> So.. If we consider that walk-forwarding
would eliminate all curve-fitting, then what do you consider to be the best
variable to walk-forward? RAR, CAR, Net profit%?
The best is your custom metric that fits your trading style.
As far as "generic", most simple and built-in metrics are
considered I would opt for CAR/MDD or RAR/MDD
General rule is that your metrics should always include
drawdown (risk) measure (in addition to profit measure).
Best regards, Tomasz Janeczko amibroker.com
----- Original Message -----
Sent: Wednesday, April 16, 2008 6:31
PM
Subject: Re: [amibroker] Re: Expectancy -
and related--specifically K-rato
Hi,
Thanks for your answers. I am reading Howard's
book right now (approx. half the book remaining) and my concern is really for
walk-forward as it seems better than simple optimizing.
In AB 5.08 I
can choose the variables to optimize. I tend to like k-ratio because it
shows consistency in the results; however the best k-ratios are almost never
the best net profit or CAR%. What I have read is here: http://www.addictfx.biz/15-categorie-90719.html
(It's in french... sorry for those who can't read french).
So.. If we
consider that walk-forwarding would eliminate all curve-fitting, then what do
you consider to be the best variable to walk-forward? RAR, CAR, Net
profit%?
Thanks,
Louis
2008/4/16, Tomasz Janeczko <groups@xxxxxxxxxxxxx>:
"I've read somewhere" - well the world (and Internet
specifically) is full of misinformation.
You really need to read Howard's book. Regular
optimization is NOT the same as walk forward.
Walk Forward process actually prevents/minimises curve
fitting.
-----
Original Message -----
Sent:
Wednesday, April 16, 2008 5:39 PM
Subject:
Re: [amibroker] Re: Expectancy - and related--specifically K-rato
Hi,
I've read somewhere that optimizing (or
walk-forwarding) using measures as k-ratio, RRR or max drawdown can lead
to curve-fitting and is not a good strategy. Do you
agree?
What do you think is the best optimizing
strategy?
Thanks,
Louis
2008/4/13, gerryjoz <geraldj@xxxxxxxxxx>:
Grant, in your post you asked me to elaborate on why i thought the
K-ratio was a waste of
space and RRR was simpler/better. What i have found is that k-ratio is generally lower the
higher the exposure for the same or similar trading systems in back
test. If you want a high k-ratio, according to the AB calc, don't buy or
sell! Here is a contrived (curve-fit) example (run on real data) over
a few years CAR 33% Profit factor 7 CAR/MDD 2.8 Max Sys
DD % 11.5% RRR 2.15 K-ratio .096 exposure 49% #trades 170
the
K-ratio definitio in AB
help is " K-Ratio -
Detects inconsistency in returns. Should be 1.0 or more. The higher
K ratio is the more
consistent return you may expect from the system. Linear regression
slope of equity line multiplied by square root of sum of squared
deviations of bar number divided by standard error of equity line
multiplied by square root of number of bars. More information: Stocks
& Commodities V14:3 (115-118): Measuring System Performance by
Lars N. Kestner " personally i prefer measures which are more
easily comprehended. This one isn't, even tho 40 years ago i did do
maths & stats at uni. In any case, back in May 2004 Tomasz
changed the calc... ======>
K-ratio calculation changed. following the change made by
its creator, Mr. Lars Kestner.
Quoting from the book
"Quantitative Trading Strategies" from 2003 by Lars Kestner:
[
- - - ] " The K-ratio
is a unitless measure of performance that can be compared across
markets and time periods. [ - - - ] Traders should search for
strategies yielding K-ratios greater than +0.50.
Together, the Sharpe ratio and K-ratio are the most
important measures when evaluating trading strategy performance.
Note: When I created the K-ratio in 1996, I thought I had created a robust
measure to evaluate performance. In mid-2000, trader Bob
Fuchs brought a small error to my attention regarding the scaling
of the K-ratio. He was
correct in his critique and I have corrected the error in this text.
Publications prior to 2002 will show a different formula for the
K-ratio. The updated
formula in this book is correct."
Mr Lars Kestner has
corrected his formula based on this critique: K-ratio = slope / ( sterr * per
)
slope: Linear regression slope of equity line sterr:
Standard error of slope per: Number of periods in the performance
test
Special thanks to Jeremy Berkovits who brought that to my
attention.
<====== There was quite a bit of discussion at
the time. I understand RRR intuitively, and when i look at the other
ratios i can see why one is higher or lower (with a bit of
checking).
Is it possible that there was a typo in the K-ratio correction? Perhaps
Mr Kestner has made another change? I don't have his books or
articles, i just gave up on the k-ratio because i didn't think it was telling me
anything useful.
I would be interested if you or anyone else have
run some examples where K-ratio is high and exposure is high, and what are the
other backtest numbers.
regards
Gerry
__._,_.___
Please note that this group is for discussion between users only.
To get support from AmiBroker please send an e-mail directly to
SUPPORT {at} amibroker.com
For NEW RELEASE ANNOUNCEMENTS and other news always check DEVLOG:
http://www.amibroker.com/devlog/
For other support material please check also:
http://www.amibroker.com/support.html
__,_._,___
|