Machine learning in trading: theory, models, practice and algo-trading - page 1846

[Maxim Dmitrievsky]

Mihail Marchukajtes:

I have no history, I was building on VPS, but now I gave it up. It is one thing to assemble it, but it is another thing to make a reliable indicator to use it in the Expert Advisor. The quality increases by 20-30%. I judge by the number of selected inputs. If I do not have OM, then 100-120, but as soon as I add OM to the file, it will immediately select 150-180 significant entrances where OM is not in the last place. But I cannot use it automatically. I have an Expert Advisor that successfully saves it in a file. But the indicator that builds it looks ok, but when I feed it to my EA it gets shifted by one bar and I get full slippage. If I can re-do it I would be thankful.

I don't know if I can do that, but I don't know if I can do that.

[Mihail Marchukajtes]

Maxim Dmitrievsky:

And if you collect, there will still be skips... maybe there are sites with archives?

There will be no skips if you collect on the UPU server in the absence of connection interruptions. So it collects on ticks and is quite reliable. There were no problems with holes... The code itself is written there correctly enough.

Another thing is that for my TS I need at least 3 to 6 months to start using it at least somehow, but overall I'm very sorry that I had to temporarily exclude OM from the TS...

[Maxim Dmitrievsky]

Mihail Marchukajtes:
There will be no skips if you build on the UPU server when there are no connection failures. So it collects about ticks and is quite reliable. There were no problems with gaps... The code itself is written quite correctly.

Ok, I'll contact you in a year when I'll get it ))

[Roman]

Mihail Marchukajtes:
The problem is that MO needs exact numbers, while the foundation is half made up of abstract values. First it will be necessary to convert these values into numbers, and then feed them into the AI. Now, this conversion itself would be a Nobel prize so.... Translate Bernanke's speech into numbers (as an example) I would listen with interest.

Fundamental data has a lot of indicators that give numerical values.
Even here on the site, the news calendar gives statistics on events.
Yes, I agree, in speech statements, numerical values are missing.
For this reason, such data should probably be classified at 0 1.
The main thing is to teach to distinguish between positive speech, or negative ))
But here, this is also an idea, to think! ))

[Mihail Marchukajtes]

Maxim Dmitrievsky:

OK, in a year we'll talk when I get it))

I would not see the indicator if it works correctly when it is called from the EA. If I started to use it especially in real time it would be a problem, and I wish I could teach how to use it in a strategy tester. I've also stopped using it, because in the tester I'm doing the final selection of the resulting model. In addition to the basic selection of models obtained in the optimizer, there is also a final selection and it is exactly what is done in the tester. Of course, I've already bypassed this point through a text file, but still...

[Mihail Marchukajtes]

Roman:

Fundamental data has many indicators that give numerical values.
Even here on the site, the news calendar gives statistics on events.
Yes, I agree, in speech statements, numerical values are missing.
For this reason such data should probably be classified at 0 1.
The main thing is to teach how to distinguish between positive or negative speech ))
But here, this is also an idea, for thought! ))

I have a below called Fundamental Market Analysis. It contains the economic indicators of countries in charts. The unemployment rate, GDP, the refinancing rate and a number of others. I think that they can be used, even conditionally, since I trade on М15 and have a lot of signals, while the unemployment rate is published once a week or even more rarely. What should I do? Same question ....

[mytarmailS]

Maxim Dmitrievsky:

I don't know, I'm lazy.) I have a bot that trains normally, I gave screenshots of it... I'm not trading it... too lazy

So maybe you're not lazy, but the bot is useless?

[Maxim Dmitrievsky]

mytarmailS:

So maybe it's not laziness, but the bot is useless?

Not many deals, an average of 1-2 per day. After arbitrage is not interesting

[Mihail Marchukajtes]

Maxim Dmitrievsky:

few trades, an average of 1-2 per day. After arbitrage is not interesting

What's up, it looks just right to me....

[mytarmailS]

Maxim Dmitrievsky:

few trades, an average of 1-2 per day. After arbitrage is not interesting

what is the probability of profit? where is that screen?

What happened to arbitrage?