Using Neural Networks in Trading. - page 15
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
>> both statements are wrong!
Why are they wrong?
1. is the price uninterrupted in moments of illiquidity (news announcement and strong momentum)? It's not uninterrupted, it's just not there! It simply isn't there because the instrument is illiquid.
2. What do we care if it is even continuous? There is a well-known example of a Weierstrass function (sum of sinusoids) that is continuous at any point (seems even uniformly continuous on a segment; Quant, correct me if I'm wrong) and simultaneously not differentiable at any point.
Why are they wrong?
from experience.
the formation of a quote is discrete.
Quoted prices and volumes tend to bounce between two consecutive points in time.
consequently are not continuous values.
quoted prices and volumes tend to be subject to jumps between two consecutive points in time.
By the way, the highlighted is a very strong argument against price continuity. Philosophy: any continuity must be confirmed from other sides. "Continuity" of a price changing a hundred times in one minute and only five times in the next is too unstable.
The same argument also argues against the continuity of transaction time itself: the volume per unit astronomical time, i.e. the flow of transaction time, can jump at will.
That's why those who have worked with spectra have seen that on minute spectra there is a lot of noise, and the reason is the sampling noise.
I think sampling noise is not very significant compared to the market noise itself. And the higher the TF the less this noise.
from experience.
the formation of a quote is discrete.
Quoted prices and volumes tend to jump between two consecutive points in time.
consequently are not continuous values.
once again.
the price is continuous.
The price quoting process is discrete. It is the digitization of a continuous signal in radio engineering terms. Whether you like it or not, there is always a price. It is absurd to say that the price does not exist because the internet is down on Saturday or Sunday. The price is there. There is a demand for currency and there is supply. We just do not have a quote during that time = no price. The same applies to a gap. The following effect can occur when there is a gap, the rate of change in the process, is greater than the quote rate. That again coming back to DSP in its terms - Kotelnikov's theorem is not fulfilled.
Only LeeoV seems to have understood the essence of what I said. The higher the TF the less this noise appears...
once again.
the price is continuous.
the price quoting process is discrete. It is the digitisation of a continuous signal in radio engineering terms. Whether you like it or not, there is always a price. It is absurd to say that the price does not exist because the internet is down on Saturday or Sunday. The price is there. There is a demand for currency and there is supply. We just do not have a quote during that time = no price. The same applies to a gap. The following effect can occur when there is a gap, the rate of change in the process, is greater than the quote rate. That again coming back to DSP in its terms - Kotelnikov's theorem is not fulfilled.
Only LeeoV seems to have understood the essence of what I said. The higher the TF the less this noise appears...
When you can't buy or sell, but you have a large position to close, there is no liquidity and no price, because the uncertainty is high.
It is not the broker's fault that the liquidity is not there.
The situation when there is almost no supply and demand arises from time to time and is quite critical.
On the market side (up to the dealer, price formation) the price is not continuous.
what does the continuity property give you? are you going to derive some kind of formula?
once again.
the price is continuous.
the price quoting process is discrete...
there! finally I've made myself clear...
let's deal with the rest...
first of all, what we see on the screen is not a price at all! :)
it's an indicative quote, which the DC forms in a way it understands only... the only thing we can say that DT quotes are based on quotes that they receive from news agencies ... and most often indicative quotes of DT do not differ from the quotes of news agencies... but, of course, no brokerage company will disdain to shift this indicative quote by a few pips for their own purposes...
In short: Sometimes the price you need (our transaction price) coincides with an indicative quote, but sometimes it doesn't...
so in essence what you see on the screen is wrong... The price we need (the price of a possible deal) we will have to simulate with some degree of certainty... this is called reconstructing a signal from noise... (man, I don't like this DSP terminology, it's confusing)... of course, I didn't suggest the best way - just combine the prices of the minute bars... but never mind, it's not about interpolation methods...
Moreover, this restored signal will itself contain a noise component - i.e. some variable that we cannot specify analytically or predict with 100% accuracy...
secondly,
Prival wrote >>
The higher the TF, the less this noise appears
not at all... are you saying that you predict the price on large timeframes more accurately than on small ones? i.e. the variance of the random member there is less?
Thirdly, of course I haven't thought about illiquidity in the stock markets... and I didn't think of the Weierstrass function... but you're not going to approximate the price with a Weierstrass function!
And as for illiquidity: I'm sorry... choose less frequency of sampling (even if it is a week - broker will find a counteragent in a week ;))...
when we talk about price function (exactly the price at which we could execute a trade with our broker), we mean that we can execute the trade at that exact moment... normal brokers have in the Regulations this 'moment' is most often limited to 3 minutes (in Forex)... that is, according to the regulations, the dealer must offer a price in 3 minutes maximum ... if this does not happen, the request is rejected... then the adviser has to look again to see if the signal to open is saved...
Smart people gave up predicting prices back in the mid-90s. So if you're in the business of predicting price give it up quickly, it's a pointless exercise.
not at all... are you saying that you predict the price on large timeframes more accurately than on small ones?
Yes