Machine learning in trading: theory, models, practice and algo-trading - page 3362

New comment
 

On the subject of OOS interpretations.

On the right OOS is excellent, on the left - very bad. How to evaluate such "robustness" is unclear. 6K non-overlapping positions.

 
fxsaber #:

On the subject of OOS interpretations.

On the right OOS is excellent, on the left - very bad. How to evaluate such "robustness" is unclear. 6K non-overlapping positions.

What do you mean "non-overlapping positions"?

That's just random ness. It is also random to have a result where the result before Sample is the same as the result after Sample in this case, and vice versa.

 
Anatoli Kazharski #:

What do you mean by "non-overlapping positions"?

At any given moment, the number of positions is no more than one.

 
fxsaber #:

At any given time, the number of positions is no more than one.

ahahahahahahahahaha

 
fxsaber #:

On the subject of OOS interpretations.

On the right OOS is excellent, on the left - very bad. How to evaluate such "robustness" is unclear. 6K non-overlapping positions.

Well, on the left the connection of indicators to trades remained, but drifted. The trend has changed or whatever. It's a conditionally good TS, if the connection in the future will be the same as on the right. But you should not bet much on it, I guess.

I've been getting similar curves on the eurobucks for the last month. The front is good, the back not so much.
 
fxsaber #:

On the subject of OOS interpretations.

On the right OOS is excellent, on the left - very bad. How to evaluate such "robustness" is unclear. 6K non-overlapping positions.

1) it is necessary to evaluate such robustness badly.
In fact, on the graph we clearly see three different states of the system, while we should see one.

2) do you have optimisation parameters?
How many are there and what are they
 
mytarmailS #:
1) it is necessary to evaluate such a slavery badly.

In fact, on the graph we clearly see three different states of the system, while we should see one.

I don't see TC states. It's just a backtest balance graph of the best set after optimisation.

2) do you have optimisation parameters?
How many are there and what are they

Five pieces.

 

Maxim Dmitrievsky #:
Ну слева связь индикаторов к сделкам осталась, но удрейфоввла. Тренд изменился или что там. Условно годная ТС, если связь в будущем будет такой же как справа. Но на нее много не надо ставить, наверное.

That's how you can explain anything. Obviously, there's nothing concrete to say. I myself think it's an adjustment. Because the beginning of the "drift" to the left coincides very much with the Sample start point. In such a situation OOS can be explained, of course.

Forum on trading, automated trading systems and testing trading strategies

Machine Learning in Trading: Theory, Models, Practice and Algorithm Trading

Anatoli Kazharski, 2024.01.03 18:52

It's just random ness. You can also get a result by chance when there is a result before Sample, as in this case after Sample, and vice versa.


I've been getting similar curves on my eurobucks for the last month. The front is good, the back is not so good.

This is also EURUSD. The OOS on the right is the last four months of 2023. Optimisation - the rest of 2023.

 
fxsaber #:

I don't see any TC states. It's just a backtest balance graph of the best set after optimisation.

in fact it is states, there are three of them normal trade, very good and just good

fxsaber #:

Five pieces.

are the values constants or dynamic?

 
mytarmailS #:

It's basically states, there's three of them normal trade, very good and just good.

I knew it was simple.

are the values constant or dynamic?

Answered this question.

1...335533563357335833593360336133623363336433653366336733683369...3583
New comment