Machine learning in trading: theory, models, practice and algo-trading - page 3109

 
mytarmailS #:
Some tactic I don't know about.

Like the buy signal is long term, but the buy positions open and close often?
If so, why not just hold the buy

few trades, uninteresting, and the closing signal may not be as effective as if you close a trade earlier or reopen several trades.

profits grow faster if you use increasing lot size as equity grows.

it's a matter of preference, the model is the same.

 
Maxim Dmitrievsky #:

few trades, uninteresting, and the closing signal may not be as effective as if you close a trade earlier or reopen several trades

profits grow faster if you use increasing lots as equity grows

it's a matter of preference, the model is the same.

Can there be many long term models, or will they be highly correlated with each other?
 
mytarmailS #:
And can there be many long-term patterns, or will they be highly correlated with each other?

If the same signs are used, they are similar, but there is still some variation in the variants.

Many models are difficult to control afterwards.
 
Maxim Dmitrievsky #:

If the same features are used, they are similar, but there is still some variation in the variants.

a lot of models are hard to control afterwards.
I mean that the risk in this strategy is not controllable, or highly volatile at least.
It makes sense to run several non-correlated strategies at once, thus stabilising drawdowns/risk.
 
mytarmailS #:
My point is that the risk in this strategy is not controllable, or highly volatile at least.
It makes sense to run several non-correlated strategies at once, thus stabilising drawdowns/risk

I take the maximum loss series allowed on the history. If it is reached on the real, then to the trash or re-training.

If for each model to prescribe such conditions that it stops automatically, it is possible. So that you don't have to control it yourself.

But they will have to be reweighed by lot volume and performance. New tricks will start uncontrolled.
 
The main thing is that the TC earns 100%... even a little, even a little bit, even a little, but 100%...
Then you can do anything you want, any kind of candy.
 
Maxim Dmitrievsky #:

I tested the bot that I threw here on the 15th of May, a month has passed. With different sl and tp different results, but on average all on the growth.


Well, it works in this period, for a couple of years ago the test is not very good.

 
Maxim Dmitrievsky #:

I take the maximum loss series allowed on the history. If it is reached in real trading, then to the trash or re-training.

If for each model to prescribe such conditions that it stops automatically, it is possible. So that you don't have to control it yourself.

But they will have to be reweighed by lot volume and performance. New tricks will start uncontrolled.

Somehow, I'm thinking in the direction of the characteristics of the row.

 
mytarmailS #:
Takeout at 5 pips stop 50 ?

Judging by the green triangles there are not enough stops.

 
Valeriy Yastremskiy #:

Well this period works, for a couple of years back the test is not so good.

it's the same everywhere. It's for the Euro, not the New Zealander.