Commercial solutions
30 views
Skip to first unread message
TraderS
May 18, 2014, 10:33:14 AM5/18/14
to automated-tradi...@googlegroups.com
Besides tssb are there any other solutions you have tried, preferably free or better open source? I ma still trying to figure out tssb. I think this is more of a statistician toolbox rather than a practical practical system design software.
Dave Witkin
May 18, 2014, 10:41:39 AM5/18/14
to automated-tradi...@googlegroups.com
Consider QuantShare (www.quantshare.com). Not free, but inexpensive and lots of well-written tutorials.
On Sun, May 18, 2014 at 10:33 AM, TraderS <serg...@gmail.com> wrote:
Besides tssb are there any other solutions you have tried, preferably free or better open source? I ma still trying to figure out tssb. I think this is more of a statistician toolbox rather than a practical practical system design software.
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-sys...@googlegroups.com.
To post to this group, send email to automated-tradi...@googlegroups.com.
Visit this group at http://groups.google.com/group/automated-trading-system-design.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/17784a8c-2129-41d0-9090-e4a3e5274c2f%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
JUAN LUIS MARTÍN MOYA
May 18, 2014, 10:51:32 AM5/18/14
to automated-tradi...@googlegroups.com
Try http://www.asirikuy.com, and Kantu. It is very interesting....
El 18/05/2014 16:33, "TraderS" <serg...@gmail.com> escribió:
Besides tssb are there any other solutions you have tried, preferably free or better open source? I ma still trying to figure out tssb. I think this is more of a statistician toolbox rather than a practical practical system design software.
--
Message has been deleted
Sergio Sokales
May 18, 2014, 10:59:32 AM5/18/14
to automated-tradi...@googlegroups.com
Do they have a demo? I will look into it. I never heard of it. Is it genetic programming based or NN predictor based?
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/CAMQOku6hPBWj%2BAOeP65Ehue_rMJju%2BnFBqs9NPpcDkQXmLW8Wg%40mail.gmail.com.
JUAN LUIS MARTÍN MOYA
May 18, 2014, 11:03:21 AM5/18/14
to automated-tradi...@googlegroups.com
TraderS
May 18, 2014, 11:05:19 AM5/18/14
to
Interesting but after reading this I suspect the developers undermine the issues of data-mining
- Optimize list of rules to find which combination of rules performs best.
- Optimize ranking systems to find which combination of nodes gives the best output.
- Optimize trading systems to find which set of parameters leads to the most profitable trading system.
- Optimize prediction models to find which model has the better prediction accuracy.
- Create your own fitness formula to fully control the optimization process."
"- Two algorithms: Genetic algorithm or Population based incremental learning algorithm.
- Optimize list of rules to find which combination of rules performs best.
- Optimize ranking systems to find which combination of nodes gives the best output.
- Optimize trading systems to find which set of parameters leads to the most profitable trading system.
- Optimize prediction models to find which model has the better prediction accuracy.
- Create your own fitness formula to fully control the optimization process."
What they describe above is a recipe of failure. Optimize means to fool yourself and there is no way around it as when market conditions change the system fails.
On Sunday, May 18, 2014 5:41:39 PM UTC+3, dwitkin wrote:
Consider QuantShare (www.quantshare.com). Not free, but inexpensive and lots of well-written tutorials.
On Sun, May 18, 2014 at 10:33 AM, TraderS wrote:
Besides tssb are there any other solutions you have tried, preferably free or better open source? I ma still trying to figure out tssb. I think this is more of a statistician toolbox rather than a practical practical system design software.
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-system-design+unsubscribe@googlegroups.com.
To post to this group, send email to automated-trading-system-des...@googlegroups.com.
TraderS
May 18, 2014, 11:08:46 AM5/18/14
to automated-tradi...@googlegroups.com
Thanks. Will take a look and respond with my opinions.
On Sunday, May 18, 2014 6:03:21 PM UTC+3, juan73luis wrote:
El 18/05/2014 16:59, "Sergio Sokales" escribió:
Do they have a demo? I will look into it. I never heard of it. Is it genetic programming based or NN predictor based?
On Sun, May 18, 2014 at 5:51 PM, JUAN LUIS MARTÍN MOYA <juan7...@gmail.com> wrote:
Try http://www.asirikuy.com, and Kantu. It is very interesting....
El 18/05/2014 16:33, "TraderS"
Besides tssb are there any other solutions you have tried, preferably free or better open source? I ma still trying to figure out tssb. I think this is more of a statistician toolbox rather than a practical practical system design software.
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-system-design+unsubscribe@googlegroups.com.
To post to this group, send email to automated-trading-system-des...@googlegroups.com.
Visit this group at http://groups.google.com/group/automated-trading-system-design.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/17784a8c-2129-41d0-9090-e4a3e5274c2f%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-system-design+unsubscribe@googlegroups.com.
To post to this group, send email to automated-trading-system-des...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/CAMQOku6hPBWj%2BAOeP65Ehue_rMJju%2BnFBqs9NPpcDkQXmLW8Wg%40mail.gmail.com.
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-system-design+unsubscribe@googlegroups.com.
To post to this group, send email to automated-trading-system-des...@googlegroups.com.
FXtrader
May 20, 2014, 10:34:55 AM5/20/14
to automated-tradi...@googlegroups.com
Hi, have you tried the genetic algorithm? Is this for optimization of systems you define yourself or it also finds systems from scratch?
On Sunday, May 18, 2014 10:41:39 AM UTC-4, dwitkin wrote:
Consider QuantShare (www.quantshare.com). Not free, but inexpensive and lots of well-written tutorials.
On Sun, May 18, 2014 at 10:33 AM, TraderS <serg...@gmail.com> wrote:
Besides tssb are there any other solutions you have tried, preferably free or better open source? I ma still trying to figure out tssb. I think this is more of a statistician toolbox rather than a practical practical system design software.
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-system-design+unsubscribe@googlegroups.com.
To post to this group, send email to automated-trading-system-des...@googlegroups.com.
FXtrader
May 20, 2014, 10:37:03 AM5/20/14
to automated-tradi...@googlegroups.com
Thanks for the link.
On Sunday, May 18, 2014 11:03:21 AM UTC-4, juan73luis wrote:
El 18/05/2014 16:59, "Sergio Sokales" <serg...@gmail.com> escribió:
Do they have a demo? I will look into it. I never heard of it. Is it genetic programming based or NN predictor based?
On Sun, May 18, 2014 at 5:51 PM, JUAN LUIS MARTÍN MOYA <juan7...@gmail.com> wrote:
Try http://www.asirikuy.com, and Kantu. It is very interesting....
El 18/05/2014 16:33, "TraderS" <serg...@gmail.com> escribió:
Besides tssb are there any other solutions you have tried, preferably free or better open source? I ma still trying to figure out tssb. I think this is more of a statistician toolbox rather than a practical practical system design software.--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-system-design+unsubscribe@googlegroups.com.
To post to this group, send email to automated-trading-system-des...@googlegroups.com.
Visit this group at http://groups.google.com/group/automated-trading-system-design.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/17784a8c-2129-41d0-9090-e4a3e5274c2f%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-system-design+unsubscribe@googlegroups.com.
To post to this group, send email to automated-trading-system-des...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/CAMQOku6hPBWj%2BAOeP65Ehue_rMJju%2BnFBqs9NPpcDkQXmLW8Wg%40mail.gmail.com.
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-system-design+unsubscribe@googlegroups.com.
To post to this group, send email to automated-trading-system-des...@googlegroups.com.
FXtrader
May 20, 2014, 10:50:46 AM5/20/14
to automated-tradi...@googlegroups.com
Dave Witkin
May 18, 2014, 11:08:09 AM5/18/14
to automated-tradi...@googlegroups.com
A tool is just a tool. Any tool can be misused.
Funny you reference over-fitting. My business partner just won a $10,000 award from NAAIM for a methodology for avoiding data mining bias (i.e., over-fitting). And take a guess what software he and I use?
You avoid over-fitting in a number of ways. There is no substitute for good development / testing processes regardless of the software you use.Funny you reference over-fitting. My business partner just won a $10,000 award from NAAIM for a methodology for avoiding data mining bias (i.e., over-fitting). And take a guess what software he and I use?
On Sun, May 18, 2014 at 10:57 AM, Sergio Sokales <serg...@gmail.com> wrote:
Interesting but after reading this I suspect the developers undermine the issues of data-ming
"- Two algorithms: Genetic algorithm or Population based incremental learning algorithm.
- Optimize list of rules to find which combination of rules performs best.
- Optimize ranking systems to find which combination of nodes gives the best output.
- Optimize trading systems to find which set of parameters leads to the most profitable trading system.
- Optimize prediction models to find which model has the better prediction accuracy.
- Create your own fitness formula to fully control the optimization process."
What they describe above is a recipe of failure. Optimize means to fool yourself and there is no way around it as when market conditions change the system fails.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/CAN-W%3D-V2%2BZLcb3RcP8M5r%2BQmLVMUrvFWf7AxWeGGB5R2mtKRKw%40mail.gmail.com.
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-sys...@googlegroups.com.
To post to this group, send email to automated-tradi...@googlegroups.com.
Visit this group at http://groups.google.com/group/automated-trading-system-design.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/CAOX0msn_Mom0VoYOMH%2BHjuF-zwHGX6Q15DXDWcQVe5xwZ%2B0VEg%40mail.gmail.com.
TraderS
May 22, 2014, 6:56:23 AM5/22/14
to automated-tradi...@googlegroups.com
Thanks for the link. Interesting site and blog.
TraderS
May 22, 2014, 7:06:32 AM5/22/14
to automated-tradi...@googlegroups.com
I downloaded the demo of kantu system generator and tried it for 3 days but I gave up. It was a frustrating exercise. My comments:
(1) The program crashes quite often without generating any error messages and settings are lost
(2) Documentation is poor in the form of a blog post or similar, not of a commercial product.
(3) There are no error checks and no warnings
(4) Figuring out the correct settings to use is a monumental task even more difficult than finding an edge by trial and error. This is OK is this is a personal system but not for a commercial product
(5) I wasn't able to generate even one system even though I tried many different settings. Maybe I was doing something wrong but regardless the program generated no warnings and just crashes or hang in there
(6) I personally think the principles based on which this program was built are not sound but I admit that opinions vary and I am be the one who is wrong.
(7) The data format is related to some specific platform and it is uncommon and hard to generate by other platforms. For example I could not find a way to use simpler mm/dd/yyyy,o,h,l,c,vol format
(8) The WFA part is interesting but I did not get to that. I do not believe it can add any value however for eliminating curve fitting.
TraderS
May 22, 2014, 7:19:26 AM5/22/14
to automated-tradi...@googlegroups.com
I agree that one must have some way of avoiding data mining bias. Usually users of backtesting and charting programs, mainly retail traders, are not aware of the issues and when vendors promote optimization as a tool for developing systems they are in a way mislead, partly due to their ignorance. Vendors may be involved in a race who is going to provide the faster optimization when it is a fact that, as you said, that without a method for dealing with the bias it is a potentially disastrous tool.
Any link to your partners work? Thanks
On Sunday, May 18, 2014 6:08:09 PM UTC+3, dwitkin wrote:
A tool is just a tool. Any tool can be misused.You avoid over-fitting in a number of ways. There is no substitute for good development / testing processes regardless of the software you use.
Funny you reference over-fitting. My business partner just won a $10,000 award from NAAIM for a methodology for avoiding data mining bias (i.e., over-fitting). And take a guess what software he and I use?
On Sun, May 18, 2014 at 10:57 AM, Sergio Sokales wrote:
Interesting but after reading this I suspect the developers undermine the issues of data-ming"- Two algorithms: Genetic algorithm or Population based incremental learning algorithm.
- Optimize list of rules to find which combination of rules performs best.
- Optimize ranking systems to find which combination of nodes gives the best output.
- Optimize trading systems to find which set of parameters leads to the most profitable trading system.
- Optimize prediction models to find which model has the better prediction accuracy.
- Create your own fitness formula to fully control the optimization process."
What they describe above is a recipe of failure. Optimize means to fool yourself and there is no way around it as when market conditions change the system fails.
On Sun, May 18, 2014 at 5:41 PM, Dave Witkin <dwi...@wconsultingllc.com> wrote:
Consider QuantShare (www.quantshare.com). Not free, but inexpensive and lots of well-written tutorials.
On Sun, May 18, 2014 at 10:33 AM, TraderS wrote:
Besides tssb are there any other solutions you have tried, preferably free or better open source? I ma still trying to figure out tssb. I think this is more of a statistician toolbox rather than a practical practical system design software.
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-system-design+unsubscribe@googlegroups.com.
To post to this group, send email to automated-trading-system-des...@googlegroups.com.
Visit this group at http://groups.google.com/group/automated-trading-system-design.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/17784a8c-2129-41d0-9090-e4a3e5274c2f%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-system-design+unsubscribe@googlegroups.com.
To post to this group, send email to automated-trading-system-des...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/CAN-W%3D-V2%2BZLcb3RcP8M5r%2BQmLVMUrvFWf7AxWeGGB5R2mtKRKw%40mail.gmail.com.
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-system-design+unsubscribe@googlegroups.com.
To post to this group, send email to automated-trading-system-des...@googlegroups.com.
dwitkin
May 22, 2014, 7:34:59 AM5/22/14
to automated-tradi...@googlegroups.com
TraderS - Sure... here are a few links to the paper.
This page has links here to the paper: http://www.statistrade.com/dave-walton-wagner-award.html
And it doesn't sound like you need it, but other readers who aren't familiar with Data Mining Bias and how it lures people in might benefit from this simplified summary we wrote for NAAIM:
http://www.statistrade.com/support-files/understanding_2014_wagner_award_paper.pdf
This page has links here to the paper: http://www.statistrade.com/dave-walton-wagner-award.html
And it doesn't sound like you need it, but other readers who aren't familiar with Data Mining Bias and how it lures people in might benefit from this simplified summary we wrote for NAAIM:
http://www.statistrade.com/support-files/understanding_2014_wagner_award_paper.pdf
On Thursday, May 22, 2014 7:19:26 AM UTC-4, TraderS wrote:
I agree that one must have some way of avoiding data mining bias. Usually users of backtesting and charting programs, mainly retail traders, are not aware of the issues and when vendors promote optimization as a tool for developing systems they are in a way mislead, partly due to their ignorance. Vendors may be involved in a race who is going to provide the faster optimization when it is a fact that, as you said, that without a method for dealing with the bias it is a potentially disastrous tool.Any link to your partners work? Thanks
On Sunday, May 18, 2014 6:08:09 PM UTC+3, dwitkin wrote:
A tool is just a tool. Any tool can be misused.You avoid over-fitting in a number of ways. There is no substitute for good development / testing processes regardless of the software you use.
Funny you reference over-fitting. My business partner just won a $10,000 award from NAAIM for a methodology for avoiding data mining bias (i.e., over-fitting). And take a guess what software he and I use?
--
On Sun, May 18, 2014 at 10:57 AM, Sergio Sokales wrote:
Interesting but after reading this I suspect the developers undermine the issues of data-ming"- Two algorithms: Genetic algorithm or Population based incremental learning algorithm.
- Optimize list of rules to find which combination of rules performs best.
- Optimize ranking systems to find which combination of nodes gives the best output.
- Optimize trading systems to find which set of parameters leads to the most profitable trading system.
- Optimize prediction models to find which model has the better prediction accuracy.
- Create your own fitness formula to fully control the optimization process."
What they describe above is a recipe of failure. Optimize means to fool yourself and there is no way around it as when market conditions change the system fails.
On Sun, May 18, 2014 at 5:41 PM, Dave Witkin <dwi...@wconsultingllc.com> wrote:
Consider QuantShare (www.quantshare.com). Not free, but inexpensive and lots of well-written tutorials.
--
Dave Witkin
On Sun, May 18, 2014 at 10:33 AM, TraderS wrote:Besides tssb are there any other solutions you have tried, preferably free or better open source? I ma still trying to figure out tssb. I think this is more of a statistician toolbox rather than a practical practical system design software.--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-system-design+unsubscribe@googlegroups.com.
To post to this group, send email to automated-trading-system-des...@googlegroups.com.
Visit this group at http://groups.google.com/group/automated-trading-system-design.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/17784a8c-2129-41d0-9090-e4a3e5274c2f%40googlegroups.com.
For more options, visit https://groups.google.com/d/optout.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/CAN-W%3D-V2%2BZLcb3RcP8M5r%2BQmLVMUrvFWf7AxWeGGB5R2mtKRKw%40mail.gmail.com.--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-system-design+unsubscribe@googlegroups.com.
To post to this group, send email to automated-trading-system-des...@googlegroups.com.
Visit this group at http://groups.google.com/group/automated-trading-system-design.
juauis
May 23, 2014, 1:43:56 AM5/23/14
to automated-tradi...@googlegroups.com
Yes, you are right. The software is in a very early stage, just mentioned for reference. I'm looking for any software/method that generate traging rules, and the underlaying methods are very interesting for me. I began with Adaptrade, but I feel not sure with this procedure, I feel I go constantly to anywhere...
Price Action Lab looks interesting, I did not know the site(confused with Trading System Lab). FXtrader, thanks for the reference, and thanks too to dwtkin, a very nice article, I will dig into that topic.
Message has been deleted
FXtrader
May 25, 2014, 2:40:41 AM5/25/14
to automated-tradi...@googlegroups.com
The summary pdf does not explain the process and also makes a few unsupported claims. The original paper thesis rests on the assumption of choosing the correct parameter interval for the method to work. Therefore the method is not independent of guessing parameters to start with and as a result is subject to bias itself. Also the paper ignores the large body of works on this subject by universities and other researchers and their proposed solutions.
TraderS
May 25, 2014, 7:24:46 AM5/25/14
to automated-tradi...@googlegroups.com
Good paper but I don't think it solves the data-mining problem. There is also danger of introducing more bias if the system developer repeats the procedure many times using the same data.
Dave Witkin
May 25, 2014, 9:55:04 AM5/25/14
to automated-tradi...@googlegroups.com
Hi. Yes, the 2-page summary does not describe the process. The audience for the 2-pager was a group who, we understood, had very limited exposure to what data mining bias was. It was intended to help them understand it. I thought I mentioned that it was for readers who weren't familiar with the Data Mining Bias and how it can lure people in, but I might not have mentioned that. We were provided 2 pages and decided how best to use them based on our understanding of the audience. I won't comment on whether it makes unsupported claims, other than to say it is illustrative, not scientific. You could probably argue every word is unsupported since we didn't cite scientific references in the document.
Yes, the SPP method relies on the judgement of the developer. I'm not nearly as versed in the topic as my partner so I don't know if he would agree, but I'm not aware of a perfect solution to the issue - all the ones I've seen have advantages and disadvantages (e.g., simplicity versus complexity, ability to reuse data versus not). We personally use a number of methods, all of which have advantages and disadvantages, in the hope that through using a variety of imperfect methods we will reduce the bias enough to identify useful trading systems.I'm sure we are all interested in anything that will push the discussion forward. For my own trading, I would be interested in any methods in the body of literature you mentioned that are easy for an average system developer to apply.
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-sys...@googlegroups.com.
To post to this group, send email to automated-tradi...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/25399e97-fbef-42fa-9e94-e45ed030cb56%40googlegroups.com.
Dave Witkin
May 25, 2014, 10:00:20 AM5/25/14
to automated-tradi...@googlegroups.com
I'm in complete agreement the paper doesn't solve the data mining bias. I don't believe there is one single, perfect solution, so a variety of imperfect solutions may be the best we can do. I am interested in understanding people's views on better solutions, and why they believe they are better.
My personal hope would be that good solutions are relatively easy to apply. If I have to take painstaking care to meet the criteria of some models, I personally find them too difficult to use.--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-sys...@googlegroups.com.
To post to this group, send email to automated-tradi...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/5b497570-31eb-4f48-ad9f-5c017298885f%40googlegroups.com.
FXtrader
May 26, 2014, 3:02:23 AM5/26/14
to automated-tradi...@googlegroups.com
I remember reading a similar article in stocks and commodities years ago that argued in favor of using all the data and varying the parameters to get a distribution. This is nothing complex or new. The problem is that in the paper you posted is not entirely clear what the choice of the final parameters to use is and based on which distributions they are selected from. The task is left up to the developer to choose the range of parameters and the distributions. So the only novelty of the paper appears to be the use of the whole available sample in optimization of parameters but it does not provide a concrete process for selecting the final parameters. Therefore I think there is little usefulness here as the whole process is too vague to apply anyways. The SPA and the Romano and Wolf tests for data snooping bias are much more powerful and specific and have been used extensively by the industry (Ref. http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1087044)
I also ran a few tests with this method for a few trading systems and it doesn't seem to provide useful info as the proper distributions depend on the choice of parameter range and the median performance changes accordingly.
Hugh Hammer
May 26, 2014, 8:06:51 AM5/26/14
to automated-tradi...@googlegroups.com
I found this paper interesting
Hugh
--
You received this message because you are subscribed to the Google Groups "Automated Trading System Design" group.
To unsubscribe from this group and stop receiving emails from it, send an email to automated-trading-sys...@googlegroups.com.
To post to this group, send email to automated-tradi...@googlegroups.com.
Visit this group at http://groups.google.com/group/automated-trading-system-design.
To view this discussion on the web visit https://groups.google.com/d/msgid/automated-trading-system-design/25219e2a-f78e-4dd5-9801-3d571d473a2a%40googlegroups.com.
TraderS
May 29, 2014, 6:51:13 AM5/29/14
to
It may be interesting but it does not offer anything new. The effectiveness of the method depends on the selection of a right interval of parameters to use. If the selection is wrong, the results will be wrong. If another selection is made then this introduces data mining bias. A true method must remove guesswork. On top of that there is no justification why the method may work. At least one blogger has examples where this method fails. In my opinion the paper itself was the result of data mining.
Reply all
Reply to author
Forward
0 new messages