|
PureBytes Links
Trading Reference Links
|
Most all systems can be "curve fitted".
The 4% VLE system could most certainly be the result of "curve
fitting" through optimization. Simply optimize the (4%) parameter
for values between 2.5% and 10% in .25% increments and see what what
happens to total performance. For example the result of the
optimization would be suspect if using 4% worked great, 4.25%
bombed, but using say 5% generated gains while 6% showed showed
losses.
I admit I did optimize for standard deviations and moving average
length on the Bolinger Band system. The system is surprisingly
robust for most parameters between 1 and 2.5 standard deviations
(.1% steps)
and 5 to 15 days on the moving average length. I tested the
optimized parameters (1.6 standard deviation & 6 day MA) over
multiple begining and ending dates with similiar good results. It
looks great trading the DIA and pretty good on the QQQ too.
In general a profitable trend following sytstem (like the 4% VLE)
will generate more losing trades than winners. While a profitable
contra-trend system (BBs) will generate more winning trades than
losers.
I like Bollinger bands systems for long/short trading. They buy
weakness and sell stregth and standard deviation is based on
volatility which is adaptive to current market conditions.
--- In equismetastock@xxxxxxxxxxxxxxx, manohohman <no_reply@xxxx>
wrote:
> This is an interesting systems as opposed to the 4% VLE system
that I posted.
>
> This system has the potential to be curve fitted. Curve fitting is
> simply changing variables in the expert equations, or adding
trading rules to a trading system to the point where the system is
tuned only for the historical data it is being tested on.
>
> In this system you have two variables which impact the outcome.
You have the BB periods and BB std devs. Both of which can be
tweaked for one security over a specified range of historical data.
BB systems are easily over optimized.
>
> In the 4% VlE system you have the VLE which can't be manipulated.
You have the 4% rule which can't be changed and you have a 20 dma
which is very common. The trading rules are take whichever signal
comes first. So there's only one trading rule. Once moving average,
which was not optimized (ie 20 not 22 or 36 or whatever) and you
have the 4% price differential which is fixed. That makes it really
hard to curve fit it. However, it is not robust in that it only
trades the price movement of the SPY, or the market as a whole, as
it was intended.
>
> The BB system is not robust either in that it is only applied to
the SPY. If it were applied to other markets, it would likely
produce much lower results. (Actually I did test it and it does
produce lower results as expected. Since it wasn't presented as a
robust trading system, I didn't bother sharing those results.)
>
> To test the degree of curve fit in the BB, I did a walk-forward
out-of-sample data test and in sample test. Walk-forward tests are
> difficult to do with MS because it does not have that as an
optional choice.
>
> Here are the results.
>
> Annual B&H Trades DrwDwn
> Ten Years 18.24% 5.84% 138/206 >0.003
>
> Seven Years 18.67% 10.42% 81/120 0.015
>
> OutSam
>
> Two Years 16.94% -3.62% 25/38 0.041
>
> One Year 31.20% -27.4% 14/18 0
>
> One Year 18.97% 27.98% 7/12 0.038
>
> Six Months 6.65% -6.68% 5/9 0.045
>
> Six Months 25.30% 23.00% 10/15 0.016
>
> You have to be careful how short your walk forward periods are
> because if they are too short, you'll chop off parts of the
drawdowns periods.
>
> In addition, in almost every year tested the lossing trades, both
> average and highest, were worse than the winning trades. However
> because winning trades out number losing trades the system is
> profitable.
>
> The TradeStation report on the 4% VLE system is different from MS
> reports. It ended its tests in 2002. As I said I didn't run it,
but I
> have the results. The annual rate of return was 21.7% vs about 6%
for
> the market. Tradestation doesn't give me the exact raito between
> winners and losers but it looks like 140 or so out 207 were good
> trades. In addition, the winning trades returned a much higher
> percentage of profit than losses.
>
> The 4% VLE system can't be curve fitted--at least by any
conventional
> means that I am aware of. I programmed it uniquely using a
highlight
> code which can't be used direclty in the systems tester. I am
> thinking about writing an expert for it and if I do, I will try to
> put it in a form that can be tested in MS. The way I want to do it
is
> a little tricky because I am not coding for the SPY, I will be
coding
> for the VLE and then transfering the entry and exit points to the
> SPY. How much fun can one guy have!
>
> Anyway, you can read the data and draw your own conclusions. If
you
> think that you need to do your own systems tests because you feel
> more comfortable with whatever you think you're doing that other
> people aren't, please do so, and then share the results. Recognize
> that there probably won't be a way to reconcile the
differences,and I
> just can't help you with figuring out how to do your own tests. If
> you don't know how to do these tests, the results probably aren't
> going to be reliable anyway.
>
> I think the BB system paired with some other non-curve fit non-
> colliner indicators might work well for the SPY. That's my opinion.
>
> JO
>
>
>
> --- In equismetastock@xxxxxxxxxxxxxxx, "personal592002"
> <personal592002@xxxx> wrote:
> > For your system backtesting enjoyment:
> >
> > Use 6 day Bollinger Bands +/-1.6 stadard deviations on end of
day
> > data on SPY. Buy at open when the closing price crosses above
the
> > lower Bolinger Band, sell when the closing price crosses below
the
> > upper Bollinger Band. Trade both long and short. Very simple &
> back
> > tests great on the SPY. Ten year performance over 2X that of buy
> and
> > hold, 207 trades in ten years 140 of which were profitable....
> >
> > Code -
> >
> > Buy & Buy to cover:
> >
> > Cross(CLOSE, BBandBot(CLOSE, 6, SIMPLE, 1.6))
> >
> > Sell & Sell Short:
> >
> > Cross(CLOSE, BBandTop(CLOSE, 6, SIMPLE, 1.6))
------------------------ Yahoo! Groups Sponsor ---------------------~-->
Buy Ink Cartridges or Refill Kits for your HP, Epson, Canon or Lexmark
Printer at MyInks.com. Free s/h on orders $50 or more to the US & Canada.
http://www.c1tracking.com/l.asp?cid=5511
http://us.click.yahoo.com/mOAaAA/3exGAA/qnsNAA/BefplB/TM
---------------------------------------------------------------------~->
To unsubscribe from this group, send an email to:
equismetastock-unsubscribe@xxxxxxxxxxxxxxx
Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
|