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

RE: [amibroker] Optimizing & Robustness of single parameters



PureBytes Links

Trading Reference Links




Hello Dave,Most of your questions 
will be answered when you run the code and step through your stocks in the 
Workspace. You will immediately note various types of behavior, either on a 
single stock or a group of stocks.I am very pleased if a system behaves 
at least 'satisfactory' on a number of stocks, if it only works on one i don't 
trust it. The CMO5 works very nice on a significant number of stocks. Similarly 
i don't trust systems that work only one year out of five. That doesn't mean it 
could no be traded... just that I would not feel conmformtable doing 
so.This is just another method of learning more about the behaviour of 
your parameters, I am not suggesting that this is the most important or only one 
you should use.Some more inline below...happy 
trading,Herman.-----Original Message-----From: Dave 
Merrill [mailto:dmerrill@xxxxxxx]Sent: 
Thursday, September 04, 2003 11:07 AMTo: 
amibroker@xxxxxxxxxxxxxxxSubject: RE: [amibroker] Optimizing & 
Robustness of single parametersvery interesting herman, thanks very much 
for this. looking forward tochecking it out.I'm slightly confused 
about where this fits in overall system evaluationthough. I see that the 
plot shows smoothness or uneveness of equity inresponse to changes in the 
parameter under test, but I have some questions.(please understand that 
I'm not disagreeing with anything at all. I'm justtrying to get my head 
around this approach system evaluation. I have a sensethat many schemes that 
backtest well do so only out of curve fit, and thatperformance can degrade 
at any time. IMO there's no more critical piece ofthe puzzle than how to 
distinguish a nice curve fit from something moregenerally applicable, and if 
parameters need periodic adjustment, how toknow how and when to do 
that.)- I assume you'd want to test all reasonable parameters to fully 
evaluate asystem. what does it mean if a system provides even response to 
changes insome parameters, but uneven response to others?<FONT 
color=#0000ff>        Even responses would 
indicate predictable behavior under changing market conditions.- 
what exactly does uneven response mean? that the system isn't robust 
andshould be discarded, or used with caution? might you also see 
unevenperformance if the test exceeded the usable range of the parameter, 
forinstance if an MA intended to be the faster of two MAs crossed over 
andbecame the slower 
one?       <FONT 
color=#0000ff>        An uneven response 
might indicate your system loses sync with the market, even under minor cahnges 
in overal market rhythm.- why would you expect behavior for all 
stocks to synchronize? somehistorically move much faster, or much more, and 
an MA, for instance, thatdidn't match the stock well wouldn't do very 
well.
    Never said that ALL 
stocks would synchronize, but some will. There are market leaders. You may want 
to read up on DT's Trade the Market systems.- what does it mean 
if equity is smooth over one time period, say 2000 topresent, but uneven, or 
consistently negative, for other time periods, say1992 to present? one 
possibility is that market behavior has changed speed,and a system would 
need to adjust. would that make it an unwise system touse, or just one that 
needs recalibrating periodically?
    You have to set your own 
testing range, some systems may work well over ten years others only start in 
1999. Never deny facts, accept what you see and try to work with 
it.thanks again for your thinking on this, very important. are 
there any booksyou'd recommend about techniques for evaluating and properly 
optimizingsystems?
    There must be but i don't 
know about any. Perhaps somebody else can suggest. If you haven't you may want 
to read the tutorial on system design on the AB 
site...dave------------------------ Yahoo! Groups 
Sponsor ---------------------~-->Buy Ink Cartridges or Refill Kits for 
Your HP, Epson, Canon or LexmarkPrinter at Myinks.com. Free s/h on orders 
$50 or more to the US & Canada. <A 
href="" 
target=_blank>http://www.c1tracking.com/l.asp?cid=5511<A 
href="" 
target=_blank>http://us.click.yahoo.com/l.m7sD/LIdGAA/qnsNAA/GHeqlB/TM---------------------------------------------------------------------~->Send 
BUG REPORTS to bugs@xxxxxxxxxxxxxSend SUGGESTIONS to 
suggest@xxxxxxxxxxxxx-----------------------------------------Post 
AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx(Web page: <A 
href="" 
target=_blank>http://groups.yahoo.com/group/amiquote/messages/)--------------------------------------------Check 
group FAQ at: <A 
href="" 
target=_blank>http://groups.yahoo.com/group/amibroker/files/groupfaq.htmlYour 
use of Yahoo! Groups is subject to <A href="" 
target=_blank>http://docs.yahoo.com/info/terms/






Yahoo! Groups Sponsor


  ADVERTISEMENT 









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.