A great book on testing and optimisation - page 21

New comment
 
FOXXXi >> :

... Not a single losing trade in the last year.

Out of how many closed?

 
goldtrader >> :

>> Of how many closed?

>> 127, but of course there is no guarantee that this will continue.

 
FOXXXi писал(а) >>

I have not dealt with the concept of risk, but with real losses, and that's OK, they are natural, just like profits with a high statistical advantage. Not a single losing trade in the last year.

Now you're going to hang on to the word "book" and make a big deal out of it.

You're not going to get it from you in short.

1. default risk and increased credit risk on one of the legs.

2. a general increase in rates.

3. higher direct transaction costs. higher market sensitivity to own orders/ lower liquidity.

4. correlation risk. total loss of return-to-average property.

there are others, but they are unlikely to be of interest anymore, the main ones are listed.

I think you are unlikely to have done a "forward" assessment of the above risks.

 
Quant >> :

in short, you're not going to get it from you, player.

1. default risk and increased credit risk on one of the legs.

2. a general increase in rates.

3. increased direct transaction costs. increased market sensitivity to own orders/reduced liquidity.

4. correlation risk. total loss of return-to-average property.

there are others, but they are unlikely to be of interest anymore. these are the main ones.

Thank you, but the student FOXXXi already passed the exam. Yes, the losses were 4) and 3) at low st.dev.And they say that there are no risks in classical arbitrage, they are there too.Then what, you can immediately leave everything, everywhere you look, all risks.

 
FOXXXi писал(а) >>

Thank you, but the student FOXXXi has already passed the exam. Yes, there were losses on 4) and 3) at low st.dev.And they say that there are no risks in classical arbitrage, they are there too.What's next, you can leave everything at once, everywhere you look, all risks.

Do not take chances.

For this reason, you need technology to go out and come back in time.

 

The first three points are understandable, they are beyond our control. But the 4th point is the inadequacy of the model. So, the test for I(0) was not done well enough. I do not believe that the Nobel laureates did not provide statistical criteria for the significance of the stationarity hypothesis.

 
Mathemat писал(а) >>

The first three points are understandable, they are beyond our control. But the 4th point is the inadequacy of the model. So, the test for I(0) was not done well enough. I do not believe that the Nobel laureates did not provide statistical criteria for the significance of the stationarity hypothesis.

let me add.

correlation risk can be modelled using, for example, multivariate Garch models.

it is also hedged.

 
Quant >> :

4. correlation risk. total loss of the return-to-average property.

Had the "pleasure" of experiencing it first hand.

And a colleague, believing in the graality of the method, and not wanting to take a loss in time, actually lost out on the fall of GM.

 
FOXXXi писал(а) >> Ripped the pieces out in your favour.

I wrote facts, nothing but facts. Unfortunately, the facts are not in your favour. It happens. No big deal.

FOXXXi wrote >> you're playing the poor sheep again.

Unfortunately, I couldn't find a place where I could pretend to be that poor sheep.

FOXXXi wrote(a) >> Instead of rudeness you should write facts and more facts.

I did write the facts. Nothing personal. Facts are not pleasant. But you'll get over it, I think....))))

FOXXXi wrote(a) >> Once again, I confirm that this is mansplaining.

You have a sick imagination. Out-of-sample is not a disease or a "bad" habit. Where you get that from, I don't understand. If you don't know what it is, then I suggest you study what it is first and then write about similarities to onanism or all sorts of clinical cases.

FOXXXi wrote(a) >> You got yourself in trouble.

I didn't come across as I didn't write that -

FOXXXXXi wrote(a) >> Out-of-sample,forward and other crap is a clinical case.

I don't understand your comparison of out-of-sample to onanism or some clinical diseases. So could you please be more specific in explaining their similarities?

FOXXXi wrote(a) >> Come on show me your system exactly by TA or neural network, show me now how cool you are.Tell me what tools you work on, I can't refute the air.
I didn't claim to be Cool anywhere. Moreover, I am not Igor Krutoy - you have me confused with him. And I don't owe you anything, thank God. You are the one who compared Out-of-sample with masturbation and clinical viral infections. So everyone is wondering how relevant this comparison is in today's world and how it affects the equity tilt? .....)))))
 

You're bickering in vain, gentlemen. Especially in such a manner.

It is normal to use OOS (Out Of Sample).

In science and engineering the method of sliding control or boot-strap, which implies multistage OOS testing, has been known and applied for a long time.

The method itself is heavy, but its effectiveness is beyond doubt.

For us traders, it is usually neutered to a single run on OOS, so it works a little worse, and sometimes does not work at all.

So there you go. )))

1...141516171819202122
New comment