Good article about data mining, Nicolas. Thanks ...

Regards, Ton.

  ----- Original Message ----- 
  From: discuslorraine 
  To: [email protected] 
  Sent: Wednesday, May 30, 2007 2:59 PM
  Subject: [amibroker] Re: Corr Matrix via Gfx functions


  Hello Ton,

  > BTW if you know the book of David Aronson - "Evidence Based TA" - my
  remark about randomness and TA is where EBTA is all about. 

  Yes, I know it, really a great book, lot of precious insight into
  statistics in general and into the data-mining bias in particular, I
  wished I knew it earlier : ) 

  For the ones not owning the book, this paper is an excellent start on
  the data mining bias: 
  http://kdl.cs.umass.edu/people/jensen/papers/mlj2000.pdf
  but the experiments in EBTA are much more useful for the trader .. 

  >I have mailed 10 time series of stocks to several technical analysts
  >around the world. Most of them in the Belenux. Question was to select
  >the 4 correct stocks. 6 were home made stocks with a randomizer. Of
  >the 49 answers I got the result was that only 1 analyst could select
  >the correct combination of 4 stocks. He could not tell me why he
  >elected these stocks. Most of them or 38 had zero correct stocks, 5
  >only 1 correct, 4 had 2 correct stocks and 1 had 3 correct stocks
  >selected. 

  Interesting, there are many textbooks containing artifical price time
  series following random walks, and indeed they look very real .. 

  Regards,

  Nicolas




   

Reply via email to