Do you know how to make canals? - page 11

New comment
 
Aleksey Ivanov:

I.e. the point is to minimise the relaxation time of model parameter changes. + depot should not suffer during this time, for which(Maxim Dmitrievsky) switch (deversifier) strategies.

That's not how I understand it: if the model is stable, then it will surely come back on deviation. And this is a property of the model and it doesn't depend on the market - it's like a doll: you can sit through the drawdown. There are quantitative characteristics and they are known when designing the model. It is expressed as the sum of coefficients of the model. If it is more than 1, it is not stable, and the less than 1, the faster it returns after a shock (news) deviation.
 
СанСаныч Фоменко:
That's not how I understand it: if the model is stable, then it will definitely come back on deviation. And this is a property of the model and it doesn't depend on the market - it's like an unrollable doll: you can survive a drawdown. There are quantitative characteristics and they are known when designing the model. It is expressed as the sum of coefficients of the model. If it is more than 1, it is not stable, and the less than 1, the faster it returns after a shock (news) deviation.

So the strategy should sit the relaxation time of the model, i.e. the drawdown, while the parameters of the model are walking, should not exceed the given MM limits.

 
Alexander Laur:

you are confused with zero variance, or just a constant

and in fact the triangle example is an example of comparing the same time series with itself
 
Aleksey Ivanov:

I calculate a moving probability density and set the probability level - let's say 0.9 - and build a band where the price hits with that probability, which is the channel.


Yes, I forgot to specify, I was constructing those moving probability distributions as undistributed(moving averages constructing by 2n+1 points lag n points, the same is true for distributions, of course), for which just using the

GARCH forecasted a number of points and created a model of non-degenerate distribution at the end part of history (which is important), considering additional statistics provided by them. Question to SanSanych(SanSanych Fomenko): "Will this approach be more correct for jumps or will it also fail?"
 
Aleksey Ivanov:

Yes, I forgot to specify, that I've made these moving distributions of probability as undistributed(moving averages of 2n+1 points lag behind by n points, the same, of course, is true for distributions), for which just by GARCH model I predicted some points, creating a model of undistributed distribution on the final part of history (which is important), considering the additional statistics, already given by them. Question to SanSanych: "Will such approach be more correct for jumps or will it also fail?

I cannot assess your method and give an answer.

You are trying to consider an idea, of which there are countless on the market, but like an overwhelming number of idea authors, you do not ask yourself the question: on what basis will what you see in the historical data be repeated in the future? Or more precisely: does your idea even have a predictive ability?

The authors of GARCH didn't come to this model immediately, and incidentally, in a bitter struggle with the ideologists of the efficient market, which they understood as stationary.

We know from statistics that stationary processes can be predicted, but non-stationary ones are very poorly predicted. This is precisely the problem. Non-stationarity has rendered useless mountains of mathematics extremely effective in other areas.

GARCH ideology:

  • The underlying premise is NOT stationarity
  • we precisely formulate the meaning of the word non-stationarity
  • start to go from NOT to stationarity to stationarity bit by bit.
  • The closer the stationarity, the greater the ability to predict the future the algorithm has


Does your idea go this way?

 
СанСаныч Фоменко:

I cannot assess your method and give an answer.

You are trying to consider an idea, of which there are countless on the market, but like the vast majority of idea authors, you do not ask yourself the question: on what basis will what you see in the historical data be repeated in the future? Or more precisely: does your idea even have a predictive ability?

The authors of GARCH didn't come to this model immediately, and incidentally, in a bitter struggle with the ideologists of the efficient market, which they understood as stationary.

We know from statistics that stationary processes can be predicted, but non-stationary ones are very poorly predicted. This is precisely the problem. Non-stationarity has rendered useless mountains of mathematics extremely effective in other areas.

GARCH ideology:

  • The underlying premise is NOT stationarity
  • we precisely formulate the meaning of the word non-stationarity
  • start to go from NOT to stationarity to stationarity bit by bit.
  • The closer the stationarity, the greater the ability to predict the future the algorithm has


Does your idea go this way?

Thank you! There's a lot to think about. I was honoured to receive your advice. Happy holidays!
 
It is better to put the question another way: Do you know how to make deals in channels?
 

I have a graph presented as a folding metre, and each knee individually simplifies the analysis of the whole metre. Everything is complicated in its simplicity.

 
Aleksey Ivanov:
Thank you! There's a lot to think about. I was honoured to receive your advice. Happy holidays!
A pleasure to talk to you as well. Happy holidays!
 
Alexander_K2:
That's exactly right. I've been talking about it for two months on my thread about how to do it. And some of the most sophisticated people have absolutely no idea how to do it. They're clueless, to put it simply. They should be playing dominoes by now :))))

Wow - that's a picture of an owl!

1...456789101112
New comment