Machine learning in trading: theory, models, practice and algo-trading - page 1527

New comment
 
Maxim Dmitrievsky:

I have to finish the topic this fall, otherwise it will be boring. A lot of time was spent on studying neural network add-ons, and then the theory of application, which nobody developed for financial markets at all. It, for some reason, is squeamishly avoided by datascientists.

Yes, we should finish it, of course.

I suspect that MO is up to the task, after all. However, people like Warlock are in no hurry to tell you everything here. And probably rightly so.

And those who can't do it, think that it's 50/50 all around and give up... All right.

 
Alexander_K:

and return to origin = 66% in two-dimensional wandering

Of course it will, when it collects all the stops ))))

 
It was interesting to take the main pairs and one system, and at least by optimization to find the days of parameters for all pairs, so that at least came to zero for the last year.
 
Alexander_K:

Why is the MoD failing in this task? A philosophical, conceptual question. I don't know the answer to it...

Again, the wrong direction a lot of people are digging.
Predicting and trying to achieve at least a 95% probability of SB is a futile exercise.
You have to look at the market from a different angle, not head-on.
There are precise mathematical tools that perfectly determine the necessary market characteristics (on history).
The main thing is to improve these mathematical tools to suit our needs.
That is, we should not try to predict the future, but to use the past and the present.

As for the options, nobody cancelled the delta hedge, between the options and the spot.
Or any other kind of arbitrage.

 
Maxim Dmitrievsky:

We need to finish the subject for this fall...

and at the end there will be only to defeat cerberus, as orpheus, psyche or hercules...)

 
Maxim Dmitrievsky:

But they all work with options, mostly. They think that it is proven that there is nothing to catch on the spot.

for options the strategy is different - you can buy several options of one instrument at different strikes (the furthest and near the current price? or with different dates? ....),

and it seems that the delta of these strings is traded, plus the analysis of the "smile"

i thought about how to check these option strategies on the spot, i couldn't think of anything similar, the principles are different: in the spot there is the current price to enter and exit according to our forecast, in options there are many strike prices where i can enter and exit when option is out of the money... these are different strategies



i need some software for options, maybe there is some truth in MO for options, but where to get historical data for options ?

 
Igor Makanu:

For options strategies are different - you can buy several options of one instrument at once at different strikes (the furthest one and near the current price like ? or with different dates ? .... ),

and it seems that the delta of these strings is traded, plus the analysis of the "smile"

i thought about how to check these option strategies on the spot, i couldn't think of anything similar, the principles are different: in the spot there is the current price to enter and exit according to our forecast, in options there are many strike prices where i can enter and exit when option is out of the money... these are different strategies



you need some software for options, maybe there is some truth in MO for options somewhere, but where to get the historical data for options ?

It's a little more complicated than that... it is forecasted using Black-Scholes equation or stochastic equations like Merton jump and MO is used to adjust parameters. I don't know exactly how it is traded, but if there is a quality forecast, I think, it is not difficult.

All you need for opts is a forecast by will

Here's the stuff from the paid subscription. For me it's a bit complicated, as well as for anyone, I think :)

 
In the same way coefficients for BP forecasting are selected, through stochastic equations. Fewer free terms, less overtraining, and the model is built on stochastic equations rather than on whatever
 
I do not know where to test these strategies:

It's a bit more complicated than that... it's forecasted using Black-Scholes equation or stochastic equations like Merton jump, and MO is used to adjust the parameters. I don't know exactly how it is traded, but if there is a quality forecast, I think, it is not difficult.

All you need for opts is a forecast by will

Here's the stuff from the paid subscription. For me, it's a little complicated, and for anyone, I think :)

Thanks, I'll read it tonight at work.

the overall problem with option trading, in runet 99.9% of articles "smartly" quote each other on the subject of Put and Call options, which takes 3/4 of an article))) - When I look at the analytical part of the article I don't knowwhere and how to test these strategies.

 
Igor Makanu:

Thanks, I'll read it at night. at work.

the problem with option trading itself, in runet 99.9% of articles "smartly" quote each other on what a Put and Call option is, which takes up 3/4 of the article ))) - I do not know where and how to test these strategies.

This is all for morons articles. Those who deal with science - they're usually full of mathematics and have almost no free access to it.

1...152015211522152315241525152615271528152915301531153215331534...3629
New comment