Imaginary "Don't risk your money on myths"

Quote from Argent:

And as for you, your + only means that you can't think for yourself and have nothing ora-ginal to add.

Lol, that does sound like an arrogant prick :D

I have a very good network with pro traders. Most systems that make money (which execute on a account 1 mil +) these days use machine learning. Join some quant groups like http://stackoverflow.com/, http://www.quantconnection.com/forum and you will be surprised by where the world has moved on. Whatever strategy you have (strategy, not tactics or a bunch of rules) machine learning can make it much more productive.

Old traders like me whose primary edge is discretionary experience are being overtaken by bots :(
 
Quote from trader#21:

Lol, that does sound like an arrogant prick :D

I have a very good network with pro traders. Most systems that make money (which execute on a account 1 mil +) these days use machine learning. Join some quant groups like http://stackoverflow.com/, http://www.quantconnection.com/forum and you will be surprised by where the world has moved on. Whatever strategy you have (strategy, not tactics or a bunch of rules) machine learning can make it much more productive.

Old traders like me whose primary edge is discretionary experience are being overtaken by bots :(

Interesting. But, shit, if I had a mil in cash I sure as shit wouldn't be trading. I be on a warm beach summers with a Hurricane IV getting my second blowjob of the morning. I don't care what the brave new world is, it is perfectly oblivious that the same old shit still goes down every day in the index futures. How could it not? They exist to fleece US, and if they changed anything THEY would be competing with each other.
 
Quote from Argent:

I have never tried it, not being an Excel wizard. The thought of the sheet sizes required is daunting. Interesting that you need to measure noise, whatever that is.
Interesting how you put it: "noise, whatever that is" because that is precisely the problem (and the solution!) During my weekend forage of wikipedia for new insights, I ran across this definition:

"Noise is an unwanted perturbation to a wanted signal."

Brilliant! This is why we (the generic "we") can't agree on what noise is. In order to precisely measure noise, you must first decide what you mean by "signal", and that not only varies from context to context, it varies from trader to trader in the trading world. A trend trader and a RTM trader are NOT going to agree on what is signal, therefore they aren't going to agree on what is noise.

Note that the words "random", "measurement error" and other picayune distractions do not appear in the above definition. First you decide what is signal in your given context, then and only then can you determine if the nitpicking details of randomness, etc. are germane.

My initial goal is optimal noise reduction in my data analysis. I thought I was doing that with a common metric but it turns out that metric is entirely wrong for my particular signal. I hope this helps other traders as well.
 
Quote from kut2k2:

Interesting how you put it: "noise, whatever that is" because that is precisely the problem (and the solution!) During my weekend scourge of wikipedia for new insights, I ran across this definition:

"Noise is an unwanted perturbation to a wanted signal."

Brilliant! This is why we (the generic "we") can't agree on what noise is. In order to precisely measure noise, you must first decide what you mean by "signal", and that not only varies from context to context, it varies from trader to trader in the trading world. A trend trader and a RTM trader are NOT going to agree on what is signal, therefore they aren't going to agree on what is noise.

Note that the words "random", "measurement error" and other picayune distractions do not appear in the above definition. First you decide what is signal in your given context, then and only then can you determine if the nitpicking details of randomness, etc. are germane.

My initial goal is optimal noise reduction in my data analysis. I thought I was doing that with a common metric but it turns out that metric is entirely wrong for my particular signal. I hope this helps other traders as well.

I have a mite of a background in digital signal processing. The best engineering solution for measuring and predicting signal in a noisy environment is the Kalman filter. The theory recognizes two sources of noise: noise in the measurement process itself and noise in the signal being measured. T'ain't neither one in the market given a clean data stream. It's all signal. You can define and calculate various moments in the signal statistics, but don't cornfuse those with noise.
 
Quote from Argent:

I have a mite of a background in digital signal processing. The best engineering solution for measuring and predicting signal in a noisy environment is the Kalman filter. The theory recognizes two sources of noise: noise in the measurement process itself and noise in the signal being measured. T'ain't neither one in the market given a clean data stream. It's all signal. You can define and calculate various moments in the signal statistics, but don't cornfuse those with noise.
Looks like a disagreement over semantics here. To me, it's not all signal because it's not all tradeable. Price signal = tradeable movements; price noise = untradeable movements. Unless you've found a way to trade every single price movement, it ain't all signal.
 
Quote from kut2k2:

Looks like a disagreement over semantics here. To me, it's not all signal because it's not all tradeable. Price signal = tradeable movements; price noise = untradeable movements. Unless you've found a way to trade every single price movement, it ain't all signal.

So a price pattern is signal if it's tradeable and noise if it's not?

Jist made a bit 'a poke salat. Moughtn't be long far t'is wurld if'n ah dun it rong.
 
Quote from Argent:

So a price pattern is signal if it's tradeable and noise if it's not?

Jist made a bit 'a poke salat. Moughtn't be long far t'is wurld if'n ah dun it rong.
Always refer to the basic wiki definition of noise. Whenever price moves in a fashion that counters your method of trading, be it your desire to enter the market in a favorable manner or to exit in a favorable manner, then you've encountered noise. You thought you had a signal to enter or to exit and suddenly you find yourself moving in an undesirable direction. So the goal is find a method of data analysis that mitigates the noise as much as possible. The amount of residual noise remaining may still be undesirable, which means you wait for a stronger signal or you find a better (less "noisy") market to trade. At least that's my trading philosophy. YMMV.
 
Quote from kut2k2:

Always refer to the basic wiki definition of noise. Whenever price moves in a fashion that counters your method of trading, be it your desire to enter the market in a favorable manner or to exit in a favorable manner, then you've encountered noise. You thought you had a signal to enter or to exit and suddenly you find yourself moving in an undesirable direction. So the goal is find a method of data analysis that mitigates the noise as much as possible. The amount of residual noise remaining may still be undesirable, which means you wait for a stronger signal or you find a better (less "noisy") market to trade. At least that's my trading philosophy. YMMV.

Nivver seed 'at afore. (Ah'm a brushin' up own ma 'Plachan English far a trip home.) That is a totally bizarre notion to me. When I write algorithms there are only two conditions: the pattern is thar, or 't'ain't. It would never occur to me to define the absence of signal as noise. En fait, I am at a total loss to remember any time I have written a trade algorithm that used anything that one would remotely consider to be a "noise" measure. But whuddeye no? Jist a simple cuntry boy.
 
I like the definition: "Whatever the market does between your entry and exit is noise". If it's not noise I will add/reduce my position.
 
Back
Top