logo
Welcome Guest! To enable all features please Login or Register.

Notification

Icon
Error

3 Pages123>
Options
Go to last post Go to first unread
Anton Beijer  
#1 Posted : Sunday, February 13, 2011 2:42:03 PM(UTC)
Anton Beijer

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/16/2006(UTC)
Posts: 39

I have been testing and testing systems in the systemtester but there are as good as no systems that produce any better results then buy and hold (i.e. without optimization and equally good results in different time frames). I also don't like having to make 100 trades (with all the timings risks) for a only 5% better result.

Are we looking for the holy grail and is everybody looking and searching for the formula for turning lead into gold or ...?

The best multi stock-multi periods tested system I could produce was:

buy: Mo(Mov(HHV(C,21),70,E),10)>100
sell: Mo(Mov(HHV(C,21),70,E),10)<100

Anyone with a better system ???? (and then I don't mean the standard Metastock pp:: CCI modified with 200 transactions in 500 daily periods)

Anton
wabbit  
#2 Posted : Monday, February 14, 2011 4:33:16 AM(UTC)
wabbit

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers, Unverified Users
Joined: 10/28/2004(UTC)
Posts: 3,111
Location: Perth, Western Australia

Was thanked: 16 time(s) in 16 post(s)
IMHO, there is no one system that will work on every component of every market for every trader; if there was I am sure with the amount of backtesting that goes on, someone would have found it by now.

I believe trading is very personal; find your own trading style and apply this to the markets you are interested in. Find your own universe of stocks/currencies/etc and work on getting reliable profits from these. If buy and hold produces the best results, then I'd be buying and holding (nothing wrong with that, just ask Buffett and co.) You are going to have to discover your own tolerance to risk; if I said there was a trading system which could triple the account in 6 months, but the chance of it going bust was 75%, would you trade it? What about the system which will only make 15% in the same period but with a busting chance of only say 10%. Each trader has their own tolerance. I don't believe the systems in MS are any good straight out of the box as the chance of it perfectly suiting you would be quite low; see how the code is written and modify them to suit your own risks, goals and markets.


wabbit [:D]

Anton Beijer  
#3 Posted : Monday, February 14, 2011 10:15:56 AM(UTC)
Anton Beijer

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/16/2006(UTC)
Posts: 39

Thank Wabbit for your honest answer.

If backtesting won't work, future use will most certainly not going to work either. Isn't the cornerstone of Technical analysis that the past patterns are most likely to be continued and repeated with a certain certainty in the future? Without this fundamental there remains nothing more then gambling.

Every casino player also has it's own system en lucky periods. If the use a computer and analysis (or change computing) tells me that the chances of even keeping your money will never exceed 36 out of 37 on roulette, you better stay out because the changes of consistent winning are nil on the long term.

Blackjack card counting works and it disliked by the casino's because a computer can provide and compute faster and better high chance situations and keep you out of chance situations better than the human mind/memory. On the long end this way of playing black jack provides predictable profit into an otherwise predictable loss situation

The system I showed can be tested in very marked on every randomly chosen group of stocks from a index and it will at least beat buy and hold. (have tested it backwards on 350 periods, 450 etc, incremental with 100 periods per step up tot 3150 periods. In every period it beats the buy and hold results of a group of indices or a group of stocks. On individual stocks it may underperform in 10-30% but overall it outperforms buy and hold consistently in every historical situation the past 15 years.

Are those not the systems we are looking for and need computer power for: systems that rule out luck of market situations and use chance computing in our advantage?


henry1224  
#4 Posted : Monday, February 14, 2011 11:21:39 AM(UTC)
henry1224

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 10/29/2004(UTC)
Posts: 1,394
Location: Glastonbury, CT

Was thanked: 2 time(s) in 2 post(s)
here are a few charts showing your system

UserPostedImage

UserPostedImage

one chart is your system only long positions over a 12 year period

the second chart is your system but using a trend filter

notice how the trend filter reduces your risk and improves your return
Anton Beijer  
#5 Posted : Monday, February 14, 2011 2:27:44 PM(UTC)
Anton Beijer

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/16/2006(UTC)
Posts: 39

Nice improvement !

Which trend filter (or how??) do you use ???

Anton


henry1224  
#6 Posted : Monday, February 14, 2011 6:13:08 PM(UTC)
henry1224

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 10/29/2004(UTC)
Posts: 1,394
Location: Glastonbury, CT

Was thanked: 2 time(s) in 2 post(s)
the MTF Filter comes with the VST Pro add-on from Metastocktools.com

it is based on the Sp500 index
mstt  
#7 Posted : Monday, February 14, 2011 8:33:59 PM(UTC)
mstt

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 7/25/2005(UTC)
Posts: 1,042

Was thanked: 57 time(s) in 54 post(s)
Hi Henry Is the first chart correct? The equity curve and the under water histogram both give the appearance of being always in a trade, and yet both charts are marked as Long only. The EST and Trade Equity both show significant periods out of a trade, and quite different equity plots that aren't explained just by the different generation methods used by VST Pro. Thanks Roy
jjstein  
#8 Posted : Monday, February 14, 2011 10:40:06 PM(UTC)
jjstein

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 5/13/2005(UTC)
Posts: 715
Location: Midwest, USA

Was thanked: 1 time(s) in 1 post(s)
Henry: It looks like your VSTmod indicator is showing signals a bit too early -- is the XTEND code missing at the bottom?

Roy: Here's the same chart, with the bottom two windows showing INDIVIDUAL trade profit (green/top) and risk (red/bottom), rather than CUMULATIVE:


UserPostedImage


--Johnathan

P.S. Is there an EASY way to insert an image to a message? The thing doesn't seem to fit right when I used the IMG thingy, so here's the full image: http://i.imgur.com/9NfmH.png
jjstein  
#9 Posted : Monday, February 14, 2011 10:47:00 PM(UTC)
jjstein

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 5/13/2005(UTC)
Posts: 715
Location: Midwest, USA

Was thanked: 1 time(s) in 1 post(s)
Oops, guess I shoulda used PrtScrn instead of SaveAs HTML.

UserPostedImage




Anton Beijer  
#10 Posted : Tuesday, February 15, 2011 1:40:51 AM(UTC)
Anton Beijer

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/16/2006(UTC)
Posts: 39

Thanks to you both for your colorful input.

I am always a bit suspicious with a historically tailored system to one situation/exchange because the risk with such filters is that you are backtesting with a certain historical insider information built in that filter. The same reason ZIG-ZAG based systems produce fantastic results when tested backwards.

Thats why I put this topic here: it's difficult to get really unbiased system that don't need excuses, conditions, or luck/specified future developments to be profitable on the long run.

Have you tested multiple stocks from other (non USA) stock exchanges with that SP500 tailored filter?


henry1224  
#11 Posted : Tuesday, February 15, 2011 7:45:21 AM(UTC)
henry1224

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 10/29/2004(UTC)
Posts: 1,394
Location: Glastonbury, CT

Was thanked: 2 time(s) in 2 post(s)
Johnathan, here is the code I used

{* User inputs *}
pds1:=Input("HHV periods",1,260,21);
pds2:=Input("MA periods",1,260,70);
pds3:=Input("MO periods",1,260,10);
Sig:=If( Mo(Mov(HHV(C,pds1),pds2,E),pds3)>100,1,
If(Mo(Mov(HHV(C,pds1),pds2,E),pds3)<100,-1,0));
rev:=Input("Reverse trade signals? [1]Yes [0]No",0,1,0);

{* Entry conditions *}
entry:= {<- Do not change entry parameter name}
Sig=1 AND Ref(Sig,-1)<>1;

{* Exit conditions *}
exit:= {<- Do not change exit parameter name}
Sig<>1 AND Ref(Sig,-1)=1;
jjstein  
#12 Posted : Tuesday, February 15, 2011 9:06:03 AM(UTC)
jjstein

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 5/13/2005(UTC)
Posts: 715
Location: Midwest, USA

Was thanked: 1 time(s) in 1 post(s)
Henry -- I get IDENTICAL results with:

entry:= {<-- Do not change entry parameter name}
Mo(Mov(HHV(C,21),70,E),10)>100;

exit:= {<-- Do not change exit parameter name}
Mo(Mov(HHV(C,21),70,E),10)<100;

I just happened to note that there is a signal very near the beginning of your chart, which shouldn't be possible, given the 70 period lead-time.


jjstein  
#13 Posted : Tuesday, February 15, 2011 10:42:07 AM(UTC)
jjstein

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 5/13/2005(UTC)
Posts: 715
Location: Midwest, USA

Was thanked: 1 time(s) in 1 post(s)
Anton,

Needed to stretch my mind this morning, so here are the median Profit/Risk ratios of a few test explorations on the S&P 500 -- with & without Market Trend Filter (MTF) and Stops (2.5 ATR Initial, 3.5 Trailing), 3000 bars (12 years).

--Strategy (Momentum-Anton) Only: 1.4
--w/MTF: 2.3
--w/Stops: 1.0
--w/both: 1.6

MTF only: 3.4
MTF w/Stops: 3.1


Anton Beijer  
#14 Posted : Wednesday, February 16, 2011 12:20:47 PM(UTC)
Anton Beijer

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/16/2006(UTC)
Posts: 39

Ok, stops does not add extra profit. That is what I found out also on most systems. It seems that the variability and unpredictability is too great : what you gain by exiting on time is lost by entering too late.

Have you tested the MTF on let's say Hang Seng index, or DAX or Russian RTS ? because if it's a filter that filters a fundamental predictable trend it shoudl work there also. If it's historically tailored to the S&P500 it will work on all S&P stocks, because those will at large follow the S&P500 WHEN TESTED BACKWARD.

What we want are systems that work forward ! We can't make money with past trades, can't we?

At least my systems works 40% better then buy and hold, and works only on the long side. Standing on the side line in bearish situation is not always a bad thing: you sleep well and at least you get some interest while sleeping. Adding a good short system could improve the system.

Anyone with an open system with a profit/risk system better then 1.4 ? (The MTF is fine but see the remarks above and see the remarks regarding the ZIG ZAG indicator)



henry1224  
#15 Posted : Thursday, February 17, 2011 6:08:12 AM(UTC)
henry1224

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 10/29/2004(UTC)
Posts: 1,394
Location: Glastonbury, CT

Was thanked: 2 time(s) in 2 post(s)
Here are 3 corrected charts of UTX using your system.

There are 3000 periods loaded

UserPostedImage

this chart is only using the MTF Filter no system signals

UserPostedImage

This chart is your system with the MTF Filter

UserPostedImage

This chart is only using your system

here is the code for the above system

pds1:=Input("HHV periods",1,260,21);
pds2:=Input("MA periods",1,260,70);
pds3:=Input("MO periods",1,260,10);
Sig:=If( Mo(Mov(HHV(C,pds1),pds2,E),pds3)>100,1,
If(Mo(Mov(HHV(C,pds1),pds2,E),pds3)<100,-1,0));
rev:=Input("Reverse trade signals? [1]Yes [0]No",0,1,0);

{* Entry conditions *}
entry:= {<- Do not change entry parameter name}
Sig=1 AND Ref(Sig,-1)<>1;

{* Exit conditions *}
exit:= {<- Do not change exit parameter name}
Sig<>1 AND Ref(Sig,-1)=1;

jjstein  
#16 Posted : Thursday, February 17, 2011 9:24:09 AM(UTC)
jjstein

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 5/13/2005(UTC)
Posts: 715
Location: Midwest, USA

Was thanked: 1 time(s) in 1 post(s)
Henry,

That middle chart doesn't look right -- the MTF filter should improve results, not worsen them. Any chance you forgot to refresh?


jjstein  
#17 Posted : Thursday, February 17, 2011 2:19:31 PM(UTC)
jjstein

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 5/13/2005(UTC)
Posts: 715
Location: Midwest, USA

Was thanked: 1 time(s) in 1 post(s)
>>Have you tested the MTF on let's say Hang Seng index, or DAX or Russian RTS because if it's a filter that filters a fundamental predictable trend it shoudl work there also.

I'd agree, but I don't have data for the individual securities which make up the indices. I suspect the correlation of growth w/U.S. indices is high, though. And, as some have pointed out, panic tends to increase all correlations to 1 <g>.


>>If it's historically tailored to the S&P500 it will work on all S&P stocks, because those will at large follow the S&P500 WHEN TESTED BACKWARD.

Um, the word "Filter" in "Market Trend Filter" might not be the best choice of words -- perhaps a better term would be "Market Trend Supervisor" or "Market Trend Override".

The MTF isn't "tailored", it merely follows the trend of a market average. You could use a variety of methods/averages to do that, and use it to "supervise" or "override" your specific trading strategy/signals.

For example, Dow Theory has been used that way for 100+ years: Be in or out (long or short) of individual stocks -- or in more recent decades, the whole market, via VFINX or SPY.

FWIW, I've been following the MTF-only (highest P/R ratio, trend and stops) over at the main site for awhile: Testing forward -- or "Walk Forward", if you will -- which confirms similar results with Equis/Reuters data.


>>I am always a bit suspicious with a historically tailored system to one situation/exchange because the risk with such filters is that you are backtesting with a certain historical insider information built in that filter. The same reason ZIG-ZAG based systems produce fantastic results when tested backwards.

As above, "filter" was not the best choice of terms: ZIGZAG is a mathematical technique/data filter, while the MTF is hard math -- it will NEVER change signals. Comparing the two would be "apples vs. oranges".

Hope that helps,

henry1224  
#18 Posted : Friday, February 18, 2011 5:35:48 AM(UTC)
henry1224

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 10/29/2004(UTC)
Posts: 1,394
Location: Glastonbury, CT

Was thanked: 2 time(s) in 2 post(s)
Johnathan, the second chart is correct,I refreshed the chart after I adjusted the inputs.

Lets look at the chart, the first four years were a sideways market,so the MTF was choppy and the indicator not correlated and the system incurred losses

Years 2003 thru 2007 are a bull market, The MTF was still choppy at the start but the indicator and MTF were correlated towards the end and made money.

Years 2007 thru 2009 were a bear market and the MTF kept the System out of the Market.

Years 2009 thru 2011 are the start of a new bull market, the indicator will start to make new highs
and the MTF will correlate with the Indicator towards the end of the bull run and will again make profits,

The question is if the profits will justify the risk involved?
jjstein  
#19 Posted : Friday, February 18, 2011 9:49:12 AM(UTC)
jjstein

Rank: Advanced Member

Groups: Registered, Registered Users, Subscribers
Joined: 5/13/2005(UTC)
Posts: 715
Location: Midwest, USA

Was thanked: 1 time(s) in 1 post(s)
Henry -- Your middle chart is what bugs me -- Anton's system w/MTF. It doesn't show profit, other than a very small amount in 2000 & 2001. Look at the posted chart.

But, adding the MTF should IMPROVE things, but the middle chart is WORSE than the bottom one -- it shows cumulative loss!

FWIW, I get the same results for your code or Anton's, when used without MTF, but there is a significant difference between them when MTF is active. Both however, show profit, as opposed to the loss in your middle chart, below.



VSTmod-L01 - Momentum - Anton

entry:= {<-- Do not change entry parameter name}
Mo(Mov(HHV(C,21),70,E),10)>100;

exit:= {<-- Do not change exit parameter name}
Mo(Mov(HHV(C,21),70,E),10)<100;

{ ----- *** End of user-edit section *** ----- }
UserPostedImage




VSTmod-L02 - Momentum - Henry

{* User inputs *}
pds1:=Input("HHV periods",1,260,21);
pds2:=Input("MA periods",1,260,70);
pds3:=Input("MO periods",1,260,10);
Sig:=If( Mo(Mov(HHV(C,pds1),pds2,E),pds3)>100,1,
If(Mo(Mov(HHV(C,pds1),pds2,E),pds3)<100,-1,0));
rev:=Input("Reverse trade signals? [1]Yes [0]No",0,1,0);

{* Entry conditions *}
entry:= {<- Do not change entry parameter name}
Sig=1 AND Ref(Sig,-1)<>1;

{* Exit conditions *}
exit:= {<- Do not change exit parameter name}
Sig<>1 AND Ref(Sig,-1)=1;

{ -------- End of user-edit section -------- }
UserPostedImage
Anton Beijer  
#20 Posted : Sunday, February 20, 2011 3:07:39 PM(UTC)
Anton Beijer

Rank: Advanced Member

Groups: Registered, Registered Users
Joined: 4/16/2006(UTC)
Posts: 39

Do you have any insights into the trend filter system formula's or is it a black box/protected system ?

And of course not unimportant: any system that is more profitable then my system, even when testing multiple securities over multiple periodes?
Users browsing this topic
3 Pages123>
Forum Jump  
You cannot post new topics in this forum.
You cannot reply to topics in this forum.
You cannot delete your posts in this forum.
You cannot edit your posts in this forum.
You cannot create polls in this forum.
You cannot vote in polls in this forum.