How do you measure noise? - page 2

New comment
 
Noise or no noise... the only stupid radio operators left on the subject. And all the stupid electricians were electrocuted. Only I survived.
 
Алексей Тарабанов:
Noise or no noise... the only stupid radio operators left on the subject. And all the stupid electricians were electrocuted. Only I survived.
What's your vaccine?
 

Alexey Burnakov:

My view. If you're not looking for any model properties, the whole process is noise to you. Nothing is predicted. If you apply linear methods, noise is a residual of the linear model. If you apply non-linear methods, noise is a residual of that model. Tomorrow you will have access to an insider. The noise will still decrease. The day after tomorrow you will become a Forex god and you will know all the plans of the participants and their implementations and the noise will become zero.

The question is simple. Answer: Noise is something you are unable to explain.

In general, yes. But here's Hegel's - randomness is an unknowable regularity works up to certain limits, so even in this case noise will remain - the same random wanderings and drift.

Alexey Burnakov:
Another, statistical formulation of the question. There is some model that leaves noise on the data of the whole population, i.e. the whole history of existence of quotes and the future of it. You approximate that model and your noise is an estimate of the ideal noise of the ideal model. That's how abstract it is. In classical methods it is assumed that the noise of the ideal model is normal.

This model is not applicable to the market. The market in 2008 has significantly changed by 2011, then by 13. Now it is different again. The successful models of 2008 did not work in 2011, and these, in turn, do not work now.

The market is definitely not normal. The whole population will not help us in any way, because they are different systems in time. And we need, rather, short-term models. Models based on a limited sample. Well, linear models are not so bad for such samples.


The red one is the standard EMA(12), for comparison, and the blue one is one of the signal extraction attempts to calculate noise. Not rebuildable (for the skeptics), although I don't see anything wrong with it.

 

О! An intelligent person will always distinguish Gogol from Hegel, Hegel from Bebel, Bebel from Babel, Babel from cable, and cable from dog.

 
Event:
What's the vaccine?
Wheat
 
Maxim Romanov:

......However a little noise is indeed present there, it is quantisation noise, each transaction is made with a volume of finite precision, because of this noise does occur.....

Quantization noise occurs when the original analogue signal is sampled. Market movements are inherently discrete in nature - time/price. Quotes are given as: Tick(n)/Price=X*Pips, Tick(n+1)/Z*Pips. The price change is fixed in pips. There is no such a thing as a price that differs from the neighboring ticks by 0.03 or 0.97 Pips between the previous and the next tick during the quoting process. That's when quantisation noise would really occur.
 
Yuriy Asaulenko:

The market in 2008 had changed significantly by 2011, then by 13. It is now different again. The quite successful models of 2008 did not work in 2011, and these, in turn, do not work now.

How the stock market is changing:

 
Алексей Тарабанов:
Wheeny
Shouldn't we take a swing at our... William sh?
 
Yuriy Asaulenko:

..... This model is not applicable to the market. The market in 2008 changed significantly by 2011, then by 13. It is now different again. The quite successful models of 2008 no longer worked in 2011, and these, in turn, do not work now....

The market doesn't change, there may just be different market situations. If a model that was successful in 2008 doesn't work in 2011, it means the model was ad hoc and not based on the key properties of the market.

 
lilita bogachkova:

How the stock market is changing:

Yes, noise is increasing. On some instruments the noise is already a little less than signal. By the way, it is quite easy to work in such a market. Buy at the bottom of the noise, sell at the top and vice versa. The risks are low. They were. :) Now this topic is also closed by HFTs, who have computers in the hardware halls of the exchange.
123456789...11
New comment