OmniTrader Upgrade Forums
Trade Plans
Two "Exponential" Moving Average Crossover
113 replies, 12136 viewings
|
OmniTrader Upgrade Forums
Trade Plans
Two "Exponential" Moving Average Crossover
|
Last Activity 1/2/2025 4:50 PM 113 replies, 12136 viewings |
|
| Printer friendly version |
Jim Dean Sage   Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA  User Profile |
Your question in the last post re VB Lookback=0 pointed out something that I should have thought of. For the FiltSlow strategy, the 10+40 stacking sometimes does not "turn on" until a few bars AFTER the 5x20 occurs. So, it makes sense to extend the System Block vote by the same 5 bars that the DDE used, so that the two EMA-pairs have a chance to "line up" in similar situations where DDE would have seen them that way. Please change the FiltSlow strategy to have the VB Lookback = 5, instead of zero, then do the DJ30/12mo comparisons once again OTOH, in "normal" situations where a TREND is getting started, after a sufficient run in the opposite direction to build up some "pressure" from traders to move in a different direction, one would expect that by the time the 10x40 System-signal occurs in a FiltFast strategy, the 5+20 stacking would already be in place ... so the VB bars=0 makes more sense, there. HOWEVER, there are other situations where even though 10x40 has just occured, the 5+20 is not stacked properly. This can happen during a tight consolidation, where basically all the lines are "mixed up" ... or during a pullback pullback after a sufficient short-trend had driven the 10x40 to hit. In the former of those cases, *I* would prefer NOT to enter the trade. In the latter, I *might* want to enter the trade, but I'd look closely at resistance levels (and some other things) before doing so. So ... presuming that my theorizing is valid, I'd choose NOT to "extend" the 10x40 system-signal for extra bars ... I'd want to see the 5+20 stacking already in place ... and I'd probably add another filter to see if the 5 & 20 are sloped up (but that's getting ahead of ourselves). So ... if you'd like, go ahead and test the FiltFast with VB bars=5 also ... but look CAREFULLY at what ADDITIONAL trades it creates, and see if those entry points are ones that you likely would have taken. Re your comment about the filterFast and filterSlow being mirror-images ... that's the whole idea of the test. Hopefully after some of my comments above, it's a little clearer now. Let me know what the modified FiltSlow looks like ... and which of the two (Fast or Slow) appears to you to be the best emulator of the DDE. Both of the Filt strat's will permit EXTRA trades. Our primary goal at this point is to figure out which one ECHOES as many of the DDE trades as possible ... we can consider the "extra" trades separately. | ||
This accout has been deleted |
With above correction. The filtSlow is closely matched to DDE with extra signals. I concur with the decision not to employ lookback on the filtFast. I really like the filter box... these opens up to more controls that I am looking for. It seems like we can only put in one instruction per Long and Short position in the FB? I am asking because I was thinking of using the third EMA parameter to further refine the filter box commands. | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
The purpose was to keep DDE untouched ... please return it to the non-reversing condition for future use. My intent was for both FiltF/S System blocks to use N=20 with NO reversing. Reason, as described earlier, is to keep the charts as simple as possible for careful comparison. Even with the simplification, and even if we ignore the many "extra" trades, there are a LOT of differences that the FB introduces, vs the "control" DDE strategy. Please fix the system blocks (remove Reversing), and go thru the DJ30/12mo to carefully compare these 2 Filt strats to the DDE. I found TWELVE differences, across TEN symbols (ie 12x when DDE fired, but one or both of the Filt strat's did not, or fired differently) ... a HUGE variance, compared to what we've seen before. Neither of these Strat's is equivalent to DDE. Check for yourself ... how many cases did you count - s/b 12 unless I made a mistake ... of them ... how many: 1. FiltSlow matched DDE but FiltFast didn't 2. FiltFast matched DDE but FiltSlow didn't 3. Neither FiltSlow nor FiltFast matched DDE Although you will see one of those as a max, and another as a min ... none are zero. Personally, I don't think that's a clear enough sampling to draw conclusions from. Be very careful about using poor statistical samples as a basis for making major decisions. Strictly speaking, you cannot even calculate a reasonably accurate standard deviation with less than about 100 samples ... and even if the "distrib" of the samples is "perfect", less than about 30 is nearly meaningless. So ... what should we do from here, to decide which approach is better? a. One path would be to THINK about the two alternatives ... which one fits our "theoretical" idea best of what we want to trade b. Another path is to check more samples ... comparing the reversing version of DDE to versions of FiltF/S with reversing in the SB's c. A third path is to examine the EXTRA trades more carefully I'm attaching the three reversing-version strats if you'd like to pursue option b ... goal would be to answer questions #1-3 above for that set. I've already done a pretty decent job of talking through option a ... but I'm not sure it was clear enough (see prior posts). Option c could be evaluated simply by checking how many of the extra FiltFast trades were profitable, vs how many of the extra FiltSlow trades were profitable ... for the non-rev set separately from the rev-set. Getting tired yet? Hey - this is the nature of the beast, if you want to really gain useful understanding. Quick conclusions can be more dangerous than no conclusions at all. I'll keep trying to help, for as long as y'all want to keep digging in :~) Partial Results from reversing-set: I found that 26 (of 30) symbols had some mismatches with DDErev2in1, and there were a whopping 56 trades that did not match up. See if you can provide the breakout: 1. revFiltSlow matched DDErev but revFiltFast didn't 2. revFiltFast matched DDErev but revFiltSlow didn't 3. Neither revFiltSlow nor revFiltFast matched DDErev [Edited by Jim Dean on 7/13/2010 11:03 PM]  Attached file : aJDgyDualEMArev-2in1.ots (744KB - 505 downloads) Attached file : aJDgyDualEMArev-FiltFast.ots (744KB - 514 downloads) Attached file : aJDgyDualEMArev-FiltSlow.ots (744KB - 498 downloads) | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
I've tried to emphasize the importance of LOOKING AT THE CHARTS. Do simple counting exercises. Look at them from the right, left, top, bottom, green vs red, and so forth. ONCE that is done, and you've obtained a visual-feel for things, THEN it can be useful, especially in cases like this one where there is no clear answer (between Slow and Fast versions) immediately obvious ... to look at the BackTest reports. Simple instructions: 1. Don't use PortSim for this ... its extra features don't apply yet 2. Set a fixed backtest period that matches your eyeball reviews 3. Use the same Focus List as your eyeball reviews 4. Test one strategy at a time and save the Results (delete fwd test info) 5. Compare simple stat's first, then drill deeper The three snapshots below are the Summary Performance+Detail Reports for each of the three Reversal systems that you last checked ... which clearly had enough trades in the 12mo period to be meaningful statistically for simple conclusions. These reports can seem overwhelming due to their very "busy" nature ... it's a useful format, but you have to sort of "dissect" it the first couple of times you look at it so that you don't get cross-eyed. So, first off ... I set up the backtest period (using ToDo > Test Settings) to be a FIXED period of 7/1/09 through 7/13/10 ... the same as my eyeball. The fwd test, for these purposes, is not necessary since we are not optimizing or tweaking anything. This allowed me to CHOP OFF the RIGHT HALF of each of the two reports ... simplifying them. Ahhhhhhh! :~) Then, I highlighted the THREE MOST USEFUL NUMBERS (imho) in purple, blue and green. NOTICE how those numbers are REPEATED, over and over, in each pair of reports. Finally, I highlighted three more rows of numbers on the Detail report (in red) that I consider reasonably valuable at this stage in the game. Most of the other non-highlighted numbers, FOR THESE PURPOSES, are either redundant or effectively meaningless (for a variety of reasons). They all do have merit in other circumstances, however. So ... spend some time with these reports. COMPARE THE NUMBERS - just the highlighted ones. Look back at the statistics you compiled regarding how the two Filt strats "matched up" with the DDE strat. NOW ... can you draw any useful conclusions? [Edited by Jim Dean on 7/14/2010 12:08 AM] Attached file : DualEMArev-2in1 DJ30,12mo.png (226KB - 859 downloads) Attached file : DualEMArev-FiltFast DJ30,12mo.png (227KB - 755 downloads) Attached file : DualEMArev-FiltSlow DJ30,12mo.png (224KB - 841 downloads) | ||
This accout has been deleted |
In beginnning of the filtFast and filtSlow, I mentioned that you got "reversing signal" set on. Which you actually want them off. Then you attached new rev-filtFast and rev-filtSlow, in these strategies, you did turn off the reversing signals in the SB but not in FB... which I believe is why you have "rev-" label. Even I am getting confused at this point. I am turning all 'reversing signals' off in SB and FB because I want to see what the system does. I don't want to see those extra reversing signals which is making us all confused. This is for filtFast, filtSlow, and DDE strategies only. I know you want us to understand and chose our options carefully, but without fine tuning the basic system and strategy set peoperly, any 'options' we chose is going to alter the whole progress of trying to find the right system to start off with. For example, using the 5/20 and 10/40 is set that way so we are all on the same page. And using 'options' such as 'reversing signals' to look see if it is better or worse. Setting lookback periods, before signal and after signal in FB are extremely important settings. Setting N-bar of 20 or 1 is important too. What we haven't explore is setting N-bar to 30 or 40 in the system box. Some of us may have already done so for look see. If the basic system is not working well, then all tweaking and setting sill still not get us good result. Yes, they will be 'different.' [Edited by on 7/14/2010 1:00 PM] | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
First, for the ones that don't have Reversing active: 3 instances - FastFilt agreed with DDE 6 instances - SlowFilt agreed with DDE 4 instances - neither SF nor FF matched * conclusion ... there's no reliable correlation for S or F For the ones that do have Reversing active (cp to rev2in1): 26 instances - refFastFilt agreed with rev2in1 27 instances - revSlowFilt agreed with rev2in1 3 instances - neither revSF nor revFF matched * conclusion ... SF and FF are just not "equivalent" to DDE ----- Stan: If you're getting frustrated then I'm perfectly OK with calling a halt to this. I indicated that the Reversing flag in the Filter block has NO effect on how the Filter Block works. I referred you to the graphic that I created. Yes - this method of training is difficult and limiting. The file attachments are provided to make it easier for you, but sometimes I do make mistakes, so I try to correct them. My preferred method of training is by videos ... followed by forum discussion about the videos. Things can be covered a lot more efficiently and quickly that way. I'll be covering all this stuff, and more, in future AOTC sessions. I was just trying to do my best to help out here, for the community at large. However, if it's getting too confusing ... or if the time required to do the carefully-structured "research" is asking too much of you, then maybe it's best if we call it a day. It's been taking me roughly two hours straight for each installment, on the average. So, if y'all don't have time to do the followup ... or if my presentation is just too difficult to follow ... then I'll use the time in other ways. Up2U. One way or t'other, I hope that what we have covered has been helpful to you, and that it will make you more effective in your future efforts. [Edited by Jim Dean on 7/15/2010 7:24 AM] | ||
This accout has been deleted |
Some of your individual post takes me far longer to understand which is why I posted some time later if not by evening to figure out. You are teaching me far more than I can believe. I think these materials on this post should be in a book. For others: To get the reports that list all the strategies selected... go to File... Print reports... Strategy tab... print or view. Looking at the report, it makes it easier to see which strategies are 'better" than other, of tall the latest strategies... rev-filtslow is the best. For the rest of the strategies, it is what I thought... breakeven or loss. Please continue... | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
First, clarifications: 1. The last five OTS files I uploaded are CORRECT, for the purposes I outlined. I provided an explanation as to why. It *is truly confusing* if we start working on the basis of different strat-definitions - that's why I provided them. If you change them, then we're no longer working together ... and I've already, at that point, put in a lot of effort using the ones I uploaded. 2. To be absolutely clear ... and you can see for yourself ... the "reversing" checkbox **in the FB** has NO effect one way or the other on the strat performance. I've explained the "normal" purpose of the reversing block is for optimization ... we are not using that now. 3. I've explained (and "sermonized") enough about understanding overall structure first, before tweaking the details. That's why I've consistently steered the discussion away from changing parameters ... or, for instance, from playing with N=## values, or Lookback bars. There are other things that logically should be addressed, first. "Parameter-playtime" has an essential place in the process ... but not until the basics are fully understood. At least, not in my playbook. ;~) Here are some additional overall-structure alternatives worth dealing with, BEFORE playing with parameters - this is how I'll gradually continue the process, via the AOTC video-based presentations: a. Look at how the 2in1 system block would work in a strategy that ALSO contained either the Fast or Slow Filter block - that is, two MV2's in the SB, and also either a 5+20 or a 10+40 stacking in the filter ... or both. b. Switch the Orders Block to a Next Pivot Point stop. This will occasion a lot of discussion about NPP's - their value and their dangers. We'll be able to use Backtesting to more usefully evaluate some of our earlier strategies at this point (with the arbitrary N=10 out of the picture). c. Examine three alternative canned stops (using the Orders Block) that logically would "complement" the strategy thinking ... Moving Average Crossover Stop, Two Moving Average Stop, and Reversing Signals Stop. This will occasion a lot of discussion and investigations. Possibly will start looking at Portfolio Simulation at this point. d. Extend investigation of Exits (which are btw MUCH more important than Entries) by looking at several other important "generic" Stops: Fixed Profit, Fixed Loss, Trailing Profit, Breakeven, and maybe Eighths. This will really open up discussions about the reasoning behind the uses of stops, and how MULTIPLE stops can be combined together interactively. e. By now a more indepth look at PortSim would be appropriate, using what we've developed thus far. POSITION SIZING is even more important than Exits ... and FAR more important than Entries. A lot of investigation about the various reports and statistics, and what's safe to "shortcut" in the evaluation-process (sometimes). f. From here the natural step is to hit the rudiments of Trade Plans, and what features they offer that Orders Blocks don't. Initially focus on different Order types, and on the structure of TP's (Steps, Rows, Conditions, Jumps, Long/Short). By this time, PortSim is in full use. **g.** Now, finally, we'll have an adequate understanding of the OT engine and strategy, to start talking about "optimization". Duxx recently posted a question about this in the Pro forum ... it's a pretty involved topic. We'd start with a discussion about how it works, and what its strengths are, and what its weaknesses are. This will lead us into discussions about appropriate backtest/forwardtest procedures to avoid excessive "curve fitting". This is hugely important. h. Our first exercises, for VERY GOOD REASON, WILL NOT use the "optimize" checkboxes and engines in the various blocks that we've been looking at. We will start with **Strategy Wizard** - a superb Add-on that any serious OT user should own. Any other approach would actually HURT our understanding and mislead us as to how to move forward. i. After SW has carved a path for us, and has set reasonable boundaries for final-tweaking-optimization of parameters using the native features of the strategy blocks, we'll tie that process in to the mix and see the advantages and potential pitfalls of it. j. Up till now, our doughty DoubleDualEMA core-systems have been our focus. The next natural step is to put together some alternative Systems and look at how the Performance Block works, and the Vote Block really starts "doing its thing". This is a powerful feature of OT ... it's not "held off" to this late because it's unimportant, but only to keep the prior activities as "clean" as possible. k. Once the System-Voting methods are mastered and understood, then the only other major piece of the puzzle is to try out multiple-Strategy models with the voting process available to them. ****** This entire journey obviously CANNOT take place in this Thread 8~> ... and truly ... it is not sensible, practical or efficient to do it all in writing ... thus the AOTC-video approach. However, useful snippets of each major step can likely be dealt with, in a simplified way, within the OT Forum ... So - I fully expect this to "wear thin" after a while ... hopefully this "big picture" will help you put the "tweak the parameters" urges into perspective - in light of the presumed goal of REALLY USING the broad capabilities of OT. ... and there is SO MUCH MORE ... OT Pro custom indicators, systems and stops, OmniScan, the new MTC Setups, Seasonality and Group Trader, many other Systems and Strategies, etc ... by the time all that we have now is really absorbed, Nirvana will have created a couple dozen MORE. Wheeee! Maybe now is a good time to call a "rest stop". It will give time for anyone who's interested in this discussion to go back over it, and play with things as they choose. Besides, I need to focus a bit more time for a while on some AOTC work ;~) About that - we've just gotten our brand-spanking new Forum in place, and I'm going to shift many of these kinds of extended presentations to that venue, since it's been **intentionally structured** for educational and knowledgebase use. As we "grow" its content, the entire N community is welcome to visit ... either to "lurk" or actively participate ... the more the merrier :~)) ... see the link, in my signature below. | ||
This accout has been deleted |
Confirmed by visual and strategy report. | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
Djenfer ... please check your Private Mail inbox. There's much more to the issue than you are aware of. And, for the record, I FULLY AGREE with you that it would be WONDERFUL if OT's Orders Block and Trade Plans block really did WORK that way (MOC at the end of the Setup bar) ... with real money, for real trades, using Integrated Brokerage ... at the HRE ... but they currently don't - and that is "by design" - N's design. Currently, anyway :~/ My desire is to have backtest and portsim info *accurately mirror* what OT *actually does* in real live trading, as opposed to it reporting test-performance that *differs* from real live trading - especially if it might be biasing the results (inadvertently) in a positive manner, by doing so. But, once again, *I do STRONGLY agree* with you that the MOC order should be applied at the end of the Setup bar, as long as that and function consistently works the same *across the board* in all aspects of OT's operation ... so that the entry ACTUALLY is sent TO THE BROKER (by OT's engine) at that time (end of the Setup bar), when a strategy is used for live, real-money trading. This goes for both EOD and RT, btw. The plain, tested, "by design" FACT is that at this point in time, OT does NOT work that way for MOC (or any other order type, for that matter) ... it does not have the ability to send any Entry orders to the broker until the first tick of the Entry bar has appeared. [Edited by Jim Dean on 7/14/2010 10:20 PM] | ||
This accout has been deleted |
I find the 'Help" not helpful... I really wished there is a book for this. It is strange how powerful OT is but lacks in textbooks. I used to use Telescan super long ago... 15 years ago. At least that came with a book. I also tried Tradestation 2000 long ago and they did give a book for that but I was perplexed at the complexity of EasyLanguage... I currently do have Tradestation but haven't forage into their strategy trading... augh! | ||
| Jim Dean Sage Posts: 3022 Joined: 9/21/2006 Location: L'ville, GA User Profile |
Please forgive the repeated mention of AOTC ... but continuation of this particular journey will take place over there ... in the future ... and I don't want to leave folks who have been following this thread closely "hanging" in the midst of the process. A pure written format is just too limiting for broad-scope topics like this to carry on for too long. Click here to visit AOTC [Edited by Jim Dean on 7/15/2010 7:26 AM] | ||
This accout has been deleted |
| ||
| SteveL Veteran   Posts: 262 Joined: 8/19/2005 Location: Boulder, CO User Profile |
Thanks for the lesson. I have a better understanding, particularly the confirmation block and how the before/after tolerance works, and the consequences of the reversing choice in the system and confirm blocks. And of course, the start on a framework for how to approach the problem of developing a strategy that I thoroughly understand. Sorry I didn't have the time to keep up with you two during the week. Look forward to continuing it at AOTC. Steve |
|
|
|
Administrator
Forum Moderator |
Registered User
Unregistered User |
 E-Mail this thread to a friend |
Toggle e-mail notification |
