Comments on: A Top Down and Bottoms Up Approach to Neural Net Models

By: Tom

Tom — Mon, 10 Mar 2008 10:30:32 +0000

dc: I’m glad you like the linear regression slope, it works nicely. Did you download the macro for it?

By: dc

dc — Sun, 09 Mar 2008 09:26:50 +0000

i forgot to say the reason for this topic origanally was i read that you wanted a good indicator to go with your model and you did mention slope regression once ,and i do like the approh of any regression model or stochastic model in the mix.(RECEPIE)

By: dc

dc — Sun, 09 Mar 2008 09:15:16 +0000

hi tom
I guess it’s sort of a labor of love looking to see if you can build something as elusive
and some people say impossible to accomplish. I’m Always trying to find the automated
(hands free) system that will buy/ sell possitions on criteria that you program into it,
without you being there in person. Just one problem noise ,noise ,noise. Pulling back and away from short time frames into longer time frame might be the answer but not necessarily. Since you yourself dabble in time price prediction i was thinking you might have that missing ingrediant for a recepie to success. Otherwise i am forced to go all out and go fuzzy logic based on possotive volume/negative volume momentum combo’s ,also accumilation/distrabution fades with trends.there are 20 different approches the only problem i have is GREMLINITESS.Just when you spot the holy grail in this mist it vanishes like a mirrage.IT CAN GET DISSAPIONTING. oh well

SUCK IT UP AND MARCH ON SOLDIER, RIGHT.

By: Tom

Tom — Sat, 08 Mar 2008 12:05:21 +0000

DC: It sounds like you have a system tailored for you! Are you trading it?

By: dc

dc — Fri, 07 Mar 2008 10:45:11 +0000

hi tom
i was wondering if you build automated trading programs(stratagies).
The reason i”m asking is i been at it for about 2 years and i think i found the elusive holy grail(but don’t tell anyone).just teasing
The problem is american platforms “suck” as if they are made for you to fail before you even start ,It almost seems like you are being hacked by someone trying to screw with you.HMMM(I think i should give them a piece of my mind and talk down to them as to how it”s completely unethical,and i”m sure even Bush would call these people EVIL DOERS,> don’t mind me i’m just venting)

But any how,i found out to my dissapiontment that most indicators won”t work all the time do to different market conditions.One beiing trends and the other is chop.In the trends your are making money in the chop you lose it right back

so if you”re interested let me give you some indicators that i have tweeked throughout my quest see what you think

FOR SCALPING & GENERAL PURPOSE.
1) linear regression(7) crossabove hull (10) to buy,crossbelow to sell.
At 180 minutes in trends it usually comes in at 2.3+ profit 75% of the time but then you have to be out of the market the other 25% not to lose your gains.though that might be controlled by using some other indicator that will keep you out of the chop.
So see how robust you can make it, if you can, it could become your favorite.

FOR TRENDS
2) stochastic d(18) k(3) smooth(1) and you overlay on top another stochastics d(3) k(14)smooth(1) I know it sounds a little strange and you may have to blend everything out with the background except the d(18) on the first one that becomes your slow and k(14) becomes your fast everything else should be blended out into the back ground (to blend, make unessasary lines white if the back ground is white ,black if it”s black,etc,etc,etc)
As i said before it’s only for trends and it might be a good screener for stocks at
3-6 hour time frame I think it works a little better then MACD.Though MACD (10,26,5) could be used as your exit scale out stratagy against this stochastic trend.
HOPE YOU ENJOY THESE 2 INDICATORS (SORT OF LIKE FOOD FOR THOUGHT)

I’m just getting a little fed up with these EVIL DOER GREMLINS! see if you have better luck.