Machine learning in trading: theory, models, practice and algo-trading - page 3666

[Forester]

Maxim Dmitrievsky #:
Marked by the duration of the trades. From one to 15 bars, with random selection.

With filtering clearly to the right and up, I need to learn more.

What period M bars?

Profit on traine 7 pts per trade, but OOS 0,7 pips per 1 trade on average. Profitability has fallen very much((
is falling as much? Like 10 times?

[Maxim Dmitrievsky]

Forester #:

Bars of what period M?

Profit on trayne 7 pts per trade, but OOS 0.7 pips per 1 trade on average. Profitability has dropped very badly((
is it dropping as much? Like 10 times?

The same one he's training on.

Well, it's one model, you have to learn in packages of 10.

I'll make a video.

[Maxim Dmitrievsky]

Forester #:

I'm not getting anything. Is there a link to the rootube itself?

It's being moderated.

[Forester]

Maxim Dmitrievsky #:

On moderation

You have described such markup in the article, everything is clear with it. The features on MAs are also clear.

But the markup on H1 is more interesting. In 15 hours the price can go to 2000 pts. You catch big enough movements, that's why the profitability per deal is good. But as I said deals are not enough in the end. Now I'll finish testing 25 pts, then 10 pts. And I will try 500 and 1000 more, maybe also at least up to 10 pts on average.

[Maxim Dmitrievsky]

Forester #:

You have described this markup in the article, it is all clear. The features on MAs are also clear.

But the markup on H1 is more interesting. In 15 hours the price can go to 2000 pts. You catch big enough movements, that's why the profitability per deal is good. But as I said deals are not enough in the end. Now I'll finish testing 25 pts, then 10 pts. And I will try 500 and 1000 more.

You can then just reduce the take and add trades. You'll get more trades.

And get signals on smaller TFs.

[Maxim Dmitrievsky]

Another approach in progress.

[Ivan Butko]

Maxim Dmitrievsky #:
Another approach in progress.

6000 trades and a penny on forward in plus(!) ?

Man, good one



I really hope you do similar tests with other pairs, as the euro has been a cheater for the last few years. Everything seems to work on it. Moreover, before 2012, it drains. And vice versa - if something does not work after 2012, before - probably works.
In general, a wider testing. To kill the scepticism. (wish)

[Maxim Dmitrievsky]

Ivan Butko #:

6000 trades and a penny on forward in plus(!) ?

Man, good one



I really hope you will do similar tests with other pairs, as the euro has been a cheater for the last few years. Everything seems to work on it. Moreover, before 2012, it drains. And vice versa - if something does not work after 2012, before - probably works.
In general, a wider testing. To kill the scepticism. (wish)

Yeah, I've looked at others too, they seem to give results too. I don't have enough hands to test everything yet.

[Maxim Dmitrievsky]

The clustering approach performs better than the model optimisation approach. Fact.

Maybe I'll drop the second topic.

[Forester]

Maxim Dmitrievsky #:

The clustering approach performs better than the model optimisation approach. Fact.

Maybe I'll drop the second topic.

With model training?