FAO Charty

Dr Iraj

Dear Charty,

Glad to see you on SI. I wish I had your email to inform you of the new site. I believe you are into developing a new future prediction model. This is excellent. I started a research into this subject around 1990 ish where I was a research fellow at Manchester Poly (Currently known as Manchester Metropolitan University, All saints building) and by 1995 with lady colleague of mine from Imperial college we Developed a model using various Artificial Intelligence Neural Net strategies to forecast the short term price projection of stocks. Unfortunately the model did not perform due to our lack of understanding of data series representing the historical data. Our assumption was that there was a definite correlation between the past and future and we were going to exploit this bond. As our research continued we noticed that this was a naïve assumption and the assumed correlation did not exist at all times (Both the existence and degree of correlation was under question). Our system failed to predict the future price trend with an acceptable degree of accuracy.
Around 1997 I studied the work of Dr Andrew Colin and Doyne Farmer who is well known in wall street(Farmer started a company called the Prediction Company to hack the stock market, by beating the Odds on wall street). I soon realized that although stock prices move in a chaotic non-linear fashion but there was a HIDDEN ORDER which was far from random behavior and I could use this order to predict the next few days stock movements.
Any way to cut the long story short I managed to develop a new model which works fine and shifts the odds in traders side in all markets. I shorted BLM on the back of my model on 29thNOV (I think it was around 330 to under 290) with a tidy profit in a very short time. My trading model is not the best around but uses the state of the art technology (Neural Network lag_recurrent system) to model the stock behavior. A copy of the prediction was sent to Mr. SAQ_MUIR of share magazine as well as few of my close friends. (Saq suggested to introduce this concept in Share mag and it was published last week Issue 51 page 54. Please let me know if you wish to have the full report as well as the prediction model for BLM. Don’t forget to leave me you email.)

Any way If you are thinking of designing a future prediction model, let me know I can save you years of research work. (look at CRACKING WALL STREET to start you of).

I feel our small but intelligent community at SI is now ready to move on to topics such as

1) Does TA work for all stocks? Hmmmm (Not for Tech stocks IMHO, due to excess randomness within the data series)
2) Rigid and Adaptive indicators (God this is so important for any new technician)
3) Correct settings for indicators
4) Time frames (weekly daily)
5) Day trading
6) The most efficient stock selection criteria in choppy market

I am up to my neck in stocks now with some positions down at the moment but I will contribute weekends as much as I can.

If any of members feel there is any interest in Neural Network Modeling of stock behavior I will be more than please to open a thread on this subject

DOGS DON’T KNOW MATH YET CAN BE TRAINED TO PREDICTIVLEY CALCULATE THE PATH OF A FRISBEE AND CATCH IT PRECISLEY. (I have made an invalid assumption in here; any body knows what that is?)

See you all

PS My neural network model is called CHEETAH.

[This message has been edited by Dr Iraj (edited 02-01-2001).]

[This message has been edited by Dr Iraj (edited 02-01-2001).]
Happy New Year Dr Iraj

Are you assuming your dogs make a single predictive calculation, rather than a series of adjustments in mid flight, ie they are adaptive.

Where can we learn more about the subjects you list
Nice to see you on the BB again doc,
the more i learn about trading stocks the more i realise that sometimes the gut reaction is useful. Sometimes you just know its gonna be a bad day. TA is without a doubt the best tool we have to time our buying and selling and it still amazes me the amount of folks out there who buy shares without even looking at any kind of graph at all. They buy, cos the paper says its a buy.

Will anyone ever find that magical system? i don't think so. Will they reduce the odds of them loosing? yes if they use TA,FA, news, education. and a whole host of other things that i haven't even found myself yet.
But the most important thing for them to find is ENJOYMENT in doing the whole process. cos if you ain't got that you ain't got nothing. So why am i rabbling on! cos the best times i have had have been whilst on this (ci now si ) site.

thanks to all of you for giving me a hobbie that i'm really enjoying.

Dr. Iraj,

Nice to see you back and hear from you.

My model is relatively simple.

It contain certain parameters that need to be fullfilled at the same time, for the model to give a trading signal.
These parameters are quite strict, and, as a result, virtually never I get a trading signal.

In nearly 1 year, I got only 2 signals, the 2nd one being the sell on the Dow.

The model is also strict in terms of projected target.

The target the model gave me for the Dow, for example, was 10800ish. THIS DOES NOT MEAN I HAVE CLOSED MY SHORT ON THE DOW!

However, I am suspecting (and secretly hoping!!), that I have undervalued my model.

It is too early, but I will come back to this point.

So far, suffice it to say that my model has proved reliable twice out of 2 signals. Still too early to draw conclusions, I am afraid.

I wish you health and serenity. It may sound bizarre, but I treat technical analysis in a kind of "yoga" way.

It helps me to find the balance, on one side, while, on the other, I have learnt that, unless I feel mentally and phisically OK, there is no point in trading, since you lose money, at least myself!



Thank you for your offer of help, I will take the liberty of taking it as an open invitation and may come back to you in due course

The above does not represent financial advice.
Thanks safty. Lets hope the bloody market picks up. I am sick and tired of this hopeless market. I might bet in Casinos the odds are rather better. (Joking of course)


You are correct in your view. Neural network constantly correct themselves by eliminating the error in each output model. The dog example is the same. The dog constantly calculates and measures the error to make sure it catches the frisbee. I had assumed that there was no wind to change the direction of the frisbee (in stock's term one assumes there is no down grades, profit warning,etc) and also no fog to reduce the poor dog's vision (in stock term no low sentiment).

I wonder if the dog still would catch the frisbee if there was the wind and fog on its way?
My answer to this would be that it is easier for a dog to catch a freesbee than for us to make money on the markets.

The above does not represent financial advice.