Machine learning in trading: theory, models, practice and algo-trading - page 1577

[[Deleted]]

Dmitry:

There can be many quality criteria. For synthetic it can be constant MO or variance.

If the average trade lasts 4-6 months, EXAMPLE, then the forward is 6 months. After that, the system can be over-optimized.

Fools who think that the model can "work" for a year or more without loss of quality will be punished by the market itself. It's only the geeks who test their Expert Advisors on a sample of 10 years... Well, also maxima....

There is a simple algorithm that does not require overoptimization at all. It is based on a breakout of the volatility of the past day (period).

Buy if:

((High[0]-Low[0] > High[1]-Low[1] ) & Close[0]>Open[0])

Sell:

((High[0]-Low[0] > High[1]-Low[1] ) &

Close[0]<Open[0]))

[Aleksey Mavrin]

Boris:

The problem is that we open an order today, but it is closed almost a year later (10-12 months)

and then we risk to find out right away if the "fundamental trend" is over (((

But it is not a matter of principle

on other TFs there may be different settings and we still don't know how to relate to them, if the settings are not some random, but work within a fairly large range of parameters

By the way, here I'm solving a similar problemhttps://www.mql5.com/ru/forum/329200#comment_14374536

If there are fewer trades, then the error will simply be larger. Assuming, of course, that it uses patterns that have not changed over time.

If you know your TS indicators (% of positive trades, MO), you can calculate for yours.

[Boris]

Another hard question for the venerable dons

for example, there are 10+ time series, coming out approximately from one point

how is it possible to trade on the divergence of the sum of series, if we do not know beforehand, which BP will be higher or lower than the others?

[Дмитрий]

Boris:

Another hard question for the venerable dons

For example, there are 10+ time series coming out of approximately the same point.

how is it possible to trade on divergence of the sum of series, if we do not know beforehand, which BP will be above or below the others?

Perhaps those above the average curve, we sell, and those below, we buy?

[Boris]

Dmitry:

Perhaps the ones above the average curve we sell, and the ones below that we buy?

Well, that's a possibility.

But how do I check it?

although, if the synthetic is only 2-3 or just an odd number, and even with the fact that they can cross over as they go, then apparently not an option

[Дмитрий]

Boris:

Well, that's an option.

but how do you check it?

Although, if the synthetic is only 2-3 or just an odd number, and even with the fact that they can cross as you go along, it's probably not an option

Check it with tests.

Even or odd makes no difference. Plot the average curve on the graph.

[Boris]

Dmitry:

Check with tests.

Even or odd, it makes no difference. Plot an average curve on the graph.

it doesn't work

The graphs cross and can do it many times

There is a situation when those above the average go down and vice versa

[Дмитрий]

Boris:

it doesn't work

The charts cross, and they can do it many times.

As you can see, it's easy to see that those above the average go down and vice versa.

So the ones above the average should go down. I take it you're trading the spread - it should converge to 0.

[Forester]

Dmitry:

So those above the average should go down. I take it you're trading the spread - it should converge to 0.

No one owes you anything. Go flub in the branch on spreads.

[Boris]

elibrarius:
No one owes you anything. Go flub in the spreads thread.

That's right, no one promised to converge to zero

The question was about how to trade with a mixture of divergent time series, and we don't know which one would be at the bottom or at the top.