System Improvisation-Thought process!

[kuriako]

"A trading system can be either a computer based program or a daily advisory service that instructs you on what trades to take. Most popularly trading systems are software that runs on a computer. The computer instructs you on what and when to enter and exit trades.
There are trading systems that trend trade, day trade or perform a combination of the two. Some trade one market or sector while others are broad enough to trade many markets."-Chadwick Investment Group

A System is an approach to buying and selling stocks, bonds, commodities or anything that is bought and sold on a regular basis, especially in an organized market. A system has the following components:
1. Rules for when Buy or Sell (entry and exit – Including Optimization)
2. Calculation of a purchase or sale price
3. Calculation of an optimum quantity of shares to purchase (money management)
4. Rules to prevent huge losses (Stop Loss)
5. Rules for profiting when the market is going down (going Short)


This thread is for helping me find ways to INVENT a replacement for the holy grail!
In most of the system development process we read books and copy the ideas presented in them and implement the system to make profit.

And most Technical analysts advocate for a simple system.
A system is simple or complex makes no difference if it is working fine in current market! If it is draining your account then yours system is under performing or it is having some flue!

Here the discussion is oriented to improve a system! If you are having a good system you will be still searching methods to improve it.

In this multi dimensional world where knowledge is power, we exchange it in this system to get it multiplied

I am a newbie to the market and is having ZERO trade experience in real market! I am here to discuss with fellow mates, professional traders and with my brothers. The discussion will be oriented to methods of improving the system.

Be it improving the data fed to system, the process happening inside system, the output given by system...

But as always lets KISS- Keep It Simple and Spiffy!!! (we are no stupids!!!)
Simple things are easier to understand and hence we go by simple means!

But the technical stuff could go to any degree of complexity. It is upto the readers to take it or leave it!

Lets kindly refrain from arguments and make time for discussions

Thank You
Kuriako

 

[kuriako]

I left Saint's thread, leaving it for New members to learn! I myself am a new member but Saint got unhappy with my ROCKET science discussion and the Linda Raschke episode
I AM A MECHANICAL ENGINEERing student, having zero experience in trading. I haven’t done a single trade in any market except for local grocery market where I sell my farm products! And as Saint said I am not to young to begin! Aged 21 now and is in final years of engineering. Warren Buffet started it in 11th year…Geez how I missed my golden 10 years-morning over it right now!
My way of doing things is experimenting! In fact here I am trying to mix and make an NEW system where the system will religiously prevent whipsaws all by itself!
I hate the idea of Fuzzy logic and neural network because “I haven’t learned it yet”. I have noticed similar comments from some members on this idea. Just review it. If a guy who failed for Mathematics starting from 3rd standard onwards can understand it you intelligent people can surely make GOLD coins from this brass.
The author is a just pass mathematic! And somewhat good engineer and a better manager and a voracious reader! These all caused this thread!
Let me give a concise format of what was discussed in other threads:-
"I was surprised yesterday when I saw a youngster in the chat room who has not even done a single trade searching for a complex trading system."-Kartik
That youngster is non other than ME this humble author!
And I asked for a complex system not to make money multiply X times in one second...But to know how a complex system will look like and WHAT YOU Pro`s call as complexity. Young age may be reason of curiosity!
I am thinking about applying "MODULATION" technique to the original data (eod price) and to modulate it as a signal. In modulation we will be amplifying the signal at some points( say the turning points are elevated more and the troughs dug more down)
The logic is very similar to FM or the plane old AM
This elaborated and amplified signal can be fed to the SO CALLED SYSTEM and can be processed...
End effect: An enhanced view of market and data CLARITY
Justification PART!
I still don’t know why I typed this
"We dig a well from TOP still not knowing whether water is there!"
And I just took the approach of the well digging...
THEY CALL IT CONTRA
In the usual approach we study in KG, then to the 1-10 standards, matriculate to 11th and then graduate! Finally to add some essence will get masters too!
I am following the GURUKULA study model combined with LAZY PROGRAMMING. In LAZY method we will study about a thing only when the need for that thing arises. We won’t learn it and prepare for the future. Just learn it the moment you want to apply it. This learning exercise has the advantage of learning only the required things. And if requirements are more, we learn a lot. This world is full of information and that too in accessible format! But everybody want to learn everything and to memorise it.
Sorry I am a LAZY program!
And I move from complexities untying them to make simple things to come out! Gurukula study model in the sense student learns by himself and will tell the GURUJI about what he observed. Guruji will correct him and guide him to more knowledge and ways to untie complexities. And for me the GURU is this forum!

 

[kuriako]

Continued:-
When we are digitising a wave we are losing much information. And the buy sell signals are just based on an event. In most case the event lags behind the actual event in market ( MABIUTS, MA... lag). If we are enhancing/optimising our signal we will be able to get more info fed to the system in a CLEAR MAGNIFIED manner. In that case I guess the system will be able to reduce some lag. I think the OPTIMISED signal part is clear. The pros there have experience with MA systems where they will be reducing lag.
But the problem is whipsaws in that case is evident. I was impressed by the Scientist Asanavales work on MABIUTS system of Karthik. The MABIUTS HB system was wonderful and I am yet to test is with my SIGNAL wave
The problem of that whipsaw is caused because we are smoothening our signal and there by losing informations from signal...
If we are smoothening an enhanced signal then we may be able to reduce the lag and also whipsaws...
I first posted all these under label FOOLISHNESS.
But the backtests proved that it was not complete foolishness!
And uasish commented like this-
“Plz avoid Saint's thread,leave it for New members to learn.
Kuriakose,
With my common sense i understand modulation may not be desirable becoz in FM we amplify the Signal while transmitting becoz as it distorts while traversing through Noise so enhanced Gain / attenuation when after distortion lands up at receiver points there is more clarity. But here in actual Mkt we want to filter the NOISE from Signal by an ideal 'passband' filter with (no gain/attenuation through out) to get a better through put. SMA as karthik mentions is FIR (Finite Impulse Response) in general means Finite becoz there is no feed back from impulse to filter. Mathematically it is a Discrete Distribution (known as Geometric Distribution means a discrete analog of the exponential distribution)
EMA is IIR (Infinite Impulse Response) becoz there is continuous feed back from impulse to filter.
Mathematically it is a Exponential Distribution (means a continuous analog of the geometric distribution)
Though i personally is not in tune with Karthik becoz in both these Distribution the basic premises is 'memory less ness ' & i strongly believe Price signals have in built Memory.
So Kuriakose your non understanding of Mkt (due to just inexperience nothing else)
1)You want to make more Clear Signals traversing through Noise ,by modulation;
2)We want to extract Clear signals from Noise+ Trend , by Filters.
Hope to interact with you in future as you are having better knowledge than me on these filter issues though i always have our Resident Specialist Qualified Engineer Karthik to fall back upon.
Asish”

I replied to Asish like this!

As far as I analysed the market signal is very simple and with out noise!!!
It is only a combination of four signals namely OPEN HIGH CLOSE and LOW
These signal propagate through the space called PRICE and causes variation!
Another signal is volume! Sometimes the signal volume interfere with the price signals! I think this is called as SUPPORT and RESISTANCE POINTS!
I was thinking about developing a system! So as per several posts here I made a frame work!

When I analysed the parts of system I looked at the whole information from market. In the first step of all(all most all) systems we shunned OPEN HIGH and LOW information. We started working on the CLOSE price only. So we filtered the signal on the first place itself and made it a single simple signal!
I could nt believe it! Because how could we get more info from system if we are feeding less info!
The moving on to next steps I found the signals being smoothed! (with MA)
And at the very point we got lags because we lost more info again!
And then we made a condition! When signal 1 interferes with another signal 2 (smoothened forms of the same parent signal) we GET a Binary signal! Means a DIGITAL wave is formed from the signal!
In all the systems we processed and destroyed the pure market signal!
An the the signal being CYCLIC(markets are cyclic-correct me if wrong) we expect there to be noise in the signal!
But actually there is no noise in the signal itself but it COULD be the interaction of price signal with volume signal that causes distortion and imparts cyclic nature.
I hope it is clear why I dont think there is noise!!!
Asish then asked for my view on the VOLUME PRICE relationship- as I was unbiased

 

[kuriako]

My reply-
We need to look at the price and volume as two separate waves! And one wave causes some effect on other wave in "SOME" form. As far as I know price influence volume and volume influence price. In a market where the number of trades are high( we got another signal!!! Number of trades ) there used to be HEARD mentality. If the small traders see a huge volume increase they do something. (what is that thing? Buy or sell- Inexperienced uneducated fellow here!)
So this causes the price to go up or down!
In other cases when price goes up, the increased number of trades jump up to buy more. So volume increases!
The volume-price relationship is dependent more on number of trades I guess!!!
From what I saw, When a large trader sells in large quantity if one large trader comes in and buy, there is no remarkable effect on both our waves!(price wave and number of trade wave) Only volume wave leaps! But when the large trader comes in and sells it in pieces(multiple trades), there is considerable effect on prices!”

Later I went through the HOME ASSIGNMENT given by CV. He is cruel than my ENGINEERING TUTORS!!!
He had given me the tedious reading exercise through a 500+300+100 page book to have a better insight! Poor me started reading and understanding it! The latter is somewhat false. Actually started reading it!
It was at this time a thing clicked me
Market has noise! If there was no noise then how could we have four different prices (OHLC) on the same day? So some noise is affecting price at every point (day) on the price signal. I considered price signal as a wave send from same transmitter and received by another receiver! So in that case we get the four signals transmitted and all with varying amount of noise received at receiver!
When I saw the advertisement of a Wi-Fi receiver (can only dream about those things) it had two antennas for reception. Naturally the two waves will be having varied noise. Moving with that thought I found ways to reduce noise and combine signals.
I was not easy as Karthik said. But untying the simple idea from that complex system we get a real simple logic. But still the simple idea is a complex mathematical equation. This is the problem of QUANTIFYING! But helpfully there were cosine and sine functions available in MATLAB and when I checked that was available too in AmiBroker.
So directly coded the small beautiful long equation! At first when I press arrow keys the cursor movement from one day to another took about a second!!! I was thinking of the complexity then!
A reference to the HIGHER ENGINEERING MATHEMATICS-Grewal made wonders!!!
[I never bought that book! And never ever opened it before. And it was not at all mine!!!]
The equation was reduced with some simple Cos(A+B)=some long something and some equations like that!!!
Most of the variables cancelled each other and a simple formula was obtained!
The wave of price was thus plotted!

 

[kuriako]

To be continued with PRICE-VOLUME modulation!

 

[kuriako]

Simple jumping!

What I noted with the SIGNAL wave!

The system now is able to see market somewhat clearly(Still vague!!!)
It was able to follow the uptrend cleanly!

But the important draw back noted was lag in buy and sell sell signal at peaks and trough by 2-3 days!!!

Overall profitability is improved and less risk as per back test. Back test period 12/12/98 to 12/12/2006
Portfolio backtest on all stocks in NSE with 5 lakh initial cash

Please comment!

 

[sikandar]

Best wishes
Continue

 

[CreditViolet]

Later I went through the HOME ASSIGNMENT given by CV. He is cruel than my ENGINEERING TUTORS!!!

He had given me the tedious reading exercise through a 500+300+100 page book to have a better insight! Poor me started reading and understanding it! The latter is somewhat false. Actually started reading it!
It was at this time a thing clicked me
Market has noise! If there was no noise then how could we have four different prices (OHLC) on the same day? So some noise is affecting price at every point (day) on the price signal. I considered price signal as a wave send from same transmitter and received by another receiver! So in that case we get the four signals transmitted and all with varying amount of noise received at receiver!

Good start.

Side point.Crude mathematical models do not have real predictive power when there is real money on the line.

 

[kuriako]

CV! CV! CV!....

Updated...Poor me! The cost of my poor mathematics paid!

Reduced my entire funny complex formula to a simple ONE LINER!!! Not by me, but a mathematician here!

Now I understood a very important point in life- Behind every complex thing there is a simple logic! And we throw away that logic because we think it as silly!

Going through the "literature" CV!

Now I have a nice price signal with me. It is not filtered but it is a NIOSE removed signal(can call it filtering- but i call it decyphering)

I think I must not do more work on modulating volume signal on price signal!
I am planing to feed two of these waves!!!

So for now I GOT INPUT FOR MY SYSTEM!
Now where is my system...

To the concept generation part! A fresh new concept devoid of filters,smoothers(ma,ema...) and A/D convertor!!!

I still don't know whether i will be able to perform it to get just a BUY or SELL


Any comments on graphs and results, CV!!!

 

[kkseal]

Hey man Kuriakose, Talk less Work more.

And when your 'noise removed signal' becomes 'plottable' post a chart of the same for us to see (& WoW

Btw, Complex = sum of simples.

Regards,
Kalyan.