Welcome, Guest. Please Login
Algorithmic Trading Forum
 
  HomeHelpSearchLogin  
 
Pages: 1 2 3 
Send Topic Print
A must to read BOOK(s) on Algorithmic Trading... (Read 83971 times)
Co0olCat
Forum Moderator
*****
Offline



Posts: 128

Gender: male
A must to read BOOK(s) on Algorithmic Trading...
07/15/08 at 11:38:34
 
Did you read a book on Algorithmic Trading which you think belongs to "a must to read" category?

Share your opinion here

A poll to be started as soon as we collect several titles...
Back to top
 
 

"Yesterday is history. Tomorrow is a mystery. But today is a gift, and that is why it's called the present."
WWW timuryusupov   IP Logged
Co0olCat
Forum Moderator
*****
Offline



Posts: 128

Gender: male
Re: A must to read BOOKs on Algorithmic Trading...
Reply #1 - 07/15/08 at 11:43:37
 
Ernie Chan (Author), November 2008, Wiley Trading

Quantitative Trading: How to Build Your Own Algorithmic Trading Business
http://eu.wiley.com/WileyCDA/WileyTitle/productCd-0470284889,descCd-description....

Description

An innovative guide to understanding and implementing highly effective algorithmic trading techniques

The business of algorithmic trading was an activity once reserved for only traders at hedge funds or the proprietary trading operations of financial institutions. Not so anymore says author Ernie Chan, a proprietary trader and blogger who runs the quantitative trading blog site, epchan.blogspot.com. In Quantitative Trading, Chan shows investors how to use Excel and MATLAB(r) to build their own algorithmic trading tools using a budget even a home day trader can afford. He then reveals how to conduct quantitative research and analysis, and discusses what it takes to turn quantitative trading strategies into profits using stocks, ETFs, and other financial instruments. Chan also provides downloadable spreadsheets and MATLAB programs that tie into material covered throughout this book.

About the author: Ernest P. Chan, PhD (New York, NY), is a quantitative trader and consultant who advises clients on how to implement automated, statistical trading strategies.
Back to top
 
 

"Yesterday is history. Tomorrow is a mystery. But today is a gift, and that is why it's called the present."
WWW timuryusupov   IP Logged
Co0olCat
Forum Moderator
*****
Offline



Posts: 128

Gender: male
Re: A must to read BOOKs on Algorithmic Trading...
Reply #2 - 07/15/08 at 11:58:30
 
Gençay, R. et al (Authors), May 2001, Academic Press

An Introduction to High-Frequency Finance (Hardcover)
http://www.amazon.com/Introduction-High-Frequency-Finance-Ramazan-Gen%C3%A7ay/dp...

Description

An Introduction to High-Frequency Finance is the first source of unified information about high-frequency data. It provides a framework for the analysis, modeling, and inference of high-frequency financial time series. With particular emphasis on foreign exchange markets, as well as currency, interest rate, and bond futures markets, this unified view of high-frequency time series methods investigates the price formation process and concludes by reviewing techniques for constructing systematic trading models for financial assets.
Back to top
 
 

"Yesterday is history. Tomorrow is a mystery. But today is a gift, and that is why it's called the present."
WWW timuryusupov   IP Logged
qroach
Full Member
***
Offline



Posts: 102
London
Gender: male
Re: A must to read BOOK(s) on Algorithmic Trading...
Reply #3 - 07/15/08 at 16:05:52
 
But Chan's book has not even been released. One can only guess what is in there. Why are you so sure it is a "must read"?
Back to top
 
 
  IP Logged
Co0olCat
Forum Moderator
*****
Offline



Posts: 128

Gender: male
Re: A must to read BOOK(s) on Algorithmic Trading...
Reply #4 - 07/15/08 at 16:14:08
 
It is a valid point. The book is not available yet. But that makes it interesting...
Besides, I only suggested a candidate for "a must read" category. Later the community will decide who is the winner (I mean which book) 8)
Back to top
 
 

"Yesterday is history. Tomorrow is a mystery. But today is a gift, and that is why it's called the present."
WWW timuryusupov   IP Logged
qroach
Full Member
***
Offline



Posts: 102
London
Gender: male
Re: A must to read BOOK(s) on Algorithmic Trading...
Reply #5 - 07/15/08 at 16:30:48
 
Ok, then.
I will venture this book:
David R Aronson, "Evidence-Based Technical Analysis: Applying the Scientific Method and Statistical Inference to Trading Signals"
http://www.amazon.co.uk/Evidence-Based-Technical-Analysis-Scientific-Statistical...

Product Description
Evidence Based Technical Analysis examines how you can apply the scientific method, and recently developed statistical tests, to determine the true effectiveness of technical trading signals. Throughout the book, expert David Aronson provides you with comprehensive coverage of this new methodology, which is specifically designed for evaluating the performance of rules/signals that are discovered by data mining.


My own opinion is that it is packed with trivialities and that it makes rather long introduction into some very basic statistics, but, in addition, it does make a strong point about data-mining and data-snooping biases, discusses EMH, provides good condensed description of Reality Check test and leads the way to AI in trading.
To many people still experimenting with simple trading strategies on platforms like Metatrader, Metastock etc... this book can be an eye-opener though.
Back to top
 
 
  IP Logged
Diego
Newbie
*
Offline

I like this forum!

Posts: 1

Re: A must to read BOOK(s) on Algorithmic Trading...
Reply #6 - 07/22/08 at 14:41:59
 
I'm actually reading Evidence Based Technical Analysis by David Aronson.

This book covers a lot about the scientific method and psychology of traders and people in general. It's a good book to guide you when you're developing strategies and algorithms. It helps  to keep in mind how much you can be wrong when you think you are sure.

I didn't reach yet the part of the book where trading strategies testing are discussed, but as soon as I end the book, I will write a little more about it.
Back to top
 
 
  IP Logged
qroach
Full Member
***
Offline



Posts: 102
London
Gender: male
Re: A must to read BOOK(s) on Algorithmic Trading...
Reply #7 - 07/22/08 at 15:55:19
 
It is good to show what is wrong, but hardy shows anything what's right. It goes as far as to test simple one rule based strategies and measure data-mining bias on them just to show that none of them work. Yes, I know that I suggested this book and yet it seems like I am criticising it. Ce la vie
Back to top
 
 
  IP Logged
statstrader
Newbie
*
Offline

I like this forum!

Posts: 27

Re: A must to read BOOK(s) on Algorithmic Trading...
Reply #8 - 01/22/09 at 22:52:26
 
Hello All,

New here and I've read all the books above and many more, so hopefully I'll be able to add some useful discourse.  Regarding, Evidenced based TA, one problem I have is that by bootstrapping each daily data bit, you are losing any inherent properties in serial correlation.  He acknowledges this a bit on his site, but doesn't go much farther.  Serial correlation is the foundation of most modeling strategies, such as econometrics and time series, and removing it takes away from the information content IMO.  Personally, I prefer to sample larger sequences of data when bootstrapping (even then, you are removing some dependencies).

Also, I didn't think the 1st book had much to do with algorithmic trading.  I think it would be interesting to have an objective definition of algorithmic trading, considering the site name Smiley

While I don't have a precise one, the book that best truly describes what algorithmic trading is, has the title, "electronic and algorithmic trading technology, the complete guide," by kendall kim.
Similar to the descriptions the book covers, I consider algorithmic trading to be more of a method to break up liquidity and find ways to efficiently and stealthily process orders without making an impact on the market.  Unfortunately, many of those methods, such as third party exchanges are not available to small traders.

Otherwise, regarding trading systems and methodology, I might stick more to quantitative trading.  

Great start on the forum and hopefully it attracts the best, brightest, and most humble (if that's possible).

Back to top
 
 
  IP Logged
qroach
Full Member
***
Offline



Posts: 102
London
Gender: male
Re: A must to read BOOK(s) on Algorithmic Trading...
Reply #9 - 01/22/09 at 23:03:40
 
I don't quite follow you about bootstrapping. Of course if serial correlation is to be exploited by a trading strategy then removing it by permuting or resampling make a difference that you need to measure. Therefore removing serial correlation has to be a goal rather than a side-effect. Or are you talking about different scenario?
Back to top
 
 
  IP Logged
statstrader
Newbie
*
Offline

I like this forum!

Posts: 27

Re: A must to read BOOK(s) on Algorithmic Trading...
Reply #10 - 01/22/09 at 23:26:45
 
Hi Yury,

I'm not sure that your goal should be removing serial correlation.  In the context of econometric time series modeling that is often the goal, to remove any serial dependencies in the residuals, so that your residuals are purely gaussian noise and any dependencies in the data set are explained by the model.

What I was referring to is using bootstrapping of daily data as Aroonson does, and using it as a proxy for the universe of possible alternate outcomes (or estimate of the population) that any trading system could have encountered.  Not sure how best to explain it, but I think the method itself has an inherent bias.

For example suppose you wanted to test a moving average system, and had all the dow daily data.  Next, you want to compare to alternate rule-set based on the same underlying data.  My hypothesis is that by sampling sequences of data (bootstrapping is fine) of some length greater than one, then your sampling distribution would be closer to the true dow population (since it contains serial information), than would be a bootstrapped 1 bit/sample distribution.  Do you agree or disagree with that assertion?  If you agree, then hopefully you see how that may add a bias to the compared distribution.  It's not so much that we want to compare the distribution of the index data with a gaussian/chance distribution, but rather, we want to compare a rule set that operates on a given distribution and compare it to an alternate chance based ruleset on the same distribution
(which is certainly not gaussian, IID).

By the same tolken, if I am developing some system and I wish to run some monte carlo type of simulation of rules on the underlying data, I will make sure to sample rules applied on random sequences of the underlying data, but never change the order of the underlying data itself.  There are typically sufficient amounts of data sequences to sample from so that you would not need to generate alternate bootstrapped versions of the underlying data I
Back to top
 
 
  IP Logged
qroach
Full Member
***
Offline



Posts: 102
London
Gender: male
Re: A must to read BOOK(s) on Algorithmic Trading...
Reply #11 - 01/27/09 at 18:56:12
 
I understand that when you compare your "intelligent" rule set with a "random" rule set then yes, better not permute. In which case, I would even leave the historical quotes intact and generate many-many random rules sets.

I am not sure that boostrapping (even is block-boostrapping) can produce reliable results that reflect the fat tails. Fat tails are probably be better be modeleld as a random walk with the desired properties (jumps, infinite variance etc.). Or is infinite variance caused by serial correlation? That leaves us with non-Gaussian distribution that has no fat tails, which is not a big deal. I.e. bootstrapping sequences rather than single bits you will get different results, but not much different. Or am I wrong about it?
Back to top
 
 
  IP Logged
statstrader
Newbie
*
Offline

I like this forum!

Posts: 27

Re: A must to read BOOK(s) on Algorithmic Trading...
Reply #12 - 01/27/09 at 22:51:17
 
I would argue that most fat tails are caused by exogenous shocks or "jumps" in your model descriptions.
However, serial correlation would also likely have some impact on different time frames (ex: day vs. week variance ratios) as shocks tend to autocorrelate (GARCH) in serial fashion.

However, take away the fat tails, and the problem still exists.

I included a small hypothetical example here.  System A has captured an underlying order in system B (the index it is applying rules on). The sign order in B is always - - + + XXXX, where the Xs are don't cares or don't trade because the signs can change in those variables.  According to my backtests, system A is enormously profitable, since it never loses and always captures the relevant serial information that is embedded in B.

However, according to a T-test (w/assumed unequal variances), the system I devised is no better than chance.  Therefore it is a failing system and does not reject the null that this system is no better than chance.  This is the basic premise of the entire book as I understand it.



See any problems here? P.S. thanks
for keeping the threads alive. Wish more lurkers would opine.



Back to top
 
 
  IP Logged
Z
Newbie
*
Offline

Reality is not your
friend.

Posts: 8
London
Gender: male
Re: A must to read BOOK(s) on Algorithmic Trading...
Reply #13 - 01/28/09 at 05:52:30
 
I don't think a t-Test is an appropriate tool in this particular hypothetical example.
Back to top
 
 
  IP Logged
statstrader
Newbie
*
Offline

I like this forum!

Posts: 27

Re: A must to read BOOK(s) on Algorithmic Trading...
Reply #14 - 01/28/09 at 07:38:46
 
Fair enough.  But please take the paired sample set, and explain how the test results deviate significantly from the tests proposed in the book we are discussing.

If the premise is using bootstrap sampling or monte carlo permutations as outlined in the book, then could you show (in an alternate table)  how any of the author (or his teacher's papers) are more appropriate to identify that system A is very significant?  Keep in mind I devised a rule set and generated a comparison sample representative of a market sequence of returns which should be sufficient for a t-test.

Otherwise, if you agree that none of the tests described are sufficient, then I think you are pretty much corroborating my original assertion,
No?  

P.S. Obviously, a larger sample set of my hypothetical system would be detected by a runs test, correlogram, or some other possible non-parametric tests; but let's focus on the book's approach.

P.S.S. It's a fantastic book to read btw, just that I have a few scenarios that don't seem to fit his approach.

Back to top
 
 
  IP Logged
Pages: 1 2 3 
Send Topic Print