two system blocks, and set one to generate Longs only and the other Shorts only (instead of both). And modify the vote block. It would seem that option 2 (multiple system blocks) would offer more flexibility, so I choose to follow that path for further study at this time. This snapshot shows the CAT voteline for the past 6 months for the original "DoubleDualEMAcross" and "DoubleDualEMA-Confirm" compared with the modified DoubleDualEMA-2in1sysB strategy with 2 system blocks (one long and one short), and vote block modified to require only 50% of all inputs to pass a signal. This requires that both of the long systems or both of the short systems to be concurrently firing to produce a signal. This preserves all the original signals, but also enables a reversing signal to occur sooner than with the original stategies, because the short system can fire before the slow long system signal has ended. And the Shorts only system block: And the Vote block with 50% minimum for "Voting Method" and 50% minimum for "Conflict Resolution". Steve [Edited by SteveL on 7/10/2010 3:37 PM]  Attached file : 2010-07-10 CAT DoubleDualEMA - vote block.PNG (47KB - 816 downloads) Attached file : 2010-07-10 CAT DoubleDualEMA - voteline.PNG (42KB - 725 downloads) Attached file : 2010-07-10 CAT DoubleDualEMA - Shorts only.PNG (84KB - 713 downloads) | ||
| SteveL Veteran Posts: 262 Joined: 8/19/2005 Location: Boulder, CO User Profile |
| ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
Your first solution ... using the Reversing checkbox ... was one of the two methods I was thinking of, in order to assure that no signals were lost due to the N=20 "persistence" effect. Your second solution ... using two System blocks PLUS changing the Vote block ... is one that I was not thinking about - but you're right ... it offers additional versatility - I discussed this earlier. Hooowwwever ... insofar as the immediate question applies ... it sorta violates the KISS principle ... can you think of another way to accomplish the SAME THING as checking the Reversing box, WITHOUT having to mod the Vote Block or add any more blocks? I'm not pulling your leg here ... there is another "KISS" alternative. Whether you discover it or not ... let's find out how "equivalent" your two solutions are. Do the DJ-30 + 12-mo chart "thing" that I described earlier, with your two strat's on the voteline ... I suggest you collapse them for easy checking. If ALL of the symbols for ALL of the trades in ALL of those months have EXACTLY the same voteline, then I'd be willing to grant that your second-sysblock + revised-voteblock method is functionally equivalent to the Reversal checkbox approach. And, of course, if you do figure out the other "KISS" method, add that one to the pile when you do the checks. I've NOT checked your new suggestion myself ... I have checked the other two ... let us know what you find out! | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
| ||
| SteveL Veteran Posts: 262 Joined: 8/19/2005 Location: Boulder, CO User Profile |
I don't see the other KISS solution...yet. I compared the "2in1sysB-Rev" (1 system block with "Use Reversing Signals") with my solution using 2 system blocks and vote block changes on the Dow30 for the past 12 months. There are two differences. Steve [Edited by SteveL on 7/10/2010 5:07 PM] Attached file : 2010-07-10 WMT DoubleDualEMA strategy comparison.PNG (39KB - 696 downloads) Attached file : 2010-07-10 GE DoubleDualEMA strategy comparison.PNG (171KB - 747 downloads) | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
The .OTD file attached should be copied into the C:\Program Files\Nirvana\OT2010\ChartTemplates folder. Steve - if you would not mind posting your strategy (.OTS) file, I'll take a look at your suggestion. Attached file : aJDctDoubleDualEMA.otd (108KB - 476 downloads) | ||
| SteveL Veteran Posts: 262 Joined: 8/19/2005 Location: Boulder, CO User Profile |
The 2nd solution is to set "Number of Bars = 1" in the systems block. By the way, I updated the previous post. I found that there were two differences, and have posted snaps of both. And as requested, I've attached my 2 system block with vote block changes solution. Steve Attached file : aSJLstrDoubleDualEMA-2in1sysBx2.ots (780KB - 504 downloads) | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
To allow all possible system-signals to get through to the Vote Block, there are two "KISS" methods - either of these by itself does the trick ... you can use both if you want: 1. Check the "Use Reversing Signals" box in the bottom right of the SB/CB 2. Set Performance Metric to N-bars and Number of Bars = 1 (in SB or CB) If you are doing parameter-optimization, it's likely that the second solution will be unacceptable (we might discuss this later on, but not now) ... so the first solution (reversals) would be a better bet. However, if you are doing param-opt, and you want to keep the Short and Long sides separate (usually this is wise for optimization), then Steve's solution proposed in the last few posts is an excellent one. It's tempting to discuss the Next Pivot Point option here ... but that will definitely open up a huge can of snails, which would slither all over our orderly path here. So, for this thread, NPP discussion is "verboten". Nutshell ... for MOST (but not all) situations, it's best NOT to use NPP for optimization, or for Exits or anything else. There ARE exceptions to this where NPP is the PERFECT tool ... but the snail-goo about that will have to wait :~) Any further questions about this? There is still a "task" remaining. Please set up two strat's ... our original DoubleDualEMA without "Reverse" ... and let's modify (permanently) the "2in1" strat to include Reversal ... we previously demonstrated that 2in1 produced identical results to the DDE. Now, with those two strat's active (only), do the DJ-30 12-mo process again ... SLOWLY ... and identify which of the NEW trades that appear with this revision are "Good" new trades (ie profitable), and which are "Bad" new trades (ie losers). The Advisor will help you ID them. After doing that (a list with symbol and mo/year would be great), then try to CATEGORIZE those trades ... WHY were the good ones missed by the DDE strat, and WHY were the bad ones borne, using the Rev2in1? Try to determine GENERAL explanations, so that we can evaluate how to make a GENERAL decision about which (DDE or Rev2in1) to use, moving forward ... and also so we might start thinking how to improve things further ... please NO suggestions yet. You're in "detective" mode now :~) I've attached the Rev2in1 strat so that we're on the same page. [Edited by Jim Dean on 7/10/2010 6:01 PM] Attached file : aJDgyDualEMArev-2in1.ots (744KB - 472 downloads) | ||
| SteveL Veteran Posts: 262 Joined: 8/19/2005 Location: Boulder, CO User Profile |
With the reversing signal, there were were 44 additional trades (plus 5 incomplete trades). 14 of them were losers. The best trades occurred because the stock was not doing much, i.e., EMA(40) was quite flat, prior to the trade. Without the "Use Reversing Signals", the small whipsaws hid the good signals for the next 20 days. The worst trades occurred because the stock reversed twice within 9 to 15 days (most were about 10 or 11 days between reversals). The second reversal occurred very quickly after the trade was entered. # STOCK SIGNAL DATE GAIN/LOSS L/S 1 aa 2009-07-24 12.39% LONG 2 aa 2009-11-16 1.41% LONG 3 axp 2009-07-15 3.97% LONG 4 axp 2010-03-09 3.80% LONG 5 ba 2009-11-16 3.00% LONG 6 ba 2010-06-29 ? SHORT 7 bac 2010-01-26 0.27% SHORT 8 cat 2010-02-22 2.11% LONG 9 csco 2009-07-15 11.14% LONG 10 csco 2010-02-17 3.83% LONG 11 dd 2009-12-30 0.41% LONG 12 dd 2010-02-22 4.02% LONG 13 dd 2010-07-06 ? SHORT 14 dis 2009-07-16 2.86% LONG 15 dis 2010-02-19 5.26% LONG 16 ge 2009-11-12 3.24% LONG 17 ge 2010-01-11 -3.21% LONG 18 hd 2009-10-13 -8.14% LONG 19 hd 2010-02-11 9.18% LONG 20 hpq 2010-02-24 3.73% LONG 21 ibm 2009-07-16 4.82% LONG 22 ibm 2010-03-09 -2.68% SHORT 23 ibm 2010-07-02 ? SHORT 24 intc 2009-11-17 -0.25% LONG 25 intc 2010-02-17 0.10% LONG 26 jnj 2009-11-13 3.35% LONG 27 jpm 2009-07-15 8.11% LONG 28 jpm 2010-01-25 0.92% SHORT 29 kft 2009-11-30 -0.63% SHORT 30 kft 2009-12-18 7.31% LONG 31 kft 2010-05-11 -5.85% LONG 32 ko 2010-04-07 -1.26% SHORT 33 ko 2010-04-23 0.70% SHORT 34 mcd 2009-07-27 -0.02% SHORT 35 mcd 2010-06-14 -5.22% LONG 36 mcd 2010-06-30 ? SHORT 37 mrk 2009-11-02 -10.23% SHORT 38 mrk 2010-04-07 8.14% SHORT 39 pg 2009-07-07 5.59% LONG 40 pg 2009-08-31 2.75% LONG 41 pg 2010-01-19 2.02% LONG 42 pg 2010-02-04 3.50% LONG 43 t 2010-06-25 ? SHORT 44 trv 2009-07-23 10.96% LONG 45 trv 2010-04-22 -4.33% LONG 46 utx 2010-07-17 1.78% LONG 47 wmt 2009-11-03 -7.91% SHORT 48 xom 2009-08-11 -4.16% SHORT 49 xom 2009-08-27 -1.43% LONG | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
[Edited by Jim Dean on 7/10/2010 9:38 PM] | ||
This accout has been deleted |
1. Using reversing signal is easy way for the system to fire a short within 20 n-bar set. 2. Another solution is to alter the 5/20 and 10/40 to other shorter time periods to trigger earlier reversing short... I am not sure if this is allowed under your question. I am also assuming you only allow change in the system block (SB). | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
[Edited by Jim Dean on 7/10/2010 10:03 PM] | ||
| SteveL Veteran Posts: 262 Joined: 8/19/2005 Location: Boulder, CO User Profile |
Here are two of the 4 charts showing the voteline differences: [Edited by SteveL on 7/12/2010 7:24 PM] Attached file : 2010-07-10 XOM 2000-04-18 comparison.png (91KB - 597 downloads) Attached file : 2010-07-10 CAT 2001-03-29 comparison.png (80KB - 601 downloads) | ||
This accout has been deleted |
With the reversing signal and N-Bar=1, I see extra long trades which are profitable but it missed the extended bull run on those new long trades. As with N-bar =1 or =20, there is no differences that I can see in 6 months charts on all Dow30. I used the 2in1 system and have two sets... one to 1 and other to 20... no difference at all. Reviewing what I said above, in the vote line, it made no difference, but in the raw vote lines of the system, there is a difference. Overall, in the end, the strategy vote line did not alter at all. I presume this is an important matter... yes? [Edited by on 7/10/2010 10:33 PM] |
|
|
|
Administrator
Forum Moderator |
Registered User
Unregistered User |
 E-Mail this thread to a friend |
Toggle e-mail notification |
