Why is the normal distribution not normal? - page 8
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
On the other hand, the s.c.o. estimate for a strongly fat-tailed distribution depends on the sample size, so it's not so simple here.
Absolutely right!
I got a formula which fits Gauss into the obtained distribution by MOC and faced with the fact that Gauss does not fit in a single way, but in an infinite way... Indeed, if we take the same distribution, e.g. exponential distribution, then on a logarithmetic scale we obtain a triangular-shaped distribution. We take 10 points near the centre of distribution and inscribe a Gaussian there (on a logarithmetic scale it is a parabola, because if y=exp(-a*x^2) then ln(y)=-a*x^2). We wrote it down and got the variance. We take the same distribution, but now we have 20 points instead of 10, as in the previous case. After adding the points, we obtain another parabola, i.e. another variance... and so on and so forth to infinity. It turns out that the concept of Gaussian dispersion is not defined for a non-normal distribution, so it is incorrect to estimate the distribution dispersion by adding a Gaussian one to it.
begemot61 wrote >>
How can you use first difference statistics without having an MO price?
You know the rate distribution. What's next? How does it relate to the price distribution?
There is no contradiction here. After all, when we trade a particular instrument, we are trading the PRICE, not the price itself... Understand? - Having opened a position, we are waiting for the price to change by a certain value, determined by the trading algorithm used, and upon reaching necessary increment - we close the position. In other words, we trade in terms of price increments, not its absolute values. For this reason, the analysis of the series of the first difference (FFD) from TP is very informative, it is in fact an analysis of possible transactions (if FFD is taken from the TF where TS trades).
Svinozavr wrote(a) >>
Generally speaking, putting price forecasting at the heart of TS is a road to nowhere. You can model/forecast the plots between the actions of a major. But any of their - the big guy's - actions will kill all the predictions. It's not the past time series that determines most of the strong moves in the market. All attempts (yours, others) have shown this well.
Why is the common perception that TA does not work? Because they try to predict the price. And the TA tools just capture a particular state of the market. They do not predict ANYTHING. That is not what they are for. A ruler cannot predict length. And a weight is not a weight. All this is analytics. And they are VERY valuable.
The prognostic properties of TA may be used only locally for auxiliary, short-term actions such as entries/exits/reversals according to the logic of the market mood, determined by the analysis. I say 'mood' as the term 'context' already bores me.
I disagree with you there.
I can cite as much evidence as I like for the existence of statistically significant dependencies between price increments. In other words, I claim that there is a connection between the previous price movement of an instrument and its movement on the next step. Another thing is that the profitability of TS based on exploitation of these dependences almost never exceeds the brokerage company commission... But, that's not the point, the point is that you are saying essentially that there is no relation between the past and the future in price VR, and strictly speaking, it's not so.
On the other hand, you are right that we have no reason to pay attention to something that cannot be reliably exploited. However, who is to say that there are no deeper dependencies in price BPs? There is plenty of room for research and I wouldn't put a big point on TA.
pps Avals wonder what you get if you subtract one from the other? It seems to me to be a wavelet-like attenuation.
Here's the difference (realdan-NR):
1999-2001 (no more than m15 in Excel)
Incidentally, there was a significant down-trend in this area. The weekly chart for this period:
If you look at the difference between the positive and negative increments, you get:
The down-trend was formed due to a large number of small negative increments, but the counter-trend was leading in number in the area of large increments (more than 13p - about 2co)
the difference between positive and negative increments
How is this?
How's that?
By frequency: (number of increments=1) - (number of increments=-1) etc.
I can cite as much evidence as I like for the existence of statistically significant dependencies between price increments. In other words, I argue that there is a relationship between the previous price movement of an instrument and its movement in the next step. Another thing is that the profitability of TS based on exploitation of these underlying dependences almost never exceeds the brokerage company commission... But, that's not the point, the point is that you are essentially saying there is no relation between the past and the future in price VR, and strictly speaking, it's not true.
I'm not claiming it isn't. Read (highlighted and even underlined):
Youcan model/forecast the plots between the biggies' actions. But any action by them - the biggies - will kill all the forecasts. It's not the past time series that determines most of the strong moves in the market. All attempts (yours, others) have shown this well.
On the other hand, you're right that we have no reason to pay attention to something that can't be reliably profited from. However, who's to say there aren't deeper dependencies in price BPs? There is plenty of room for research and I wouldn't put a fat point on TA.
Who's making a point? I'm just saying TA works. Most of them use it inappropriately and don't even understand what these indices show.
Most people put the cart before the locomotive, i.e. the forecast before the analysis (context definition, damn it!). They try to forecast prices by TA and build their TS on it. This is an inevitable failure.
The TA should be used where it works. It works in the analysis of the current state and short-term forecasting within this um... context.
I wonder how many cacti have been chewed up?
If you sit on the bank of a river chewing cacti for a long time, you can see the corpses of your colleagues who started chewing them earlier floating down it.
Avals писал(а) >>
By frequency: (number of increments=1) - (number of increments=-1) etc.
Got it. It's cool how it works in five digits:
I wonder how many cacti have been chewed up?
If you sit on the river bank chewing cacti for a long time, you can see the corpses of your colleagues who started chewing them earlier floating down it.
And here's one from a neighbouring thread:
You listen, you listen, you listen, you listen, you listen, you wait, you wait.
TA should be used where it works. And it works in analysing the current state and short-term forecasting within that um... context.
I agree with this one! I am working in that direction myself at the moment.