99percent
Member
...and that is all you need to know Grasshopper.
-
Please read our AutoTracker policy on the IFT deadline and remaining active. Thanks!
-
$ - Premium Service Content (Info) | AutoTracker Monthly Winners | Is Gmail et al, Blocking Our emails?
Find us on: Facebook & X | Posting Copyrighted Material
-
Join the TSP Talk AutoTracker: How to Get Started | Login | Main AutoTracker Page
-
The Forum works well on MOBILE devices without an app: Just go to: https://forum.tsptalk.com ...
Or you can now use TapaTalk again!
userque's Account talk
- Thread starter userque
- Start date
...and that is all you need to know Grasshopper.
userque
Well-known member
My Thoughts...
Tried all sorts of machine learning schemes (sometimes, to my detriment :embarrest
to trade the markets. I think I've finally satisfied most of my curiosities and am ready to 'settle down' and stick to my 'last, best, and final system.'
The only other 'thing' I can think of is to brute force every parameter and variable, of every machine learning algorithm I can. This will require flexibility only provided by a proper programming language. After much research, I've chosen Python (instead of R, etc.). It'll take a couple of months to get it up and running as I'm new to Python; then it'll take however long it does for the brute forcing to complete.
Some non-programming packages do something similar (Knime, RapidMiner, Weka, etc.), but not to the degree I'm talking about doing.
I'm out of shortcuts.
Here's what I believe:
Market price action is composed of both randomness and patterns. Pseudorandomness. This is the same sort of randomness generated by Vegas slot machines and computers in general. The patterns are created by repeating groups of random price actions. Simplified:
[random group of price actions][different random group of price actions][third random group of price actions][...etc...][repeated group of random price actions]
I believe ML algorithms get choked up trying to make sense of the random price actions. Just as a ML algorithm will try to predict lottery numbers...but fail. The signal to noise ratio is too low for ML algos.?
My final system doesn't try to make a mathematical model of price action--as traditional ML does, it only tries to detect patterns.
Tried all sorts of machine learning schemes (sometimes, to my detriment :embarrest
to trade the markets. I think I've finally satisfied most of my curiosities and am ready to 'settle down' and stick to my 'last, best, and final system.'The only other 'thing' I can think of is to brute force every parameter and variable, of every machine learning algorithm I can. This will require flexibility only provided by a proper programming language. After much research, I've chosen Python (instead of R, etc.). It'll take a couple of months to get it up and running as I'm new to Python; then it'll take however long it does for the brute forcing to complete.
Some non-programming packages do something similar (Knime, RapidMiner, Weka, etc.), but not to the degree I'm talking about doing.
I'm out of shortcuts.
Here's what I believe:
Market price action is composed of both randomness and patterns. Pseudorandomness. This is the same sort of randomness generated by Vegas slot machines and computers in general. The patterns are created by repeating groups of random price actions. Simplified:
[random group of price actions][different random group of price actions][third random group of price actions][...etc...][repeated group of random price actions]
I believe ML algorithms get choked up trying to make sense of the random price actions. Just as a ML algorithm will try to predict lottery numbers...but fail. The signal to noise ratio is too low for ML algos.?
My final system doesn't try to make a mathematical model of price action--as traditional ML does, it only tries to detect patterns.
clester
Well-known member
My Thoughts...
Tried all sorts of machine learning schemes (sometimes, to my detriment :embarrest
The only other 'thing' I can think of is to brute force every parameter and variable, of every machine learning algorithm I can. This will require flexibility only provided by a proper programming language. After much research, I've chosen Python (instead of R, etc.). It'll take a couple of months to get it up and running as I'm new to Python; then it'll take however long it does for the brute forcing to complete.
Some non-programming packages do something similar (Knime, RapidMiner, Weka, etc.), but not to the degree I'm talking about doing.
I'm out of shortcuts.
Here's what I believe:
Market price action is composed of both randomness and patterns. Pseudorandomness. This is the same sort of randomness generated by Vegas slot machines and computers in general. The patterns are created by repeating groups of random price actions. Simplified:
[random group of price actions][different random group of price actions][third random group of price actions][...etc...][repeated group of random price actions]
I believe ML algorithms get choked up trying to make sense of the random price actions. Just as a ML algorithm will try to predict lottery numbers...but fail. The signal to noise ratio is too low for ML algos.?
My final system doesn't try to make a mathematical model of price action--as traditional ML does, it only tries to detect patterns.
Just stumbled on your thread and saw you talking about machine learning. I have gotten interested in that lately and so you got my attention. I'm taking a class on ML on coursera from Stanford. I'm just near the beginning but seem interesting. They use matlab/octave for programming. Lots of matrix algebra so far mixed with a little programing at this point.
What is your background in ML? or AI ? any suggestions where I can go to learn more? I have programming background (C.S. degree years ago) but not professionally. I have done a little in Python. Have you ever used Clojure?
userque
Well-known member
Excellent! Keep us posted.
I've had some college towards EE. I work with electronics. My hobbies include ML and AI.
I learned more by doing a lot of reading of academic papers. You can restrict google to .edu domains to find them more easily.
EDIT: What exactly are you interested in learning more about? I may be able to point you to something I've already found.
Haven't used Clojure. So far, including stuff I've forgotten and only know a little about :blink:: BASIC, Z-80 Assembler, MS-DOS, FORTRAN, QBASIC, ABASIC?, C++, Visual BASIC, VBA, AutoIt, Sikuli.
Python is not as fast--execution-wise, but it seems to be the way to go as far as versatility and development speed, etc.
userque
Well-known member
k0nkuzh0n
Well-known member
Matlab is very popular and has a ton of libraries/toolkits and a lot of support. Not the most user friendly environment since there's a lot of parameter passing and documentation usually has a lot of short hand.
Im an EE who did a CE track. Most of my coursework was in C, but I've also used matlab, mathcad, assembly and, C++. and I took a SaaS class (ruby on rails) and Android applications class on Coursera/Edx. I use LabVIEW for a living...but I don't think I'd ever recommend labview for datacrunching; it's also got a lot of libraries and it's far more user friendly than matlab, but I'd never consider it efficient relative to other languages, and it's pretty terrible at digital logic/algebraic equations.
userque
Well-known member
Yeah...considered Matlab too. As you suggest, it has a steep learning curve. That's fine, except that it can't double as a general purpose language like Python can. Not to mention it is not Open Source. If I'm going to spend a lot of time getting good at a language, I want it to be able to do a lot...without a lot of effort...relatively speaking.
I've played with Android development...nothing serious.
Do a little research into Python and see what you think...including the Machine Learning and Numbers, etc. packages.
I've played with Android development...nothing serious.
Do a little research into Python and see what you think...including the Machine Learning and Numbers, etc. packages.
userque
Well-known member
userque
Well-known member
userque
Well-known member
userque
Well-known member
Just a little something I've been wanting to mention for months now:
I always follow a system and sell at a loss if the system calls for it. We have to know that every trade won't be a winner. So why sweat a loss? Some become long-term investors...holding on to a losing trade. Others 'double down' at 'better' prices. Developing and FOLLOWING a system that one is comfortable with really helps.
If I ever get felted, it'll be because I traded poorly--until the bitter end; not because I became an investor to avoid selling at a loss, and ultimately, watched a stock fall until it was delisted, etc.
One can hold on to a losing trade for a year to finally and proudly exit at a profit. However, I believe it would have been more profitable to take the loss a year ago and trade for that whole year rather than leaving money tied up in a losing trade. This concepts escapes a lot of folks, strangely.
All of this, of course, assumes the use of a system that works. One shouldn't be trading unless their system has shown to be a working one--either through paper trading or proper backtesting.
Finally, regarding 'doubling' down:
Martingale Systems and Why They Dont Work In Stock Trading
If a system gives another buy signal after the initial trade lost ground, is it then ok to 'double' down?
Depends upon your money management scheme. One would be 'doubling' their exposure and risk...so...maybe.
...IMO
Good luck during these rough waters!
______________________
Disclaimer: I am far from an expert!:blink:
I always follow a system and sell at a loss if the system calls for it. We have to know that every trade won't be a winner. So why sweat a loss? Some become long-term investors...holding on to a losing trade. Others 'double down' at 'better' prices. Developing and FOLLOWING a system that one is comfortable with really helps.
If I ever get felted, it'll be because I traded poorly--until the bitter end; not because I became an investor to avoid selling at a loss, and ultimately, watched a stock fall until it was delisted, etc.
One can hold on to a losing trade for a year to finally and proudly exit at a profit. However, I believe it would have been more profitable to take the loss a year ago and trade for that whole year rather than leaving money tied up in a losing trade. This concepts escapes a lot of folks, strangely.
All of this, of course, assumes the use of a system that works. One shouldn't be trading unless their system has shown to be a working one--either through paper trading or proper backtesting.
Finally, regarding 'doubling' down:
Martingale Systems and Why They Dont Work In Stock Trading
If a system gives another buy signal after the initial trade lost ground, is it then ok to 'double' down?
Depends upon your money management scheme. One would be 'doubling' their exposure and risk...so...maybe.
...IMO
Good luck during these rough waters!
______________________
Disclaimer: I am far from an expert!:blink:
Nice post on trading discipline.
That's good. It's a poker term to me - but is it also used for other endeavors where you lose "the lot?" :1244:
That's good. It's a poker term to me - but is it also used for other endeavors where you lose "the lot?" :1244:
userque
Well-known member
Nice post on trading discipline.
Thanks...from one poker player to another:smile:
userque
Well-known member
It was slang started by Phil Laak (allegedly). So I feel safe with repurposing the slang term. But the Urban Dictionary does allow for my use in its fourth definition:
FireWeatherMet
Well-known member
Nice read.
I guess I kind of follow that idea of not being against selling at a loss in that not every move is positive.
It comes from the old adage of "if conditions that prompted your trade have changed, its OK to change your trade".
It is true in a range-bound channel we've been in, or times of high volatility if your trade went wrong, you could give it a few days to 2-3 weeks to get positive again, and this revolves around or IFT limitations.
Did exactly this yesterday. Went in at a short term bottom a few days ago looking for a 1% pop. Well, would have gotten it except I went more than half into the I which dragged things down, but at the top of the channel we started reversing again and I sold yesterday, making for exactly 0% on my move. But I wanted to avoid todays fed notes volatility and saw no mechanism to take us out of the trade-bound channel, and still see more room to downside than upside .
Its nice to know there are others who share the "end the trade" not out of panic or fear, but just thru tactical and strategic reasons.
userque
Well-known member
Exactly. And nice trade!
I originally had non-TSP trades in mind, but you're right--we have to allow more leeway with TSP trades.
Similar threads
- Article
- Article
