|
PureBytes Links
Trading Reference Links
|
Scott, If you are going to confine your trading to a specific
universe, say the 217 symbols, then it is reasonable to run the
exploration tests on only those symbols.
You don't need to optimize again for about a year. However, the most
important aspect of making money is knowing how the system performs
under the different market conditions.
Many systems test very well because they perform good in up trends.
When you remove the up trend the system is a disaster.
You should pick historical periods for the explorations where the
overall market was in an up trend, down trend, price
consolidation/small pull back and price consolidation/pull back with
rotation. The latter is the most difficult to trade.
Most of the time those historical periods should last at least a month
for them to be long enough to influence the performance of your
system. Run explorations on 10 days out of the period and see how many
stocks the system finds each day on average, define your trading
rules, open the charts, apply your trading rules and see how many of
the candidates you would actually have bought. (If you want to see how
it works mechanicall, buy them all.)
Check the buys and figure out how many made money and how much money
they made. It's very difficult for someone to look at chart history
and not pick and choose trading rules that match the performance of
the stock. You should write down the trading rules and force yourself
to apply them exactly.
If you keep good history, you should see all of your statistics
change, often dramatically, in the different market conditions. You
may find that you only want to trade your system in up trends. Okay,
no problem. Learn to define an up trend and trade away.
There is no such thing as a system that works in all market
conditions. There are no specific trading rules that work in all
market conditions. There is no one psychological set of circumstances
that exist in all market conditions. The systems tester won't define
how your system and you react to these things.
The MS systems tester is really poor at calculating drawdowns. Most
people would be shocked to learn how many losers in a row their
winning system is going to encounter. In addition, most systems,
especially those written by inexperienced people, find far too many
trades for one human being with a reasonable capital account to take.
So how many do you take, and how much does your picking and chosing
impact the systems test results? If you don't take all the trades the
tester recommends, then you aren't going to achieve the results the
system tester did.
The average trader can not make 2500 trades a year. They don't
multiply the average trades per symbol by the number of symbols
they've tested to see how many trades they would have to make a year
to trade on a mechanical basis. On an EOD basis, a trader has to make
10 trades per day every day to hit 2500 a year. Ain't going to happen.
There are all kinds of other issues, too many to discuss in a post.
Most system tester results aren't worth the paper they're written on
because they have nothing to do with real life. Very few people can
trade a mechnaical system totally mechanically. The drawdowns are too
large, the trades can't be executed automatically without human
involvement and the normal trading account would be churned too much.
Individuals are poor choices for mechanical trading. Hedge funds and
other money managers are good choices for mechanical trading. How many
individuals could watch 50% or more of their trading account vanish
and keep believing in the system and keep trading it?
It takes a number of years of experience to put together a system that
has good numbers and is also rational from the perspective of being
able to trade it in live trading. The vast majority of traders would
have made more money using the buy and hold techniques defined in The
Four Pillars of Investing than they would have from trading some
system they created in a system tester or got from a guru.
I'm not even sure a systems tester should come with a program like MS.
I often think you shouldn't be able to buy the tester module until you
have 3 or more years of trading experience, and then I might have a
trader demonstrate they can make at least 5 successful trades using a
moving average--sort of like a driver's license test.
Have fun with the explorer!
--- In Metastockusers@xxxxxxxxxxxxxxx, "Scooter" <mariani@xxxx> wrote:
> Super,
> I re-ran the system test on 10 years of data again (Thanks again
Dusant). I
> ran them from Jan 1 to Dec 31 of each year. Out of the 800+ symbols
I have
> in my group I narrowed the list to 467 that had history back to 1995. I
> uploaded the spreadsheet up to the files area.
>
> In the 1995-2004 Worksheet you will find my summary. On the far right I
> calculated the winning percentages for each symbol. In the bottom
right cell
> I calculate the number of symbols that had a 80% or better hit rate.
I came
> up with 218 securities. I will re-run my criteria on those 218 only and
> report my results.
>
> Do you think it is reasonable to run the explorer only on the 218 best
> candidates? I would guess that I would re-run this exercise every
quarter or
> so to keep the symbols fresh.
>
> Scott
>
>
> -----Original Message-----
> From: Metastockusers@xxxxxxxxxxxxxxx
> [mailto:Metastockusers@xxxxxxxxxxxxxxx]On Behalf Of superfragalist
> Sent: Friday, April 15, 2005 1:34 AM
> To: Metastockusers@xxxxxxxxxxxxxxx
> Subject: [Metastockusers] System Tester Results Comparison with Scott's
> system
>
>
>
> Since we're having fun with numbers, I ran a comparison of part of a
> system I saw that Roy is publishing in his newsletter next month
> against the system results Scott published for us. Scott's results
> look good.
>
> I used Scott's criteria, $20,000 in equity, 100% in. All trades are on
> the open after the signal so that's a one bar delay. (Scott didn't
> tell me what he was doing with the entry/exit.) Roy's is a long only
> system. I don't know if Scott's is long only or both long and short. I
> ran Roy's on the S&P 500 and did not eliminate losing stocks. Oh, I
> almost forgot. I didn't optimize it. Scott's is optimized.
>
> Scott's dates are 4/13/xxxx to 4/13/xxxy Normally I won't test across
> skewed dates, but I wanted good comparisons so I used Scott's dates.
>
> Roy's Scott's
>
> 2004/5 $500.965 $139,048
> 2003/4 $613,302 $234,187
> 2002/3 $17,897 $106,714
> 2001/2 $198,010 $158,116
> 2000/1 $75,173 $(-29,693)
> 1999/0 $219,871 $290,114
>
> If you guys are interested in seeing how this works, get a copy of
> Roy's next newsletter. The code for this part of the system will be in
> there. Since Roy is having other trader's write articles, maybe Scott
> would be willing to share his system in Roy's newsletter also. I would
> read about it, and play with it.
>
>
>
>
>
>
>
>
>
>
>
----------------------------------------------------------------------------
> --
> Yahoo! Groups Links
>
> a.. To visit your group on the web, go to:
> http://groups.yahoo.com/group/Metastockusers/
>
> b.. To unsubscribe from this group, send an email to:
> Metastockusers-unsubscribe@xxxxxxxxxxxxxxx
>
> c.. Your use of Yahoo! Groups is subject to the Yahoo! Terms of
Service.
------------------------ Yahoo! Groups Sponsor --------------------~-->
Has someone you know been affected by illness or disease?
Network for Good is THE place to support health awareness efforts!
http://us.click.yahoo.com/Rcy2bD/UOnJAA/cosFAA/zMEolB/TM
--------------------------------------------------------------------~->
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/Metastockusers/
<*> To unsubscribe from this group, send an email to:
Metastockusers-unsubscribe@xxxxxxxxxxxxxxx
<*> Your use of Yahoo! Groups is subject to:
http://docs.yahoo.com/info/terms/
|