Calling all Quants and Statistical Analysts | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Calling to all Programmers! let's remake this 123 indicator 1 reply

Calling all Geniuses 4 replies

Calling all Divergence traders; what's your favorite indicator? 5 replies

Calling all Quants and Statistical Analysts

  • #1
  • First Post: Jan 19, 2011 10:04am Jan 19, 2011 10:04am
Hey

I'm currently studying Mathamatics and Statistics and finally getting my head around it. Just wondering if this is something anyone uses much in their trading or backtesting and how useful you find it and how advanced you take it.

Most people fail it the first time I've heard and the majority don't even turn up to the exam, its aparently the hardest subject in the degree but my teacher says that if you master it you can bet on the horses! - I guess you could apply the same logic to trading.

Anyway would be cool if anyone has some experience with it to describe the sort of things you do with it... I know how its roughly applied in stocks but just the basics for diversification and that sort of thing

Day trading is slightly different I would imagine.
Time hides Nothing
  • #2
  • Jan 19, 2011 11:22am Jan 19, 2011 11:22am
I'll go as far as the basic concept allow me nothing more or less.

From my own perspective it won't help you as a trader knowing to much about it, if you love math do it as a hobby. They will be of a true help when trying to find the expectancy of the model you are trading

math is more theoretical and very abstract, where as the statistics, will give you a solid foundation understanding probability and decision making under uncertainty.

I think that it will be of any help if you work in derivatives bc pretty much everything in that area revolves around statistics on some level, but even in that field you only need to understand math beyond basic calculus.

So I'll resume as follow: only move around the basics

Bear in mind that's just the way I see it.

all the best!!
  • #3
  • Edited 2:10pm Jan 19, 2011 1:52pm | Edited 2:10pm
That's cool! I love statistics (although I'm weak in maths)! I remember the time about a year ago when I would spend most of my free time on the website of the Estonian statistics office. Oh, the good times. They roll fast.

Can't give you much insight about the application part, though. When I took Statistics 101 I played around with some of the basic formulas and came up with some pretty neat stuff. However, I was unable to apply it to trading due to my lack of programming skills and Excel is only good enough if you're interested in dailies (I know some whiz kids are able to do better by feeding Excel with real time data, but that's not me). Those formulas are still gathering dust somewhere. Some time in the future, I intend to learn a bit of programming and put those things to good use, but you know what us humans are like...promises of doing it tomorrow or some time in the future end up being like the signs they used to have in saloons - free beer tomorrow (come back tomorrow and they would say "Come back tomorrow") haha...

Now that's a lot of talk and not a lot of useful information so I'm going to stop here. Thanks for raising this topic, though, it's definitely worth discussing.

PS. I'm not a quant or a statistical analyst. I'm here by mistake.
If you don't risk, you don't ever have to lose.
  • #4
  • Jan 19, 2011 2:18pm Jan 19, 2011 2:18pm
Cindy, good luck with your study first of all...Math and Statistics are essential components I have to say when it comes to trading based on probabilities and system design...

I am an Electrical Engineer, so I studied lots of Mathematics in my life, however, I never studied statistics since Electrical Engineering rarely deal with uncertainties with few exceptions...

I studied statistics though because I knew that trading without understanding probabilities is like walking in the dark in a mine field....

I would be more than happy to help if I can....


Regards,

Nader
  • #5
  • Jan 19, 2011 4:19pm Jan 19, 2011 4:19pm
Quoting narafa
Cindy, good luck with your study first of all...Math and Statistics are essential components I have to say when it comes to trading based on probabilities and system design...

I am an Electrical Engineer, so I studied lots of Mathematics in my life, however, I never studied statistics since Electrical Engineering rarely deal with uncertainties with few exceptions...

I studied statistics though because I knew that trading without understanding probabilities is like walking in the dark in a mine field....

I would be more than happy to help if...
it seems that we are both on the same professional career field. I'm a graduate of the Faculty of Electronics and Telecommunications, but don't love the maths anymore as I did in the past.

all the best!
  • #6
  • Jan 19, 2011 4:36pm Jan 19, 2011 4:36pm
I'll go as far to say Signal analysis was the best thing for trading that I took. Statistic -so so. Removing noise from a time series, kernel smoothing, spectral analysis... etc. It really changes the perspective. If you can find some classes about it the better.
  • #7
  • Jan 19, 2011 5:19pm Jan 19, 2011 5:19pm
Quoting triger88990
it seems that we are both on the same professional career field. I'm a graduate of the Faculty of Electronics and Telecommunications, but don't love the maths anymore as I did in the past.

all the best!
I used to adore Math in the past...Not anymore exactly like you....It seems that as one move deeper into practical life and reality and away from theory, bonds start to get loose....

All the best to you too


Regards,

Nader
  • #8
  • Jan 19, 2011 5:24pm Jan 19, 2011 5:24pm
Quoting bug
That's cool! I love statistics (although I'm weak in maths)! I remember the time about a year ago when I would spend most of my free time on the website of the Estonian statistics office. Oh, the good times. They roll fast.

Can't give you much insight about the application part, though. When I took Statistics 101 I played around with some of the basic formulas and came up with some pretty neat stuff. However, I was unable to apply it to trading due to my lack of programming skills and Excel is only good enough if you're interested in dailies...
lol bug
  • #9
  • Jan 19, 2011 6:45pm Jan 19, 2011 6:45pm
Quoting CindyXXXX
if you master it you can bet on the horses!.
My first university level stats professor is published regarding stats and horse racing. He has made a lot of dough doing it too.
  • #10
  • Jan 19, 2011 6:59pm Jan 19, 2011 6:59pm
i hear that they make programs to forecast horse racing,depending on the race course,mud,weather and all types of crap

did u r proffessor make heaps of money on that?
  • #11
  • Jan 19, 2011 7:03pm Jan 19, 2011 7:03pm
lol, are you guys for real? do you ever enter a trade without knowing your signals statistical edge? oh common...
  • #12
  • Edited 11:03pm Jan 19, 2011 10:41pm | Edited 11:03pm
Thanks for the replies

Hers some handy info if you are interested in learning about stats or anything else for that matter

http://www.khanacademy.org/ (not selling anything its a free site)

This guy is a genius and a godsend... The crappy text book they make us use makes no sense for a beginner and the lecturers don't really give a shit if we're learning or not. If they were as passionate about teaching as this guy there wouldn't be such a skills shortage IMO]]Anyway check it out he does a whole section on statistics and another one on probabilities. Hes a pure legend and helped me understand 5 weeks of material in just a day... I shit you not! And he does it for free despite being chased (wanted) by hedge funds he does this.

Anyway... As I progress through the course I may update this thread but feel free to use this thread for discussion on probabilities and statistics. I'm slowly learning to love it

And on a side not there's a guy in my town who is a multimillionaire and made his fortune on the horses. He was a pretty ordinary kid with a gift for maths (sound like a movie trailor doesn't it!) but anyway, my love is trading but just thought Id mention it.

I think he takes the trifector bets where the odds can blow out to 1000:1 so backed by investors he probably loses a whole bunch but hits the jackpot every now and then. Interesting guy

Cheers
Time hides Nothing
  • #13
  • Jan 20, 2011 3:19am Jan 20, 2011 3:19am
Khanacademy is a terrific project and I've learned a lot from the guy. Since CindyXXX brought up the topic of (free) online media, I'm going to share something here as well.

1. The most important video you'll ever see (well, that's the title at least)
Albert Bartlett talks about the concept of exponential growth and peak oil. I don't want to take up the matter of peak oil here, so if you're not convinced, keep it to yourself. There much more in his videos that just peak oil. I'm posting part 1 of 8, but you can follow the links on youtube to watch them all.

2. Game Theory with Professor Ben Polak at Yale University.
I can't say I took the course, but I took the course. The wonderful thing about this is that you don't have to be a Yale student - you can just download the material for free! You can listen or watch all the course lectures on the web (left hand side menu, click downloads). This stuff is brilliant.

The web is full of free university lectures and course materials, you just have to do some digging.
If you don't risk, you don't ever have to lose.
  • #14
  • Jan 20, 2011 4:28am Jan 20, 2011 4:28am
Quoting CindyXXXX
Thanks for the replies

Hers some handy info if you are interested in learning about stats or anything else for that matter

http://www.khanacademy.org/ (not selling anything its a free site)
WoW....An excellent resource.....And the great thing is that you can actually download the videos...

Thanks a lot for sharing this...


Regards,

Nader
  • #15
  • Jan 20, 2011 6:31am Jan 20, 2011 6:31am
Quoting narafa
WoW....An excellent resource.....And the great thing is that you can actually download the videos...

Thanks a lot for sharing this...


Regards,

Nader
No probs Nader amazing stuff really I was blown away to find that e wry single video there was from the same guy
Time hides Nothing
  • #16
  • Jan 20, 2011 6:34am Jan 20, 2011 6:34am
Quoting bug
Khanacademy is a terrific project and I've learned a lot from the guy. Since CindyXXX brought up the topic of (free) online media, I'm going to share something here as well.

1. The most important video you'll ever see (well, that's the title at least)
Albert Bartlett talks about the concept of exponential growth and peak oil. I don't want to take up the matter of peak oil here, so if you're not convinced, keep it to yourself. There much more in his videos that just peak oil. I'm posting...
Thanks Bug will add them to my study schedule Theres just not enough hours in a day!
Time hides Nothing
  • #17
  • Jan 21, 2011 10:41pm Jan 21, 2011 10:41pm
Hi Cindy,

I have a pretty technical background (MS in Nuclear Engineering, and currently working on second MS in Operations Research - half way done with it) ... and am very interesting in trading, and how statistics can be applied to it. As a side project, I’ve been working on some statistical analysis using the past 6 years of weekly price data to gain some insight into weekly behavior. I will post more when finished with this project, but here is a sample of my findings:

For EURUSD:
Summary of range:
Min. 1st Qu. Median Mean 3rd Qu. Max.
0.00920 0.02008 0.02605 0.02871 0.03295 0.13550

So the average (mean) range is 287.1 pips … and 50% of all of the weekly ranges are between 200.9 and 329.5 pips.

If its an up week, then there is a 69.5% chance of closing within the top 30% of the range. Ex: if weekly range is 300 pips, then there is a 69.5% chance of closing within 100 pips of the high

If its an up week, then there is a 50.6% chance of closing within the top 20% of the range.

If its an up week, then there is a 26.8% chance of closing within the top 10% of the range.

Its something like a 75% chance of closing within 33% of the range within the high for an up week ... pretty good odds!

So there is a better than 50/50 chance of closing within 20% of the weekly range’s high if an up week. This proves mathematically that its best to hold till the week’s close if long. If price was normally distributed for the weekly range, then it would be 50/50 chance of closing in the middle of the weekly range.

If you are a stats student, you will almost certainly use R (an open source statistical package). I've posted my R script here if you are intested:

http://www.forexfactory.com/showthre...45149&page=274

Regards,
EZCurrency
  • #18
  • Edited 11:34pm Jan 21, 2011 11:06pm | Edited 11:34pm
Hi EZ,

I applaud you for your work, however, there is something that doesn't seem quite right about the assumptions. I've seen a similar approach in the book long/short market dynamics. However, like the author's approach, what you quantified here doesn't seem to be useful other than describing information in hindsight.

Correct me if I'm in err, but when you define up week, you are stating that it will close within x% of the range after the fact (i.e. first you slice the data into up and dn weeks, and then describe some conditional statistics). It would seem more useful to look at the data up to some time range (say the duration or fraction of an up week) and then quantify how far up or down it will move FOLLOWING that week GIVEN some condition (i.e. it was UP or DN in the prior week/n periods). If I understood your description and code properly, it is useless to say that if you are in a long week then hold, because it's likely to continue up.. you don't know that the way the data was partitioned.

Quoting ezcurrency
...
So there is a better than 50/50 chance of closing within 20% of the weekly range’s high if an up week. This proves mathematically that its best to hold till the week’s close if long.
Do you agree that you can't go backwards from having knowledge of an up week to utilizing those odds? We can't time travel (a.k.a. look-ahead bias)

Hopefully that makes sense, but again, good work and I'm not knocking it, because I run a lot of these types of stats as well.

Secondly, it's useful to look at how these properties change over time, rather than a static summary of stats; you'll find the probabilities are rarely constant.
  • #19
  • Jan 21, 2011 11:34pm Jan 21, 2011 11:34pm
Quoting jamjamjam
Do you agree that you can't go backwards from having knowledge of an up week to utilizing those odds? We can't time travel.
Depends on the criterias on which you base the qualification of a week's direction. Of course, if the one criteria is the positions of the week close relative to the open, then you can't know until the end of the week where it will close. If you take the current price versus the week open, then the status can change many times between up and down during the week the statistics cant be applied.
“Nothing great in the world has ever been accomplished without passion.”
  • #20
  • Jan 21, 2011 11:41pm Jan 21, 2011 11:41pm
Quoting Deevz
Depends on the criterias on which you base the qualification of a week's direction. Of course, if the one criteria is the positions of the week close relative to the open, then you can't know until the end of the week where it will close. If you take the current price versus the week open, then the status can change many times between up and down during the week the statistics cant be applied.
Right. Take a look at the code (he kindly shared). You'll see that right off the bat, the data was already partitioned into up and down weeks, and then statistics were applied on the partitioned data. As you said, among other problems, "Of course, if the one criteria is the positions of the week close relative to the open, then you can't know until the end of the week where it will close."

Your second condition, however, can be usefully summarized. You could say that given the current price is some percentage from the open it will behave a certain way during the rest of the week. Nothing wrong with that (although, it's not very sophisticated); no problem if the status is changing, that's precisely what summary statistics are for.
0 traders viewing now