Serving the Quantitative Finance Community

 
User avatar
deskquant
Topic Author
Posts: 0
Joined: October 24th, 2010, 2:54 am

machine learning paper by german creamer.

September 18th, 2011, 4:05 pm

http://papers.ssrn.com/sol3/papers.cfm? ... 847Looking at the sharpes, seems to good to be true. Not sure if anyone had a look or had specific thoughts on a live implementation?
 
User avatar
tags
Posts: 3162
Joined: February 21st, 2010, 12:58 pm

machine learning paper by german creamer.

September 18th, 2011, 7:57 pm

##double##
Last edited by tags on September 17th, 2011, 10:00 pm, edited 1 time in total.
 
User avatar
tags
Posts: 3162
Joined: February 21st, 2010, 12:58 pm

machine learning paper by german creamer.

September 18th, 2011, 7:57 pm

has anyone ever been able to obtain results as ridiculous as those exposed in a publicly released study ??
 
User avatar
chocolatemoney
Posts: 0
Joined: October 8th, 2008, 6:50 am

machine learning paper by german creamer.

September 20th, 2011, 5:37 am

overfitting => extreme sensitivity in changes in the "trading environment" ??
 
User avatar
ktang
Posts: 0
Joined: January 15th, 2010, 7:16 pm

machine learning paper by german creamer.

September 20th, 2011, 6:27 am

QuoteOriginally posted by: edouardhas anyone ever been able to obtain results as ridiculous as those exposed in a publicly released study ?? Do you mind to share your results with us ?
 
User avatar
Hansi
Posts: 41
Joined: January 25th, 2010, 11:47 am

machine learning paper by german creamer.

September 20th, 2011, 7:40 am

QuoteOriginally posted by: ktangQuoteOriginally posted by: edouardhas anyone ever been able to obtain results as ridiculous as those exposed in a publicly released study ?? Do you mind to share your results with us ?I took his post to mean that most academic public research has better performance displayed than what you can really gain using the same methods in the market because of sample bias, data cleansing etc.
 
User avatar
ktang
Posts: 0
Joined: January 15th, 2010, 7:16 pm

machine learning paper by german creamer.

September 20th, 2011, 8:27 am

QuoteOriginally posted by: HansiQuoteOriginally posted by: ktangQuoteOriginally posted by: edouardhas anyone ever been able to obtain results as ridiculous as those exposed in a publicly released study ?? Do you mind to share your results with us ?I took his post to mean that most academic public research has better performance displayed than what you can really gain using the same methods in the market because of sample bias, data cleansing etc.Agree.Most of the research paper just use mid price and no trading cost. This is also one reason why you get a worser performance in real world.
 
User avatar
chocolatemoney
Posts: 0
Joined: October 8th, 2008, 6:50 am

machine learning paper by german creamer.

September 20th, 2011, 6:27 pm

I had a look at the paper, and I suspect that the model has been very tightly calibrated to the data.The result is a great paper performance in the calibration window, but when the market is live in the market, it gets carried away by the noise and performs poorly if the "market conditions" evolve.