Trendfollowers: When oh when are we going to start making some $$$?

[asiaprop]

then on Amazon you have a lot of disgruntled kids who have a hard time getting through even the most elementary stats classes in college. For them any book which does not get them an A aside their busy sports and social schedules sucks. I think its actually a very good book, whatever people say, cause I read and worked with it in the beginning.

And it kind of makes we wonder about you when you cite the publication year as a reason for your book disqualification. I would claim not much has happened at the basic stats and probability front since 1994 ;-)

Quote from Rodney King:

Not sure a book from 1994, $118 new and poorly reviewed on Amazon, is the right starting point.

 

[mizhael]

Quote from asiaprop:

mizhael,
However, if you run out of sample tests and get for several different forward tests the same statistically significant results then I dont see how this has anything to do with curve fitting.

So you use thousands of data points for in-sample statistical test and another thousands of data points for out-sample statistical test?

I dont want to bore you with theory go and see for yourself and tell me what you see when you observe the price pattern of AAPL over the past years versus a name such as RIM. And those two names are not even the most extreme examples and far from being outliers.

I saw on Bloomberg that if I plot the recent two years, they show very different behavior. But if I plot the past 10 years, then they are very similar indeed.

 

[mizhael]

Quote from asiaprop:

stats 101, here some references (I already told you some work also needs to be done by yourself).

* Econometric Analysis, Greene
* Mathematical Statistics and Data Analysis, John A. Rice
* Probability and Statistics, DeGroot, Schervish

Those are mostly the ones I went through in college before I delved into stochastic calculus and derivatives pricing in grad school (not MBA)

Thanks a lot!

I have learned those t-test stuff. However I just don't see what do I use t-test to: I am testing the statistical significance of trendiness? What's the null-hypothesis here? And how to do that? I am just missing something and need some enlightening...

 

[asiaprop]

a) yes I do, and sometimes tens or hundreds of thousands depending on bar frequencies or tick data.

b) Well use this: AAPL EQUITY RIMM EQUITY HS. It shows you some basic statistical properties of the correlation that goes on between those 2 names over specified time periods.

Quote from mizhael:

So you use thousands of data points for in-sample statistical test and another thousands of data points for out-sample statistical test?



I saw on Bloomberg that if I plot the recent two years, they show very different behavior. But if I plot the past 10 years, then they are very similar indeed.

 

[mizhael]

Quote from asiaprop:

absolutely, and I read. However, what they did was very inaccurate. First of all they did not use daily data (butter production in Bangladesh was probably not updated daily). So, take a look how many samples they used to make up their correlation profile. That's laughable, 11 observation points.

Secondly it does not have predictive power but simply showed how those 2 variables correlated over the same observation period, two very different concepts. At least thats what I see when I glanced over the setup, but maybe they lagged one time series, which I may have overlooked.

I am a very strong believer in statistical tests and probabilities. However, I take the discretion to override a model if something absolutely does not make sense to me, but very rarely. If you show me how butter production has predictive power in the short term and over at least 2000-3000 observations during different business cycles then I am very happy to look into it, no matter what others think of the logical relationship.

Good points! So if we found a strong correlation relation and showed at least 2000-3000 observations, and we lag one time series vs. the other time series before we throw them into the correlation calculator, then we should be able to trade the correlation, right?

We use the leading one to predict the lagging one and trade the lagging one?

 

[asiaprop]

no, not the statistical significance of trendiness (though you could use t-tests for that, too, of course). My point was to rigorously test the performance results of a trend system on one asset with in-sample data, then construct a t-test on the out-of-sample data (and you could cut the out of sample data into several clusters to test separately) with the null being that the out-of-sample performance does not hold up to the observed in-sample performance.

Quote from mizhael:

Thanks a lot!

I have learned those t-test stuff. However I just don't see what do I use t-test to: I am testing the statistical significance of trendiness? What's the null-hypothesis here? And how to do that? I am just missing something and need some enlightening...

 

[mizhael]

Quote from Equalizer:

I agree with you, that paper is good for laughs - hence it's name, all I'm trying to illustrate is that you definitely need to be aware of and always question any findings.

Many quant traders formulate a theory and then program and test the daylights out of it, but it is originally based on a market theory/observation. This does not preclude the existence of curve fitting based on data mining either, but it is waaaaaay better than throwing massive computing power at the data to sieve for spurious patterns.

Then again, mining can uncover relationships you never thought of. It is a fine balancing act.

But I'm sure you know this, it is more for the benefit of the OP.

I heard even pure data-mining works... Rentec is using that a lot.
So I think Rentec has invented a mechanism that they can try everything, even the moon-phase vs. stock price, but they know how to judge when it works and when it doesn't and then switch on and off fastly...

 

[Equalizer]

Quote from mizhael:

I heard even pure data-mining works... Rentec is using that a lot.
So I think Rentec has invented a mechanism that they can try everything, even the moon-phase vs. stock price, but they know how to judge when it works and when it doesn't and then switch on and off fastly...

This is not true.

 

[asiaprop]

sorry i am out of here, I already spent lot of time, need to finish up some work. ;-)

Have fun guys

 

[ProfLogic]

Quote from Equalizer:

LOL, unlike you asswipe, no-one in this thread (apart from you and your ass chum BoWo, you know, him being the world's greatest system and pairs trader - maybe Pears trader is more appropriate), has made any assertions about being a BSD, hence we don't need to prove shit to you jack or anyone else.

You however with your perfect trades, and ERGs (LMAO), have over the years behaved like the emperor of ET, like you know it all and that you have the one true method, without ever, EVER proving anything, and as soon as anyone questions anything you commence the ad hominem attacks and never EVER answer the questions. You should go into politics, seriously, you spin shit like there's no tomorrow.

So, since you're the mofo who thinks themselves Mr BSD-trader, how about you proves us all wrong by posting live intraday trades for a month?

No? I didn't think so dickless.

Why is it that when I hear "FrothLogic" I think of the words "Pathological Liar, Charlatan, Clown, Fraud, Crook, BS artist..."? I wonder?

Pushing them buttons am I . . . lol

You don't need to prove anything to me, you are incapable of proving anything to yourself.

You accuse me of something you can't prove because it is a lie . . . how ironic.

Search my posts moron. I've posted plenty of trades.