|
PureBytes Links
Trading Reference Links
|
Al and others,
before I comment about efficiency ratio, let me
explain better by an example:
let's call I have built a simply trading system
based on resistance and support breakouts (or rebounds), and those levels
coincide to the most recent peak and trough (hard values) respectively. Such a
system need to be optimized, in order to find the best % value for "change"into
the peak/trough formula and finally to achieve the best performance. But,
for any system like this, you can easily verify that a global optimization
(tha same "change" % for all symbols) always gives a <FONT face=Arial
size=2>meaningfully worst result than an individual optimization.
It's true that some symbols give poor gains under
every conditions, but by an individual optimization it's possible to bouncefrom
a loss to a reasonable gain, with whatever symbol.
Lastly, what makes the difference? Why,
running the same trading system, a symbol needs a value of 3 for "change",
while another one needs a value of 12, or 16, to give its best
performance?
Perhaps, the "fractal efficiency ratio" could be
the answer, but at first glance I didn't found a meaningful relation. Anyway,
I'll check it again.
P.S. The formula for fractal efficiency ratio could
be the following:
TimePeriods = <FONT face="Courier New"
color=#ff00ff size=2>500<FONT
size=2>;//(10,10000)
ratio = (<FONT face="Courier New"
color=#0000ff>abs(Close - <FONT
face="Courier New" color=#0000ff>Ref<FONT
face="Courier New">(Close,-TimePeriods)))/(<FONT
face="Courier New" color=#0000ff>Sum<FONT
face="Courier New">(<FONT face="Courier New"
color=#0000ff>abs(Close-<FONT
face="Courier New" color=#0000ff>Ref<FONT
face="Courier New">(Close,-<FONT face="Courier New"
color=#ff00ff>1<FONT
face="Courier New">)),TimePeriods));
Best
regards,
<FONT
face=Arial>Franco
<BLOCKQUOTE
>
----- Original Message -----
<DIV
>From:
Avcinci
To: <A title=amibroker@xxxxxxxxxx
href="">amibroker@xxxxxxxxxxxxxxx
Sent: Friday, November 01, 2002 2:56
PM
Subject: Re: [amibroker] a few
considerations about optimization
Franco,
Recently I was talking to a professional trader who told me that some
stocks simply don't behave well in any system and others do extremely well (in
backtesting). He attributes this to a term called the "fractal efficiency
ratio," coined by Perry Kaufman. A stock has to have some non-random
movement to be predictable. It's the total change in price over a given
period, divided by the sum of the absolute values of all the daily changes in
price. If a stock has too small a directional component, then it's a
poor candidate for any system, regardless of how many filters
or refinements you add. You're better off using all that firepower
on a better target. I've been testing a lot of stocks lately individually,
finding that many simply give very bad backtest results and very non-robust
parameter coefficients. So, I eliminate them from my watchlist and concentrate
on those stocks that behave well. This seems to be working well. I haven't had
time to write any code yet to see if the good-performing stocks have a higher
fractal efficiency ratio (personality as you call it?) than the poor
performing ones, but it's worth a try. You must test over a long enough
period of time to encompass bullish, bearish, and sideways markets, like
1/1/97 (or even earlier) to present time. If you try this idea out, let me
know how successful you are. I'm very interested in this concept. When I get a
chance, I'll try it myself. But, in theory, it seems to have merit.
Al Venosa
<BLOCKQUOTE
>
----- Original Message -----
<DIV
>From:
Franco Fornari
To: <A title=amibroker@xxxxxxxxxxxx
href="">amibroker@xxxxxxxxxxxxxxx
Sent: Friday, November 01, 2002 6:15
AM
Subject: [amibroker] a few
considerations about optimization
Hello,
trying to optimize any trading system,I think
we all have thought, sometime, we would like to avoid such a tedious process
or to do it once and for all.
It could be possible? This question badgered me
for a long time, unfortunately with no success, yet I feel there must be a
solution.
Why I say that? Because a peculiarity of
each stock, called "personality" by someone, wich seems stable enough. In
other words, I think if we were able to mathematically represent this
characteristic, we could automatically optimize any trading
systems.
But, the big matter is: what is this
characteristic (long term volatility, frequency of peaks and troughs,
price)? How could we assess or measure it? And, first of all, does such
a feature exist or is it only a mirage? How do you think
about?
Best regards,
FrancoPost
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.
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.
|