[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: Objective functions (was RE: [amibroker] Re: Optimization -- again)



PureBytes Links

Trading Reference Links




Pal,
 
For a system to be valid, it must work on all 
numbers tested, not just those with certain names and not others with 
different names.
Try telling Keith Fitchen that (the author of 
Aberration, the most successful mechanical system ever sold ... check out 
Futures Truth).  First, Keith will tell you that his wildly successful 
approach DOES NOT work on equities.  Second, Keith will tell you that 
he only trades a basket of six commodities.  I believe both these 
principles are directly contrary to your opinions about optimizing and the 
selection of issues to be traded.
 
If a system works on Bonds and not on Beans, this 
system is curve fitted over a specific set of data (Bonds) and it loses all 
statistical validity.
 
Wrong, wrong, wrong.  
If I have an approach that has worked on Bonds for ten years and it doesn't work 
on beans...BFD.  Should I abandon a robust approach to trading 
Bonds...because I can't make "beans" on 
Beans?   
 
Beware of drinking other people's bath water and 
whatever you do, don't drink the Kool Aid .   
 
Take care,
 
Steve
 
 
<BLOCKQUOTE 
>
  ----- Original Message ----- 
  <DIV 
  >From: 
  palsanand 
  
  To: <A title=amibroker@xxxxxxxxxxxxxxx 
  href="">amibroker@xxxxxxxxxxxxxxx 
  Sent: Monday, October 20, 2003 9:43 
  AM
  Subject: Objective functions (was RE: 
  [amibroker] Re: Optimization -- again)
  Hi,In my mind, curve fitting means either using 
  different systems for different markets, or using different parameters of 
  the same system for different markets, and this is not valid technical 
  analysis.Historical testing via computer means feeding a set of 
  numbers (open, ow, close prices), and receiving back an output set of 
  rules that hopefully will make money trading. The numbers themselves do 
  not have names, and the computer doesn't recognize the difference 
  between 'Beans' or 'Bonds'. For a system to be valid, it must work on 
  all numbers tested, not just those with certain names and not others 
  with different names.If a system works on Bonds and not on Beans, 
  this system is curve fitted over a specific set of data (Bonds) and it 
  loses all statistical validity. To believe it will work in the future as 
  it has worked in the past is very dangerous.Also, different 
  markets do not have different personalities. Again, they are reduced to 
  just being a set of numbers or a bunch of algorithms. If a channel 
  breakout (or any other) method is successful, then the same parameter must 
  be used for all the markets, for the same reasons as above. You cannot use 
  a 20-day channel in Silver and a 40-day channel in Corn, this also falls 
  under the crime of curve fitting.I therefore take exception to any 
  system, that either only trades one specific market or group of markets, 
  or trades different markets using different parameters or rules of the 
  same system. All this proves is what has worked best in the past, and this 
  will usually not continue to work in the future, as there is no 
  correlation under this scenario.This is not specifically written 
  to condemn vendors. This is a clarification of my definitions of 
  'optimizing' and 'curve fitting', and a warning as to what types of 
  trading systems may be valid and what to stay away 
  from.Regards,Pal--- In amibroker@xxxxxxxxxxxxxxx, "Dave 
  Merrill" <dmerrill@xxxx> wrote:> thanks, I'll check it out if 
  I can find it.> > I'm sure I'm ignorant, but how logic or sound 
  trading principles can be used> to set an MA period (for instance) 
  without examination of past history> escapes me. as does the 
  distinction between using past history or> 'experience' to do that and 
  optimization. as does the justification for> seeing optimizations 
  from one point in time as somehow blessed above all> 
  others.> > dave>   I would have to refer you to 
  an article published by Futures Magazine>   concerning 
  optimization and its research value in November 20?? by>   
  Kent Calhoun.> >   Possibly the only way to do it 
  correctly, is to first arrive at a set>   of parameters 
  and algorithm based on logic, experience, or sound>   trading 
  principals that won't be subject to change. Then do a walk>   
  forward with no attempt to improve results via optimization.> 
  >   Regards,> >   Pal> 
  > >   --- In amibroker@xxxxxxxxxxxxxxx, "Dave Merrill" 
  <dmerrill@xxxx>>   wrote:>   > Pal, 
  couple questions/comments.>   >>   > - 
  are you saying that 30 "occurrences" in any system produces 
  95%>   accuracy?>   > 30 trades? 
  regardless of the market or trading system rules or 
  time>   frame?>   > what's the basis 
  for saying this?>   >>   > - could you 
  explain "select stable parameters with an equity shift>   
  less than>   > the parameter shift after equity spikes 
  have been eliminated"? I>   don't>   > 
  understand what you mean.>   >>   > - 
  just fyi, your last paragraph seems to be trying to convince 
  me>   that>   > optimizing is good, 
  probably in response to my asking "if you>   prefer not 
  to>   > optimize parameters, how do you set them?". I 
  asked that only>   because you>   > said, 
  "I prefer a system to work without optimization", which I>   
  thought was>   > a nice goal, but one I don't understand 
  how to achieve. seems that>   you 
  don't>   > actually intend to avoid optimization either, 
  since you then>   discuss how you>   > do 
  it.>   >>   > dave>   
  >   There is a correct method to optimize any system that 
  is>   >   statistically valid, 30 occurrences 
  with 95% accuracy.>   >>   
  >   The key to optimization is to select stable parameters with 
  an>   equity>   >   shift less 
  than the parameter shift after equity spikes have been>   
  >   eliminated. This process creates stability for optimal 
  parameters>   >   shifts within the four 
  technical market phases. Parameter shift is>   
  >   always geometric, but equity shift decline relative to 
  unstable>   >   parameter selection is 
  usually exponential.>   >>   
  >   All systems are optimized to some degree. As soon as a 
  trader>   chooses>   >   to 
  enter a trade on the open as opposed to the high/low/close 
  of>   day,>   >   he has 
  made a decision as to how a system should be traded. 
  Does>   he>   >   know the 
  high/low/close of day entry is better than the next>   
  opening>   >   for an entry? If not why not? A 
  potential 28% difference in>   >   profitability 
  exists for channel system entries between opens and>   
  >   closes.>   >>   
  >   The purpose of trading is to consistently make money. This 
  is>   done by>   >   having the 
  best information available. If a trader does not know>   
  the>   >   best entry for his system, what is he 
  trying to prove? That the>   >   system 
  isn't optimized? To lose money because a trader is>   
  ignorant of>   >   his systen's best parameters 
  is foolish.>   >>   >   
  Regards,>   >>   >   
  Pal>   >   --- In amibroker@xxxxxxxxxxxxxxx, 
  "Dave Merrill" <dmerrill@xxxx>>   >   
  wrote:>   >   > one question pal: if you 
  prefer not to optimize parameters, how>   
  do>   >   you set>   
  >   > them? or do you have some kind of trading rules that 
  don't have>   time>   >   
  > constants, trigger levels, etc, that need to be set?>   
  >   >>   >   > 
  dave>   >   >>   
  >   >   I thought I might throw in my 2 
  cents.>   >   >>   
  >   >   Vendors love optimization, because it can 
  generate eye popping>   >   >   
  hypothetical profits which has no connection to real-time>   
  trading.>   >   >>   
  >   >   I prefer a system to work without 
  optimization. But if I have>   to>   
  >   do>   >   >   it, 
  I would make sure that the optimization is robust in 
  the>   >   >   following 
  manner:>   >   >>   
  >   >   1. The sample size of data should be large 
  enough to represent>   >   
  real->   >   >   time market 
  conditions - bull, bear and sideways markets.>   
  >   >>   >   >   2. 
  The look-back period should be as large as possible for 
  the>   >   same>   
  >   >   reasons.>   
  >   >>   >   >   3. 
  The testing of optimizable parameters should be on out 
  of>   >   sample>   
  >   >   data using walk-forward 
  analysis.>   >   >>   
  >   >   4. The Central Limit Theorem says that for a 
  sample to assume>   the>   
  >   >   characteristics of the population, the size 
  of sample should>   be>   
  >   >   large. The minimum sample size should be 
  around 30. But since>   an>   
  >   >   uptrend or downtrend can last for say 50 
  periods, I would>   have a>   
  >   >   minimum sample size of 100 periods making 
  sure that the full>   >   
  market>   >   >   cycle is there 
  (uptrend, downtrend and congestion).>   >   
  >>   >   >   5. The optimizable 
  parameters should be as few as possible and>   
  >   tested>   >   >   
  in a wide variety of markets.>   >   
  >>   >   >   Curve-fitting is 
  like rolling a fair dice with 1/6>   probability 
  of>   >   >   getting any number 
  from 1 to 6, rolling it 5 times, getting>   #6, 
  4>   >   out>   
  >   >   of 5 times (80%) of 
  time.>   >   >>   
  >   >   A lot of traders fall in the trap of 
  curve-fitting without>   being>   
  >   >   aware of it. So when designing a system, it 
  is important to>   keep>   
  >   your>   >   >   
  guard up as far as curve-fitting is concerned.>   
  >   >>   >   >   
  Regards,>   >   >>   
  >   >   Pal>   >   
  >   --- In amibroker@xxxxxxxxxxxxxxx, "Gary A. 
  Serkhoshian">   >   >   
  <serkhoshian777@xxxx> wrote:>   >   
  >   > Fred,>   >   
  >   >>   >   >   
  > Could you narrow-down your idea of a reasonable sample 
  size>   for>   >   
  >   backtests.  You've been hinting at rather sizeable 
  backtesting>   >   >   periods, 
  but would like to put some numbers to it.  Also>   
  wonder if>   >   >   you use # of 
  trades as a guide versus period of time for>   
  >   backtesting>   >   
  >   period.>   >   >   
  >>   >   >   > 
  Thanks,>   >   >   > 
  Gary>   >   >   
  >>   >   >   > Fred 
  <fctonetti@xxxx> wrote:>   >   
  >   > There are a lot of questions and provacative statements 
  in>   your>   >   
  >   post,>   >   >   
  > only one of which from my perspective needs an>   
  answer/response.>   >   >   
  >>   >   >   > Market 
  behavior will continually change after that ...>   
  >   >   >>   >   
  >   > Change ? from what ? into what ? I guess this is the 
  part I>   >   don't>   
  >   >   > follow.  To me there is nothing 
  new in market behavior now>   that>   
  >   >   > didn't exist last month, last year, 
  last decade, last>   century,>   
  >   but>   >   >   
  > clearly those that take a short sighted view of history 
  and>   the>   >   
  >   > market action that made up that history will clearly 
  never>   see>   >   
  it.>   >   >   > It's a forest 
  and trees thing ...>   >   >   
  >>   >   >   > --- In 
  amibroker@xxxxxxxxxxxxxxx, "Dave Merrill">   
  <dmerrill@xxxx>>   >   >   
  > wrote:>   >   >   > > 
  I'm not trying to be argumentative, honest (:-)... I'm>   
  more>   >   than>   
  >   >   a>   >   
  >   > little>   >   
  >   > > sick of saying the same thing over and over, but 
  I  j u s>   t>   >   d 
  o>   >   >   > n ' t   
  g>   >   >   > > e 
  t   i t .>   >   >   > 
  >>   >   >   > > 
  ------------------------------>   >   
  >   > >>   >   
  >   > > I fail to see the huge difference in principle 
  between>   equity>   >   
  >   > feedback and>   >   
  >   > > backtesting.>   >   
  >   > >>   >   
  >   > > let's start by assuming that backtesting 
  performance of a>   >   
  system>   >   >   > and 
  its>   >   >   > > parameters 
  over some period of past data tells you>   
  something>   >   >   
  about>   >   >   > 
  its>   >   >   > > future 
  performance. it's not a perfect predictor, but>   it's 
  the>   >   >   
  best>   >   >   > 
  evidence>   >   >   > > we 
  have. does this seem like a reasonable starting point?>   
  what>   >   >   > 
  alternative>   >   >   > > is 
  there?>   >   >   > 
  >>   >   >   > > if that's 
  true, why is it better to do it only once? what>   
  >   >   > justification is>   
  >   >   > > there for picking one examination 
  period over another?>   clearly>   
  >   >   > market>   
  >   >   > > behavior will change continually 
  after that. don't we>   need a>   
  >   way>   >   >   
  of>   >   >   > 
  working>   >   >   > > that 
  looks at what's been happening and evolves our>   
  response?>   >   >   > 
  >>   >   >   > > sounds 
  like we examine performance up to some point and>   
  >   adjust,>   >   >   
  > trade with>   >   >   > 
  > the best-choice system and parameters for a while, 
  then>   >   examine>   
  >   >   and>   >   
  >   > adjust>   >   
  >   > > again later. make sense? what alternative is 
  there?>   >   >   > 
  >>   >   >   > > so then, 
  how often do we re-examine performance history?>   
  to>   >   put it>   
  >   >   > > differently, how long do we ignore 
  any changes in market>   >   
  dynamics>   >   >   > that 
  may>   >   >   > > or may not 
  have occurred? why would intermittently>   refusing 
  to>   >   >   
  look>   >   >   > 
  and>   >   >   > > respond 
  improve system performance or reliability?>   
  >   >   > >>   
  >   >   > > if that needs to be done, why not 
  have the system itself>   do>   
  >   it,>   >   >   
  as>   >   >   > part 
  of>   >   >   > > its 
  inherent operation? why is it better for us as an>   
  outside>   >   >   
  agent>   >   >   > 
  to>   >   >   > > 
  periodically run some separate tests, reach into the>   
  >   internals of>   >   
  >   > the>   >   
  >   > > system, and change stuff?>   
  >   >   > >>   
  >   >   > > or should we just continue with 
  the system and parameters>   we>   
  >   >   choose>   >   
  >   > at the>   >   
  >   > > beginning? are they somehow more valid than what 
  we'd>   choose>   >   
  >   later,>   >   >   
  > using>   >   >   > > the 
  same backtesting methods, but on a different date>   range 
  of>   >   >   
  data?>   >   >   > 
  >>   >   >   > > 
  ------------------------------>   >   
  >   > >>   >   
  >   > > I realize that even if it seems to make sense 
  logically,>   this>   >   
  all>   >   >   
  a>   >   >   > 
  complete>   >   >   > > crock 
  if no systems put together like this even backtest>   
  well,>   >   >   > never 
  mind>   >   >   > > forward 
  testing.>   >   >   > 
  >>   >   >   > > but every 
  time I think about abandoning this line of>   
  >   research, it>   >   
  >   > seems like>   >   
  >   > > the first thing I'd want to do with a new system 
  would be>   >   (let me>   
  >   >   > guess),>   
  >   >   > > test and possibly adjust it using 
  data up to some date,>   then>   
  >   run>   >   >   
  > with it for>   >   >   > 
  > a while after that and see if equity growth is good. 
  if>   it is,>   >   
  >   I'd>   >   >   
  > want to>   >   >   > > 
  lather, rinse and repeat with other in and out of 
  sample>   >   data, to>   
  >   >   > make sure>   
  >   >   > > that wasn't 
  coincidence.>   >   >   > 
  >>   >   >   > > sounds 
  way too familiar to be a completely different>   
  animal.>   >   >   > 
  >>   >   >   > > 
  dave>   >   >   > 
  >   From: Fred [mailto:fctonetti@xxxx]>   
  >   >   > >>   
  >   >   > >   That IS what I was 
  trying to say.  I suspect because>   
  equity>   >   >   
  feed>   >   >   > 
  back>   >   >   > 
  >   is like looking in a rear view mirror, great 
  for>   letting us>   >   
  know>   >   >   > 
  >   where we were and how we could have adjusted the past 
  to>   >   make it>   
  >   >   > >   better, but that's 
  about it.> > 
  >         Yahoo! Groups 
  Sponsor> > > >   Send BUG REPORTS to 
  bugs@xxxx>   Send SUGGESTIONS to 
  suggest@xxxx>   
  ----------------------------------------->   Post 
  AmiQuote-related messages ONLY to: 
  amiquote@xxxxxxxxxxxxxxx>   (Web page: <A 
  href="">http://groups.yahoo.com/group/amiquote/messages/)>   
  -------------------------------------------->   Check group 
  FAQ at:> <A 
  href="">http://groups.yahoo.com/group/amibroker/files/groupfaq.html> 
  >   Your use of Yahoo! Groups is subject to the Yahoo! Terms 
  of Service.Send 
  BUG REPORTS to bugs@xxxxxxxxxxxxxSend SUGGESTIONS to 
  suggest@xxxxxxxxxxxxx-----------------------------------------Post 
  AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx (Web page: <A 
  href="">http://groups.yahoo.com/group/amiquote/messages/)--------------------------------------------Check 
  group FAQ at: <A 
  href="">http://groups.yahoo.com/group/amibroker/files/groupfaq.html 
  Your use of Yahoo! Groups is subject to the <A 
  href="">Yahoo! Terms of Service. 







Yahoo! Groups Sponsor












Send BUG REPORTS to bugs@xxxxxxxxxxxxx
Send SUGGESTIONS to suggest@xxxxxxxxxxxxx
-----------------------------------------
Post AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx 
(Web page: http://groups.yahoo.com/group/amiquote/messages/)
--------------------------------------------
Check group FAQ at: http://groups.yahoo.com/group/amibroker/files/groupfaq.html



Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.