|
PureBytes Links
Trading Reference Links
|
Hi list,
Since hypothetical backtested results in TS can only tell so much, below
I'm quoting OB results from some guy who has traded it real-time with
real money.
David who started this topic seems to think "trend filters" don't
improve results. My testing shows that trend filters DO improve results,
sometimes quite a bit. Nevertheless, imho OB's has had real problems
since Oct-2002 and by my testing (my feed etc) all basic variations (I
follow almost a dozen, including the infamous SOXball etc) are in a
8-month drawdown. This problem is true across a wide paramater range.
So, how have other people's OB-variants performed (OUT-OF-SAMPLE) the
last 6-9mo or so?
Regards, M
QUOTE
By RB on Wednesday, January 29, 2003 - 10:19 pm:
Regarding oddball performance:
I've taken every oddball trade since Feb 28, 2002. Profits based on
signal price since then are 238 pts on 264 trades (not including the
current open trade which is about -13). I use Real Tick for data. After
slippage I've made about 130 pts in 11 months (ave about .25 slippage on
entry and .25 slippage on exit using market orders on close of bar). The
person who says it has been in a drawdown for six months is absolutely
correct. In fact, it is currently in a 163 pt drawdown (including
slippage) and down 50 pts total since mid June. June through Oct
actually made money but losses the past three months have wiped out
profits and then some from mid June until now.
I've also tracked every signal per Mark Brown's website and they don't
agree exactly with his new report. There were a few times where Real
Tick was clearly one direction and the website went the other direction
(this happens frequently when the indicators are close to 1.0 or 3.0)
but these were extreme cases. For example, the indicator was -10 but
oddball website stayed long.. however, when the TS report was run, these
instancss have been "fixed" to reflect the proper trade. This is fine if
you used Real Tick or some other data vendor that had it right, but if
you used the website for your signals, you won't agree with his
hypothetical results. It is also another example of the inherent dangers
in hypothetical results.
bottom line is the system does make money but it is VERY difficult to
trade. Besides this 160 pt drawdown that I'm currently in, it had a 120
point drawdown and some other significant ones. The past three months
have all been negative but were preceeded by 6 straight winning months.
Also, even during the winning months there was normally a 50-75 pt
drawdown from the equity high of the month. I'm extremely disciplined
and take EVERY trade but it is not easy to do and my points per trade
aren't that impressive nor is my P&L considering the drawdown I'm in,
but the system is free and does work. Thanks to Mark for providing his
idea to us. Also I've experienced some three week periods where I've
made in excess of 100 pts and one could be right around the corner (OR
NOT). I hope it is, but the way its been struggling lately, it sure
doesn't feel like it.
Like many of you, I'm considering trading I-master instead of or in
conjunction with oddball.
Good trading to everyone and thanks for all the good dialogue on this
board. I've been at this for many years and find the current conditions
extremely challenging. I do fear the conditions will get worse, not
better, in the next couple years. Hope I'm wrong.
/QUOTE
David Colin wrote:
>
> Here is the (most of) 2002 results for Mark's "Original OB" and the "OB Time
> Change" variation I posted last year. It is an extremely small change (just
> changing the time you start taking trades from 0930 to 1100 EST) but usually
> improves the results quite a bit, both increasing net profit while reducing
> trade frequency. Actually this is the only variation I have tested that
> improved results in every time period I looked at, sometimes quite
> dramatically.
>
> Let me also say that none of those "trend filter" type variations that I
> posted held up last year (NYSE tick, Adv/ Decl line >< 0 etc). I am
> somewhat mystified by this since their results were quite good in
> backtesting and it's not like they are curve fit or anything. But they
> earned about 25% of what OB original did with a bigger DD. So whatever.
> Thanks for listening.
>
> OB Original Results Jan. 1 2002 - Dec. 19 2002
>
> Net: 108 K
> # trades: 282
> MIDD: 27,675
> PF: 1.36
> % winners: 46.1
> Avg. Trade: 383.07
>
> OB Time Change Jan. 1 2002 - Dec. 19 2002
>
> Net: 134 K
> # trades: 216
> MIDD: 22,150
> PF: 1.55
> % winners: 50
> Avg. Trade: 620.37
|