Machine learning in trading: theory, models, practice and algo-trading - page 2899
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
It was a brainstorming session, not the most successful, admittedly.
Weekends, holidays, Maksim Vladimirovich is in the bath, we can have a bit of a chat....
So to speak, one of the tasks of defining the state.
Exactly) It is necessary to formalise the determination of whether the price is in a trend or flat and from what moment it is there. If the algorithm cannot do this, then further search for levels is meaningless. But if it can do it well, then in essence it is almost half a grail).
Exactly) It is necessary to formalise the determination of whether the price is in a trend or flat and from what moment it is there. If the algorithm cannot do this, then further search for levels is meaningless. But if it can do it well, then in essence it is practically half a grail).
Well, then we need to formalise that there is a trend and a flat, then we can build boundaries between transitions.
but trend/flat is a subjective concept.
but trend/flat is a subjective perception.
Exactly. If we find an objective definition, it will be a philosopher's stone turning any TS into a grail.
You can ask a more meaningful question, how your guru defines trend and flat.
Exactly. If you find an objective definition, it would be...
One could ask a more meaningful question, how does your guru define trend and flat.
In our case, trend objectivity is probably just the majority recognising each of our trends as a trend.
Speaking of methods, a search of our website for the phrase "trend detection" yields over 600 pages of results. The first link in terms of relevance is a 12 year old article.
Any formal method will always give errors of the first and second kind (false definition and false omission of trends)
And the most important thing is that an algorithm should determine the trend without the possibility of manual intervention, otherwise there will be no possibility to test and optimise it.
If we talk about waves, they can be formalised through a zigzag, but there is an arbitrary choice of the zigzag parameter.
I have no idea
He has a video called "Why Techanalysis Doesn't Work". There he explains that indicators do not work because they are counted on history, not at this particular moment. A natural question arises - doesn't he calculate his trends and levels on history? I am not trying to undermine him in any way - I want to see if there is some deeper logic behind this external contradiction.
If we talk about waves, they can be formalised probably through a zigzag, but there immediately appears an arbitrariness of choice of the zigzag parameter.
Nah... it's better to use spetral analysis, take the smallest wave in the spetra and measure from it, or from the largest....
then there will be no parametres, only proportions of one relative to the other, at any TF, at any price area.
Doesn't he count his trends and levels on the history? I don't want to undermine him in any way - I just want to understand
He is not familiar with the scientific method and is not a programmer, so you should not expect clear formulations from him.
In classical volumes the whole theory is built from the distribution bell, but it is hardly fully formalisable. Let's assume that we have built a bell for the first 1.5 hours, allocated +-3sko. And then it is not clear. If flat, we should trade from these boundaries to the centre. But if there is a trend, we will be torn up. As if it is possible to put a condition, if the price went beyond 4 sko, then the trend, but I have little faith that this will work. The way they trade with 5pp stops on the video, I don't know if it can be formalised. The point is that taking into account the volatility, they look for trades to which the price can reach. They select only the best cases, so that there is a good sl/tp ratio, so that the trade is clearly distinguished, taking into account the trend, etc. (Seiden describes it well). At the same time, there will be a lot of limits that did not work (the price will not reach them).