Machine learning in trading: theory, models, practice and algo-trading - page 1091

 
Maxim Dmitrievsky:

I do nothing, read books, educate myself, go for a walk with the girls :)

It all works, I showed you a screenshot, but I can do better, but I'm too lazy to do it.

I don't trade yet because I'm too lazy... but I read books.

i meant the screenshot where you wrote "whatever works" or it was something new, and i missed it?

 
Farkhat Guzairov:
That's what I advised everyone, because it's a conversation for the sake of conversation and nothing more.
Farhat, at the first stage the TS chooses the optimal model, and then the NS makes a forecast after training?
 
Maxim Dmitrievsky:

Yes, that one... can still be applied to all sorts of different strategies, because that screen was really from randoms without any understanding of the strategy itself

So it's a blueprint that can be applied to all sorts of strategies, that's what we'll have to work on later.

+ try other methods of accounting for arguments

That's great, get it done soon, don't be lazy.

I had a breakthrough in my forecasting yesterday...

So maybe I have to reconsider my opinions about BP not working in the market... It turns out that I (and 99.9% of steelmen) have wrong understanding of the forecast, even in forecasting packages everything is implemented exactly wrong. It's funny, I spent a lot of years to understand that BP doesn't work, and here it is...

a forecast is a sum of sub-forecasts that are ranked and summed or just added up, and you don't have to do that, the future has many options, it's a simple but key phrase, you should not summarize, but choose the option that fits the current situation

 
Yuriy Asaulenko:

You should tell us exactly who the local Gurus are. Because since 3:34 p.m. everyone has been waiting tensely.

So it's YAJAJA!!!! What's the answer to that? Repeat it...

 
Mihail Marchukajtes:

So it's YaYaYaYaYa!!!! What's the answer to that??? Repeat it...

Can econometrics distinguish a trend from a flat?
 
Renat Akhtyamov:
Farhat, at the first stage the TS selects the optimal model, and then the NS builds the forecast after training?
In fact, I don't have TS as such, I have statistics, it is formed according to certain rules (like you can call it TS, but I wouldn't). After that the statistics is normalized for NS and NS training is carried out, the result is the probability expressed in %, then we need a second level NS for decision making, but for now without it. I want to reassure everyone at once, it's just a trained NS, unfortunately we don't have enough time or knowledge (or rather understanding, what to do next) to bring it to AI. I will say that there is a result, the only problem is what risks will be optimal to achieve the profit.
 
Of course you must understand that the NS or AI will never give a 100% true answer, it can only be a probability of the outcome as no one knows the future.
 

How realistic is it to teach a machine to distinguish between GBPUSD and EURUSD?

Where do I dig? Read? Recognize? Classify? Or?

 
Igor Makanu:

How realistic is it to teach a machine to distinguish between GBPUSD and EURUSD?

Where do I dig? Read? Recognize? Classify? Or?

What for? I can't even tell the difference myself.) Well, there are different ratios and so on. But price changes are traded, the price itself has no great importance.

 
Yuriy Asaulenko:

Why? I can't tell the difference myself.) No, well, there are different ratios, etc. But it's the price changes that are traded, the price itself has no great importance.

Well, if I'm going to write in detail that I want to distinguish between 2 types of bars on the chart, I guess they are called impulse bars? I sure as hell won't get any answers

:)

but the problem of distinguishing between two charts and two types of bars is essentially the same

I've just got interested in AI, read a little bit, watched a lot on YouTube, now I need some practice - the task is not for theory, but for practical training of AI - I can solve this task by means of OOP, but I cannot even formalize it in terms of AI - what to use? scaffolding? classification? just NS? ....

WHAT?