Signal Bender

free & uncensored discussion arena for TheRumpledOne

Moderator: moderators

User avatar
newschool
rank: 150+ posts
rank: 150+ posts
Posts: 489
Joined: Fri Aug 21, 2009 2:26 am
Reputation: 0
Gender: Male

Postby newschool » Wed May 26, 2010 3:01 pm

So, MQL would need to have a facility for using the Output from Indicator_1 and using it as Input for Indicator_2, and that is just a small example of the MetaIndicator layering.


An EA is basically already doing this, loading different external indicators [iCustom function], and then mixing their output [in form of buffers] to form a trigger. So currently I dont see this as a major obstacle.

From my understanding, MT4 starts to slow down when you significantly increase the number of Indicators running simultaneously and it is supposed to have an upper Indicator limit, beyond which you cannot load additional Indicators. In other words, MQL just can't handle the job of processing complex trading systems, fluently.


The only time MT4 slowed down on me is on the calculation of a total Bars history. Now I limit them to few hundreds.
Concerning the last phrase, as I see it, making EAs is easy and free. Coding in a fully compilable language cost time/money, so for me its a second step not a first one.

--

K833 = E828+0.0025
K834 = G828+0.0025

All your Excel cells are simply variables. Hell, why not use the same name while coding? It would then only needed to copy paste your functions, and just manually change the appropriate syntax when comparing cells/variables.

Please add www.kreslik.com to your ad blocker white list.
Thank you for your support.

User avatar
forexbob
rank: 150+ posts
rank: 150+ posts
Posts: 247
Joined: Mon Apr 20, 2009 9:52 am
Reputation: 2
Location: Planet Earth
Gender: None specified

Postby forexbob » Wed May 26, 2010 3:57 pm

newschool wrote:
So, MQL would need to have a facility for using the Output from Indicator_1 and using it as Input for Indicator_2, and that is just a small example of the MetaIndicator layering.


An EA is basically already doing this, loading different external indicators [iCustom function], and then mixing their output [in form of buffers] to form a trigger. So currently I dont see this as a major obstacle.

From my understanding, MT4 starts to slow down when you significantly increase the number of Indicators running simultaneously and it is supposed to have an upper Indicator limit, beyond which you cannot load additional Indicators. In other words, MQL just can't handle the job of processing complex trading systems, fluently.


The only time MT4 slowed down on me is on the calculation of a total Bars history. Now I limit them to few hundreds.
Concerning the last phrase, as I see it, making EAs is easy and free. Coding in a fully compilable language cost time/money, so for me its a second step not a first one.

--

K833 = E828+0.0025
K834 = G828+0.0025

All your Excel cells are simply variables. Hell, why not use the same name while coding? It would then only needed to copy paste your functions, and just manually change the appropriate syntax when comparing cells/variables.


Hi Newschool,

agree with you that with icustom you can layer indicators, and fuel indi2 with the outcome of indi1. But that is certainly not multithreaded. maybe not needed. maybe it is and then a rework will be needed. (indicators merged or ... )

so maybe it can be done, but i can not make an estimate how large Signalbender's system is. (at least i can tell referenced cells go quite high. and maybe even multiple sheets are used)
And there are more unknown factors. maybe the system can 1st be optimized before going to metatrader. etc etc

i know for sure that the level3 zigzag indi is a cpu hog, certainly when attached to multiple charts. but prob that can/is also optimized by some already.

further is fully compiled also an illusion, because metatrader goes to an intermediate code. (easily decompiled)
The highest Forex rebates: http://www.cashbackforex.info

SignalBender
rank: 50+ posts
rank: 50+ posts
Posts: 108
Joined: Mon May 10, 2010 9:36 pm
Reputation: 0
Gender: None specified

Postby SignalBender » Wed May 26, 2010 7:16 pm

forexbob wrote:?? flying ??


Shoptalk, nothing to worry about.

forexbob wrote:i would prob then build it around a DB (mysql,oracle, ..) and yes i admit
excel can be a handy part. interfacing between excel and DB can for example via excel-db http://www.excel-db.net/index.htm , but there are prob other tools that fit.


An RDBMS would be great for storage of historical data, but not great as a real-time data handler. This is the piece of the puzzle that is rather illusive to me right now, a real time data handler for the Excel engine to run its calculations against. That's why my ppt shows the Excel Surrogate object in the conceptual diagram. Building that, I believe is the key to keeping Excel in the loop because it is Excel that is doing all the heavy lifting. So, each one of its mathematical calculations against real market data, needs to point to a data source that it can work with. Thus, the Excel Surrogate idea.

If I point Excel to SQLServer, Oracle, Sybase, IBM Informix or MySQLServer (as an example); you still have to connect AND import. Additionally, before you import the data, you have to tell the Datasource Manager where you want the tables from the RDBMS to be located within Excel. Well, this presents a huge problem. If you look closely at the pic of the flat-file database that I use in Excel, you see EURUSD historical data, with the very first row containing EURUSD real-time DDE linked data from MT4.

The problem is not just giving the Engine access to data. The problem is giving the Engine access to properly formatted data that each formula can link to. In other words, the Engine expects to "see" the data in its "native" format - and for Excel, that means specific rows and columns that mean something to Excel. This is why an Excel Surrogate is needed that can also remain persistently and intelligently connected to the data feed, whereupon any disconnect from the source, will result in an auto-population of the missing data/bars in the Surrogate. This way, the Engine won't skip a beat.

I know what the solution should be - but building it, is another challenge altogether.

forexbob wrote:the tradesignals from excel could then go directly, or back to a DB.


Or (if you look at the ppt diagram), the trade signal, order profile and supporting indicator output can be passed in real-time to a .Net Form which itself would be embedded into the trading platform (either ProTrader Multistation or FXall). The trading platform would be receiving a properly formatting string (output from the signal engine in Excel) containing all trade execution instructions to be run against the trading platform. This take the database out of the business of passing trade execution instructions to a trading platform. The platform would get the instructions directly from the engine itself. Again, check the ppt diagram.


forexbob wrote: or let the whole build as a proof of concept without meaningfull 'indicators' within the excel engine.


Hmmmm. Very, very, very interesting. I had not thought of that before. I know how it would work with ProTrader Multistation. They have their own SDK that would be used to embed a .Net Form. That .Net Form would be used as the "Trade Control Panel" inside ProTrader. The Excel Engine would pass its output to the .Net Form and the Form (just in case I need to do some manual adjustments to the trade profile) will drive the trading platform. So, I think your idea make sense for development purposes, because all the .Net Form "needs to know" is the structure of the formatted order instructions - not where the instructions came from.

Hmmm. I'll flush this out some more on paper, but this might solve a significant problem that I've had for quite some time now. Thanks! :)


forexbob wrote:or get some excellent programmers do a NDA....


NDAs are broken and busted all the time as Intellectual Property Rights vary from jurisdiction to jurisdiction, all over the planet. There are probably more intellectual rights violations law suits in the world, as there are real property right violations cases. It would be a nightmare to manage and in the end, the full system would get exposed one way or another. After I'm out of the market for good, I won't mind giving it away. But, for now - it is too powerful to simply give away, for reasons that are not exactly "clear" to everyone.

There are some aspects of the system that I can talk about in public and others that must remain under wraps for now. One day, it will get released, but now is not the time. I have not yet reached my target revenue goals.

Thanks Bob, for the brainstorm. I like the 'Blind/Empty POC' idea - I'll flush that out later. ;)

SignalBender
rank: 50+ posts
rank: 50+ posts
Posts: 108
Joined: Mon May 10, 2010 9:36 pm
Reputation: 0
Gender: None specified

Postby SignalBender » Wed May 26, 2010 8:15 pm

Thanks, Newschool. You've got mail.

SignalBender
rank: 50+ posts
rank: 50+ posts
Posts: 108
Joined: Mon May 10, 2010 9:36 pm
Reputation: 0
Gender: None specified

Postby SignalBender » Wed May 26, 2010 9:01 pm

ForexBob,

I don't think MQL and MT4 were designed for robust trading systems. I had an email exchange with the people from Molanis, developers of Strategy Builder and Indicator Builder for MQL. They intimated that 100 Indicators in a trading system was a "problem." Basically, saying that if I had 100 Indicators, then there was something wrong with my design. I basically, laughed to myself and thought: He really has no idea, does he!

There is this basic theory out there that says, if you have more than a few Indicators in your trading system, that all you are doing is fractional copying of a handful of Indicators. I even read one author on the subject who said that the Trader with a large number of Indicators in their system, simply does not know what he/she is doing. Again, I smiled. :)

Most of these authors have never developed a Language the describes market behavior. Most of them make their comments based on the current set of Traditional Technical Indicators that everyone uses. RSI this, or CCI that... So, when they talk about having "too many indicators," they are typically referring to traditional average based indicators. More than 95% of the Indicators in my system are non-traditional. Custom designed indicators that have nothing to do with tradition or in most cases, not even conventional thinking.

So, when you have a trading system where "100" indicators is required to produce a trade signal - then "100" can never be "too many." Thus, these authors completely miss the point entirely, because the nature and structure of the system is what dictates the measure of "too much." Not the absolute value of the number of indicators used in any one particular system.

MT4 was never designed with Delta based trading in mind. And, to its credit, it never could have been designed that way. Among other things, this system employs a layering of indicators, all designed to measure and quantify different aspects of the market at its core. At a higher level, the system then uses the output from those layers to construct a decision support layer of signals. And, from that layer, the system then produces the final trade signal after having hashed out the "best candidate signal" suitable for the market conditions in existence at the time of execution.

This is not a simple cross over system. It is an intelligent system that learns as it ages without becoming overly restrictive, or over reactive. After learning more about MQL, it becomes clear that it is not capable of handling such a system.

However, there are still things that I'd like to do in MQL that do incorporate some of the Indicators within the system, so I hope to be able to find an MQL coder to assist with those aspects. Newschool has demonstrated some interest, so I'll move in that direction.

A system does not have to be massively complex, to be profitable. The Bugatti Veyron 16.4 is not "necessary," but it is the most outstanding example of what supercar automotive engineering can accomplish to date. Not to mention, one of the fastest production supercars on the planet.

I like the Veyron, not just because of its speed and performance (that's a given). What I appreciate the most about the Veyron, is the detailed engineering that went into its design and creation. There is simply no supercar like it on the road today, from a purely engineering standpoint - and that's why I appreciate it so much. The Koenigsegg, probably the second highest performing production (street legal) supercar in the world, is my second most favorite. It does not have anywhere near the same level of elegant engineering and systems complexity as the Veyron, but I still appreciate it, because from a pure performance standpoint, it gets the job done better than anything else - other than the Veyron itself.

Veyron - complex, massively sophisticated, elegant: High Performance.
Koenigsegg - smart, straight forward design characteristics: High Performance.

I'll take one of each (thank you very much) for completely different reasons and I won't feel cheated with either of the two. They are both outstanding examples of what "performance" means - yet two entire different paths were taken to achieve that performance.

Good trading systems, are no different. Some will be like the Veyron. Some will be like the Koenigsegg. Neither will fail to make you very wealthy in the long run. ;) I simply prefer the Veyron.

Please add www.kreslik.com to your ad blocker white list.
Thank you for your support.

User avatar
forexbob
rank: 150+ posts
rank: 150+ posts
Posts: 247
Joined: Mon Apr 20, 2009 9:52 am
Reputation: 2
Location: Planet Earth
Gender: None specified

Postby forexbob » Wed May 26, 2010 10:02 pm

SignalBender wrote:ForexBob,

I don't think MQL and MT4 were designed for robust trading systems. I had an email exchange with the people from Molanis, developers of Strategy Builder and Indicator Builder for MQL. They intimated that 100 Indicators in a trading system was a "problem." Basically, saying that if I had 100 Indicators, then there was something wrong with my design. I basically, laughed to myself and thought: He really has no idea, does he!

There is this basic theory out there that says, if you have more than a few Indicators in your trading system, that all you are doing is fractional copying of a handful of Indicators. I even read one author on the subject who said that the Trader with a large number of Indicators in their system, simply does not know what he/she is doing. Again, I smiled. :)

Most of these authors have never developed a Language the describes market behavior. Most of them make their comments based on the current set of Traditional Technical Indicators that everyone uses. RSI this, or CCI that... So, when they talk about having "too many indicators," they are typically referring to traditional average based indicators. More than 95% of the Indicators in my system are non-traditional. Custom designed indicators that have nothing to do with tradition or in most cases, not even conventional thinking.

So, when you have a trading system where "100" indicators is required to produce a trade signal - then "100" can never be "too many." Thus, these authors completely miss the point entirely, because the nature and structure of the system is what dictates the measure of "too much." Not the absolute value of the number of indicators used in any one particular system.

MT4 was never designed with Delta based trading in mind. And, to its credit, it never could have been designed that way. Among other things, this system employs a layering of indicators, all designed to measure and quantify different aspects of the market at its core. At a higher level, the system then uses the output from those layers to construct a decision support layer of signals. And, from that layer, the system then produces the final trade signal after having hashed out the "best candidate signal" suitable for the market conditions in existence at the time of execution.

This is not a simple cross over system. It is an intelligent system that learns as it ages without becoming overly restrictive, or over reactive. After learning more about MQL, it becomes clear that it is not capable of handling such a system.

However, there are still things that I'd like to do in MQL that do incorporate some of the Indicators within the system, so I hope to be able to find an MQL coder to assist with those aspects. Newschool has demonstrated some interest, so I'll move in that direction.

A system does not have to be massively complex, to be profitable. The Bugatti Veyron 16.4 is not "necessary," but it is the most outstanding example of what supercar automotive engineering can accomplish to date. Not to mention, one of the fastest production supercars on the planet.

I like the Veyron, not just because of its speed and performance (that's a given). What I appreciate the most about the Veyron, is the detailed engineering that went into its design and creation. There is simply no supercar like it on the road today, from a purely engineering standpoint - and that's why I appreciate it so much. The Koenigsegg, probably the second highest performing production (street legal) supercar in the world, is my second most favorite. It does not have anywhere near the same level of elegant engineering and systems complexity as the Veyron, but I still appreciate it, because from a pure performance standpoint, it gets the job done better than anything else - other than the Veyron itself.

Veyron - complex, massively sophisticated, elegant: High Performance.
Koenigsegg - smart, straight forward design characteristics: High Performance.

I'll take one of each (thank you very much) for completely different reasons and I won't feel cheated with either of the two. They are both outstanding examples of what "performance" means - yet two entire different paths were taken to achieve that performance.

Good trading systems, are no different. Some will be like the Veyron. Some will be like the Koenigsegg. Neither will fail to make you very wealthy in the long run. ;) I simply prefer the Veyron.


SMILE !
The highest Forex rebates: http://www.cashbackforex.info

User avatar
forexjake80
rank: 50+ posts
rank: 50+ posts
Posts: 137
Joined: Sun Nov 15, 2009 12:20 pm
Reputation: 0
Gender: Male

Postby forexjake80 » Wed May 26, 2010 10:35 pm

signal bender, would you mind posting that stuff you sent newschool? Surely, everyone at kreslik is a trustworthy person.
It's time to kick ass and chew bubble gum. And I'm all out of gum.

SignalBender
rank: 50+ posts
rank: 50+ posts
Posts: 108
Joined: Mon May 10, 2010 9:36 pm
Reputation: 0
Gender: None specified

Postby SignalBender » Thu May 27, 2010 2:08 am

forexjake80 wrote:signal bender, would you mind posting that stuff you sent newschool? Surely, everyone at kreslik is a trustworthy person.


Actually, ForexJake80.

I have not yet sent anything to anyone. I've continued to post additional Indicators in this thread (TCD Extensions) without a response from an MQL coder. So, my assumption is that nobody wants to code them here at Kreslik.

However, the more I read Kovalyov and the "Laymans" MQL reference posted by Tom H. Liles, combined with what I already know, the more I am confident that sooner or later, I'll have my EA up an running using some of the fundamental elements from the prototype.

MQL is not going to give me 7thSignalTrader in full bloom and I don't expect that. But, I still want to push MQL to the limit by using what I can from the core prototype, inside an MT4 EA, on the lower time-intervals. There are some new concepts and ideas that I have not yet encoded or integrated into the core system. I'd also like to see if MQL can handle those as well.

I just want to be able to pull the trigger more than once per day, as long as the intra-day trading system [MQL] is at least as accurate as the single-shot inter-day trading Excel based system. To give you an idea of what the core system does, here's the last public trade profile that I posted online. I don't do public trade profiles anymore on a routine basis - been there (years ago), done that - done.

Last Public Trade Profile (Tradernumber7 = SignalBender

User avatar
newschool
rank: 150+ posts
rank: 150+ posts
Posts: 489
Joined: Fri Aug 21, 2009 2:26 am
Reputation: 0
Gender: Male

Postby newschool » Thu May 27, 2010 3:42 am

It depends how you see it. Instead of making 100 indicators, its also possible to add the formulas calculation part into the EA, thus not needed to load many exernal data (saving miliseconds). Also if I think about it, Im not sure Excel really has more processing power than MQL4... its not like it was optimized for that kind of task...

But concerning the Excel cells, I still think your calculations are pretty simple (AND/OR/SUM) and a hundred of them wouldnt be a problem (I saw some kind of work like that).

By judging of the different screenshots of your "controlboard", I do believe a full .NET made by a pro would be the top solution. But like you seems to accept in your later posts, properly chosed portions of the system could be running great with EAs.

PS: The Veyron has nothing on the McLaren http://www.youtube.com/watch?v=RXqSedWSu2k

SignalBender
rank: 50+ posts
rank: 50+ posts
Posts: 108
Joined: Mon May 10, 2010 9:36 pm
Reputation: 0
Gender: None specified

Postby SignalBender » Thu May 27, 2010 5:42 am

newschool wrote:It depends how you see it. Instead of making 100 indicators, its also possible to add the formulas calculation part into the EA, thus not needed to load many exernal data (saving miliseconds).


That sounds to me like aggregation by another name? Are you saying, aggregate as much indicator output as possible and then relabel that cluster of output as the new Indicator_1? Still, won't MQL have to know what the underlying formulas are and even as much, still calculate their values? Almost like thermodynamics in the sense that you can change the state of energy, but you cannot create or destroy energy. The calculations have to get done in some form or another, in some clustering or aggregation scheme or another; but in any case, they all have to get done/processed.

newschool wrote:Also if I think about it, Im not sure Excel really has more processing power than MQL4... its not like it was optimized for that kind of task...


Well, that's the thing about Excel. From a pure calculation standpoint, it will be as powerful as the processors (CPUs) behind it. I run a dual processor system and I routinely have banged up against Excel's maximum formula size restriction. So, many of the formulas had to be stretched out across multiple cells and linked together. There are lots of those examples in the system and I never run into any problems with Excel hanging or slowing down. It takes approximately 28 seconds to save the Engine file. That's how large file has become, yet no latency issues on any real-time calculations. I streamlined where I could.

I don't yet know much about the MQL/MT4 architecture, but from I do here from people that have been with the application since the earlier versions, is that it hangs or slows down when you add a certain number of Indicators. That says nothing about what those Indicators are comprised of, but that's what I hear and read from time-to-time.

newschool wrote:But concerning the Excel cells, I still think your calculations are pretty simple (AND/OR/SUM) and a hundred of them wouldnt be a problem (I saw some kind of work like that).


The baseline calculations are simple. The only baseline data points to calculate are the Open, High, Low and Close. So, that's easy to understand and all of those calculations comprise the core functions and core indicators. The higher level functions and higher level indicators, is where things stop being all that simple because multiple Indicators are uses to produce higher level Indicators, or what I call, Metadata. It is the Metadata layer within the system that causes complexity, not the baseline calculations.

Above the MetaIndicator layer resides the MetaBrain layer. Yet, another higher tier of MetaIndicators born out of other MetaIndicators (Meta about Meta). The actual trade signal comes from the MetaBrain tier/layer. At that level, there are zero core calculations anywhere and the system is no longer relying on "market data" to make its decision. The things that I've shown thus far are simple, yes - but the full spectrum system architecture is rather unique. I've never seen anything like it in Excel for trading and I've seen some fairly amazing trading set-ups using Excel.

newschool wrote:By judging of the different screenshots of your "controlboard", I do believe a full .NET made by a pro would be the top solution. But like you seems to accept in your later posts, properly chosed portions of the system could be running great with EAs.


Yes, the Tactical Trade Panel, is my pride and joy. In all these years, I've not seen anyone build anything even remotely close to that in Excel. Now, I've seen some better looking C, C++ trading panels, but nothing that rivals this in Excel. What sits behind the panel is the most interesting part as the panel itself is self sorting on each engine update, so the entire GUI shifts its order, while maintaining the overall look & feel with each engine update. I have a Tactical Panel as well as the Global View:

Image

Note: This panel is disconnected and pointing to out of date and un-synchronized data, so the output on the panel is not live. DO NOT USE.

The difference here, is that this Tactical Panel shows only the trade profile for a single currency pair. Note that the GlobalView Panel shows the Mini Profile for all pairs in the system. The Tactical Panel is a more detailed view of the trade profile and what the market/trade is doing at any given time.

I don't expect to recreate this kind of stuff in MQL/MT4. But, yes, you are correct. I would like to do some other thing using the lower-time interval data in MQL based on some of the concepts you see here in digital form. I am for the most part a Digital Dashboard type trader as it removes much of the forward looking anticipatory angst and replaces it with hard-core numbers and trade progress message. This Panel displays real-time messages about various aspects of the both the trade itself and the market. Just some FYI.

newschool wrote:PS: The Veyron has nothing on the McLaren http://www.youtube.com/watch?v=RXqSedWSu2k


I love the McLaren! But, she is no Veyron! Not by a long shot, guy. ;) The Veyron is a wet dream and an engineering masterpiece. However, in its day, the McLaren was most definitely the one to beat at 246 mph straight line speed. However, at 254 mph, nothing touches the Veyron, while still qualifying as a street legal daily driver. Just the transmission design alone on the Veyron, is worth the price of admission. It is a dual-shaft, dual-clutch, DSG, 7-speed active drive-line transmission that spools-up and auto-synchronizes the "next" gear before it is needed, producing a shift speed less than 150 milliseconds. 0-60 mph then back to 0 mph in under 4.9 seconds. That's absurdly ridiculous. And, all of that connected to a 4-wheel drive-train.

Like I said, I love the McLaren - she's awesome. But, the Veyron transmission alone has more engineering stamina than two McLaren's stacked on top of each other. In all honesty, the Koenigsegg CCXR, is the "new" McLaren and the Zonda R, is the new "Ferrari Enzo." And, to top it all off, you can actually buy the CCXR and the R brand new - whereas, you can't buy either the McLaren or the Enzo new anymore.

Zonda R

Please add www.kreslik.com to your ad blocker white list.
Thank you for your support.


Return to “TheRumpledOne”