logo
Welcome Guest! To enable all features please Login or Register.

Notification

Icon
Error

2 Pages12>
Options
Go to last post Go to first unread
guara_riua  
#1 Posted : Thursday, October 13, 2005 2:48:04 PM(UTC)
guara_riua

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/24/2005(UTC)
Posts: 77
Location: Canada

This was an issues that was discussed some times ago with some variations about trading signals appearing/dissapearing/moving in the real-time trading environment. Until I haven't saw it for myself I didn't realized the huge impact it can have for a trading system that is based on the Pivot Types to generate the trading signals. For the last 6 months I've worked on a few versions for a trading system based on a couple of fast and adaptive moving averages that had the role to identify the trend (Up, Down, Neutral);the trade signals were generated mainly by the trend and the pivot type. I believed that I have found the holly grail because the back tests were absolutely unbelivable good: the return for 1 year for different currencies in Forex were between 20:1 up to 100:1. This kind of return sounded quite un-real, but I took all the measures to make sure I use real-time usable signals: i.e. the trading signal must be delayed up to 4 bars in order to allow the pivot to be identified in real-time, because this is what happens; a pivot can take from 2 bars and up to 4-5 bars to be identified on the chart and only then the trading signal can be computed. Here is what I was able to track and replicate: - an Intermediate Pivot was replaced by a Major Pivot 6 hours backward in the chart! - a Minor Pivot was replaced by an Intermediate Pivot 12 bars backward in the chart! Obviously at those moments is time when this happened there were new trading signals generated "post-mortem". I wanted all to know that Pivots Type or PP+ external functions based on Pivot Type could not be used for real time trading. Be aware!!! While some external functions could still be useful, anything based on Pivots Type will generate erroneous trading signals and the back testing of such treading systems will output results that will be un-achievable in real-time. I hope others will find this post useful. Guara Riua
henry1224  
#2 Posted : Thursday, October 13, 2005 11:20:12 PM(UTC)
henry1224

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 10/29/2004(UTC)
Posts: 1,394
Location: Glastonbury, CT

Was thanked: 2 time(s) in 2 post(s)
Guara Riua, I can't comment on your system, since you haven't posted it. I have a few questions for you though, 1: what time frame are you trading? 2:what are your pivot Time factors? 3: Are you using the pivots as a stop? I have found trend direction to be very stable, it is based on pivots from a higher time frame to determine the trend on a lower time frame
guara_riua  
#3 Posted : Friday, October 14, 2005 1:25:11 AM(UTC)
guara_riua

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/24/2005(UTC)
Posts: 77
Location: Canada

Henry Thank you for taking the time to reply. Unfortunately any system cannot change the behaviour of the plug-in or add-on. The only function from the PP+ I have used in my trading system is: - ExtFml("PowerPivots.Pivots") in order to identify the Pivot Type/Importance. The system algorithm is quite simple: - buy when the trend is UP and Pivots are Low Minor or less (-1,-2,-3 or -4) OR buy when the trend is Neutral and Pivots are Low Intermediate or less (-2,-3 or -4) Similar for the Sell signal only change being the values of the Pivot Types are positive. My finding is related to the fact the Pivot Types are changing while streaming real-time data. Anybody could reach the same results if will folow these steps (for any historical intraday data reccord): - change the Loaded and Displayed DATA time frame to have the same start and end date AND time - make MetaStock to load and display the data close to a recent Major/Intermediate Pivot time and date that normally has after it 50-60 bars - add bar by bar or a group of 2-3 bars every time by changing the Time Value for the Displayed and Loaded Data, and the pivot type change will show. Obviously a "test" expert advisor should be created to display only the pivot types (In the Symbols Tab). As I've learned, most of the PP+ function ARE using almost always the Pivot Types. If I am right, then its all bad news for all the PP+ owners because real-time trading using this add-on will not work as intended. The most of the erroneous signals will be for the functions that are using Pivot Types of Intermediate value and higher. Using higher time factors/time frames will not solve the problem. Only functions that are using Minor Pivots will not generate erroneous signals. I hope I've provided a good deal of info about the problems this add-on has. One last notice: I used only FOREX data for my testing. Regards Guara
henry1224  
#4 Posted : Friday, October 14, 2005 1:46:11 AM(UTC)
henry1224

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 10/29/2004(UTC)
Posts: 1,394
Location: Glastonbury, CT

Was thanked: 2 time(s) in 2 post(s)
Pivot types are meant for support and resistance and not for exact timing signals.
henry1224  
#5 Posted : Friday, October 14, 2005 1:53:37 AM(UTC)
henry1224

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 10/29/2004(UTC)
Posts: 1,394
Location: Glastonbury, CT

Was thanked: 2 time(s) in 2 post(s)
here is the formula for Nth Pivot Price i1:= Input("Pivot Type", -4, 4, 2); i2:= Input("Instance", 0, 10, 0); i3:= Input("TFactor", 1, 100, 1); ExtFml("PowerPivots.NthPivotPrice",i1,i2,i3); Plot this for both + and - pivots on top of the price chart and you will see support and resistance
guara_riua  
#6 Posted : Friday, October 14, 2005 5:15:17 PM(UTC)
guara_riua

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/24/2005(UTC)
Posts: 77
Location: Canada

Henry I always appreciate the comments and recommendations from traders with more experience than mine. Working in the electric design field I am inclined to rely on tools and systems that provide a consistent behavior for a large variations of data input and more importantly, are simulating with a high consistency. I agree that my trend-pivots following system can be not the best way to use the PP+ add-on. But even with your example that plots the resistance-support lines based on Pivots Type, the plot output will change in real-time as the pivots higher than Minor will change. This will happen also for higher time frames. What I am saying is that most probably, PP+ is providing a very good info for any given historical data, ploting good quality pivots and other indicators based on pivots type. But all this is changing in real-time data streaming. I am pretty sure that TrendDirection will change its output too, if set for a pivot type higher than minor, for any TimeFactor used. This can be verified with any historical data by cutting it shorter, close to a Major/Intermediate pivot and adding subsequently more bars to the loaded/displayed data. These are the reasons that I am rising a big question mark about using PP+ in real-time trading. I can agree also that most of the time the PP+ add-on could provide good quality signals, but any trading system based heavily on PP+ will not be able to provide good quality back testing results. This kind of system cannot have its performances consistently evaluted because the simulation will be always better than the real-time results. Tese were my 2 cents. Guara
Jose  
#7 Posted : Friday, October 14, 2005 7:34:32 PM(UTC)
Jose

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 1/19/2005(UTC)
Posts: 1,065
Location: Koh Pha-Ngan, Earth

Was thanked: 2 time(s) in 2 post(s)
Hindsight-based Power Pivots - Gasp! Who would have thought! Tell me it ain't so, Joe ... 8-[ Golly Gosh... what are we to do now with that truckload of formulae using the PP? :eek: jose '-)
Patrick  
#8 Posted : Saturday, October 15, 2005 1:47:57 AM(UTC)
Patrick

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 9/8/2004(UTC)
Posts: 2,266

Was thanked: 1 time(s) in 1 post(s)
We have discussed this issue many times ... So no, it's not a shocker Jose. Do we really need to go over it again ... :x Alright then ... I tested the power pivots plug in many times and there were a few cases of hindsight results on the contrarian and pivots functions. So yes most the system / expert that came with the plugin will report results that are actually much better than they would be in real trading. Again, as this has been said many times, the plug in itself is maybe not something I would recommend. :wink: Now regarding henry's formulae, you will also notice that most Henry's formulae use the syseval and trend direction functions ... While I have not tested all of them I'm fairly positive that they are valid and can be useful. And then again, taking the risk to repeat myself here, you have to test and paper trade ANY system no matter who wrote it ... Patrick PS: Henry and Jose, quit it ... Please ...
henry1224  
#9 Posted : Saturday, October 15, 2005 3:03:39 AM(UTC)
henry1224

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 10/29/2004(UTC)
Posts: 1,394
Location: Glastonbury, CT

Was thanked: 2 time(s) in 2 post(s)
Guara, let me explain the powerpivot function Pivots are formed when the middle bar is greater or lower than the surrounding bars. Pivots have greater importance when they have developed into primary, major, and intermediate pivots. As a trend is developing, the first recent pivot will be a minor pivot. the trend continues and the Pivot will change it's strength from minor to Intermediate and so forth. It's value as support and resistance remains constant. now let me explain trend direction As the first pivot is formed, it's a minor pivot! The price retraces, and then the trend continues to a new high and now the price retraces back, Thus forming another new pivot. Since this pivot is greater than the previous Pivot the trend direction is up. It will continue until the Pivot is lower than the previous Pivot. Now when you are trading a daily chart and using the pivots from a weekly Time frame as the trend direction, the trend direction is not likely to change before your position is stopped out!! Now let me explain the Syseval function It is just a latch that takes the first signal and waits for an exit or opposite entry signal. Since it is an external function, the calculations are done outside of Metastock, speeding up the explorations and system tests. It eliminates the use of Cum() and barsSince and Prev and set and reset variables that you see in other types of code. The user manual that comes with this add-on doesn't give enough documentation for the external functions, also the systems and experts that are prepackaged are password protected. All of these factors lead to a lot of frustration, which leads to misinformation and misuse of these functions. If the data is EOD or Real-Time, the pivots behave the same! Of course the Time factor will be different! You didn't tell me what Time frame you are trading? 5 min, 10 min bars, 30 min or 60 min.
guara_riua  
#10 Posted : Saturday, October 15, 2005 4:43:46 AM(UTC)
guara_riua

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/24/2005(UTC)
Posts: 77
Location: Canada

Henry I appreciate the time you have taken for adding some clarifications to the correct use of PP+. If the SysEval function is totally independent of Pivot Types, then I totally agree with your statement that its use should be very much recommended. Support/Resistance I agree also that the support and resistance values will remain constant, independent of the pivot type, but think about their significance: - there are Minor/Intermediate/Major and Primary support/resistance lines according to their corresponding Pivot Types. When a Minor pivot evolves into a Major pivot and so on, the significance of the support/resistance line is changing too. Unfortunately this will happen after a substantial delay in real-time. You can have a situation when a Minor support/resistance line is crossed, and after quite some time this line becomes/evolves into an Intermediate/Major support/resistance line and your stop loss/limit signal will appear "post-mortem" if it's based on a condition where the price is crossing an Intermediate/Major support/resistance line. Obviously there will be no "post-mortem" signals if the buy/sell condition for the price crossing a support/resistance line will use only those values generated by Minor pivots only. Trend Direction I agree with your affirmations about the TrendDirection function only for the case when using Minor Pivots Type. In the case when using the TrendDirection function for pivot type higher than Minor its output will change as the pivot evolves from Minor to Intermediate and so on. Using a higher time frame this will happen just the same, but it will take a much longer. I agree than meantime, the trade would have been already triggered/stopped out if the decision was based on Minor Pivots only, for the TrendDirection function. If the TrendDirection function is set up to work for Intermediate pivot types or higher then, as the pivot type changes/evolves from Minor to higher ones, so will do the function output: it will change BACKWARDS IN TIME. I agree than when using only Minor Pivot Types for trading decisions with any function from the PP+ add-on it will work OK in real-time, but I believe that ALL the functions that use Intermediate Pivot Type or higher will generate non-usable signals in real-time trading. I am gladly welcoming any suggestions/comments that will prove me wrong, because if it will happen, then I will have the chance to continue to use this product with ALL its features. I used the 5 min time frame to "catch" the Pivot Type changing event, but for trading I use 5min, 15min, 1 hr and daily time frames as 4 separate charts. I din't check the output for the TrendDirection function yet. I've just deducted logically what will hapen when the pivot type changes. I intend to check it very soon and then I will post my results. Regards Guara
guara_riua  
#11 Posted : Saturday, October 15, 2005 4:00:39 PM(UTC)
guara_riua

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/24/2005(UTC)
Posts: 77
Location: Canada

I have checked these 2 PP+ functions and below are my findings. 1) Trend Direction Turned out that Henry is right: this function is not totally dependent on the pivot Type, but rather is changing its output when the CURRENT PRICE is taking out the value of the Last Pivot Type PRICE. This behavior ensures the stability of this function in relation to the Pivot Type Dynamic Variation in real-time trading, mainly because when this event happens, it is at a time long enough after the Pivot Type already changed. 2) Nth Pivot Price This is not the case with this function. It is changing BACKWARDS with the Pivot Type Dynamic Variation. I used EUR/USD 60 min chart and in the lower inner windows is ploted the "NthPivotPrice" function from which I've substracted $1.00 and the multiplied by 100 in order to make the price variations more visible. I haven't checked for other PP+ functions yet, but I recommend to all the PP+ owners to verify thoroughly the behavior of the PP+ external functions they use with regard to the Pivot Type Dynamic variations. I hope this post clarified some very shadowed areas in the PP+ use in real-time trading. Guara
Keltner  
#12 Posted : Saturday, October 15, 2005 5:50:50 PM(UTC)
Keltner

Rank: Member

Groups: Registered, Registered Users
Joined: 10/15/2005(UTC)
Posts: 22
Location: SLC

:evil: Man I am wore out about topic . ... Pivots change and you need to use ExtFml( "PowerPivots.PivotTrigger")=1 to back test systems. Some of the systems in the plug in did not have this in the back tests with out this code your back tests will have errors. point being IF you use the nth pivot price fuction in a back test you should also use the ExtFml( "PowerPivots.PivotTrigger")=1
StorkBite  
#13 Posted : Sunday, October 16, 2005 4:25:26 AM(UTC)
StorkBite

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 3/19/2005(UTC)
Posts: 2,995

Was thanked: 14 time(s) in 10 post(s)
Hey Keltner- Welcome to the forum!
PowerPivots.PivotTrigger")=1[/quote wrote:
Thanks for clarifying use of the backtesting procedure. I'm sure it'll be useful information.
Quote:
Man I am wore out about topic ...
It's too early in your membership to be worn out. There's so much left to do! :wink
guara_riua  
#14 Posted : Monday, November 7, 2005 4:45:50 AM(UTC)
guara_riua

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/24/2005(UTC)
Posts: 77
Location: Canada

This is going to make some waves. I've generated 3 kinds of pure mathematical stimulus files to check 2 of the most important PP+ external functions: 1) TrendDirection 2) Higher Time Frame indicators that can be plotted onto the current one. Bad news for both. For high frequency input variations (low bars per cycle count) the PP+ function output is far from being consistent. It will eventually kind of stabilizing for much lower frequencies (higher bars per cycle count), but will still show at some point a major functional discontinuity (see the MedianPrice 4x in the atttached document) or a smaller inconsistent behavior for the TrendDirection. I have tried to upload the stimulus files, so everybody to be able to replicate my findings, but I've got kicked out by the server with a 2MB limit error for the file size (even if the max file size for the attachments are stated to be 8MB). I am not sure if I can upload 2 attachments, but if this will not work, you can get the describing document from my other post : "ADSI Stimulus Files" that describes also how to convert CSV files to XLS files and then to use the Downloader to convert them to MetaStock format. After these tests I've abandoned the use of the PP+ add-on. For me the most important external function was the TdataCreate/TDataLocalize that was supposed to provide a way to calculate and plot an indicator from a higher time frame onto the current one. I found the TrendDirection for Pivot Types higher than Minor are quite late for the exit, when used in a higher Time Frame. I believe a good equity management stop/loss order will kick in way sooner than a signal generated only by the TrenDirection for Pivots Type of Intermediate or higher, in a higher time frame. I hope somebody will develop soon a similar function that will work like TDataCreate/TDataLocalize. I will add it to the Forum.DLL wish list anyway. Sorry to bring these bad news. This add-on could have been an attractive one with its rich external functions list. Guara
Jose  
#15 Posted : Monday, November 7, 2005 5:53:13 AM(UTC)
Jose

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 1/19/2005(UTC)
Posts: 1,065
Location: Koh Pha-Ngan, Earth

Was thanked: 2 time(s) in 2 post(s)
The moral of this story: If you are looking for 100% trust when it comes to possible hindsight-based "shifting" functions, trust only pure MetaStock code. Test any external functions thoroughly in real-time. jose '-)
henry1224  
#16 Posted : Tuesday, November 8, 2005 1:13:13 AM(UTC)
henry1224

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 10/29/2004(UTC)
Posts: 1,394
Location: Glastonbury, CT

Was thanked: 2 time(s) in 2 post(s)
Quote:
I've generated 3 kinds of pure mathematical stimulus files to check 2 of the most important PP+ external functions: 1) TrendDirection 2) Higher Time Frame indicators that can be plotted onto the current one. You have generated 3 kinds of pure mathematical files. they are not based on any type of trading principle, just computer generated random numbers! Have you tested these findings on actual traded securities? Did you use the enhanced system tester, Or Tradesim?
Quote:
I found the TrendDirection for Pivot Types higher than Minor are quite late for the exit, when used in a higher Time Frame. I believe a good equity management stop/loss order will kick in way sooner than a signal generated only by the TrenDirection for Pivots Type of Intermediate or higher, in a higher time frame. Please disclose your system as to how you use Trend Direction? Which Pivot Type are you using? have you tested ADSI External functions ? Bollinger's Plug in?, Nison's Candle plug in? Murphy's Plug in? How about the other Add-ons?
henry1224  
#17 Posted : Tuesday, November 8, 2005 1:29:43 AM(UTC)
henry1224

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 10/29/2004(UTC)
Posts: 1,394
Location: Glastonbury, CT

Was thanked: 2 time(s) in 2 post(s)
3 csv files tested on one system? How about testing the one system on 30 securities over a 5 year period? Test it on daily charts, or hourly intraday charts. Testing any system on 1,2 or 3 assets doesn't prove anything except that there has not been enough statiscal results to rule out or prove anything! Let us hope that medical research doesn't go the way these tests are going.
henry1224  
#18 Posted : Tuesday, November 8, 2005 2:01:08 AM(UTC)
henry1224

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 10/29/2004(UTC)
Posts: 1,394
Location: Glastonbury, CT

Was thanked: 2 time(s) in 2 post(s)
ExtFml( "PowerPivots.ArrayLoad", BIN#, HOOK) ExtFml( "PowerPivots.ArrayRelease", BIN#, HOOK) ExtFml( "PowerPivots.ArraySave", BIN#, ARRAY, HOOK) ExtFml( "PowerPivots.BarsSinceNthPivot", PivotTYPE, INSTANCE) ExtFml( "PowerPivots.FibLadder", LadderLevel, Length) ExtFml( "PowerPivots.NthPivotPrice", PivotTYPE, INSTANCE, TFACTOR) ExtFml( "PowerPivots.PivotBreakout", STRENGTH_LongTerm, STRENGTH_ShortTerm) ExtFml( "PowerPivots.Pivots") ExtFml( "PowerPivots.PivotTrigger") ExtFml( "PowerPivots.Retracement", PivotSTRENGTH, Retrace%, TFactor) ExtFml( "PowerPivots.RetInd", PlotIx, PivotArray, TrendArray, PivotSTRENGTH, Retrace%) ExtFml( "PowerPivots.SysEval", LongEntry, LongExit, ShortEntry, ShortExit, HOOK) ExtFml( "PowerPivots.TDataCreate", OHLC, TFactor) ExtFml( "PowerPivots.TDataLocalize", TDataArray, TFactor, Offset) ExtFml( "PowerPivots.TimeCapsules", PRICE, TFACTOR, OFFSET) ExtFml( "PowerPivots.TrendDirection", PivotSTRENGTH, TFACTOR) Which ones work? Which ones have bugs?
guara_riua  
#19 Posted : Tuesday, November 8, 2005 2:36:31 AM(UTC)
guara_riua

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/24/2005(UTC)
Posts: 77
Location: Canada

Henry In my opinion, when a function is not consistent when using a clean mathematical stimulus it only proves it was not debugged well enough or its definition is flawed. Even if it may generate acceptable output most of the time, my test showed that the function output depend not only on the input data, but rather have a kind of feed-back that is able to generate oscillations type response. Statistically it may still provide positive results, because the inconsistent output appears relatively rarely. All I am saying is that to find that a function is instable, it doesn't have to pass through hundreds of tests, when it's failing some basic ones. It could be still used with statistically acceptable results. And I haven't had to test a trading system. The TrendDirection and TDataCreate/TDataLocalize external functions were my only test subject at this time. I am not trying to discourage the users that are happy with what they get from this add-on and intend to continue to use it. I just tried to clarify some of its external functions behavior, first to myself, and then I just shared my findings with the forum members. I am not trying to change or impose any opinions of my own onto anybody. Every trader should be able to make its own decisions. As I remember from some of the best trading books I've read, a major rule every trader shall follow is to not have his trading decisions influenced by news, rumors or other trading analysts beliefs, but to follow its own system and to improve it. These were my 2 cents. Guara
guara_riua  
#20 Posted : Tuesday, November 8, 2005 2:54:49 AM(UTC)
guara_riua

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/24/2005(UTC)
Posts: 77
Location: Canada

Henry Please download the document I've mentioned in my other post "ADSI Stimulus Files" with the Title "PPP_TF_Samples.doc" You will be able to see the type of stimulus I've used. It is not RANDOM at all. It's a continous and predictable function because is a weighted sum of 2 and respectively 3 sine waves. It also provides at the end the formulas I've used for the indicators that are part of the screen shots. I have no intention to test all the add-ons in the world. I started this one because I saw on the charts some functions behavior that rised my attention. I already mentioned the functions I have tested and I have no intention to go further to test the rest of this package. I may even try to use the SysEval for my next system. I am focused now to build a trading system based on ADSI package and Dr. Ehlers work. I will let the forum know if I will reach an interesting result. Guara
Users browsing this topic
2 Pages12>
Forum Jump  
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.