PureBytes Links
Trading Reference Links
|
I've been using ICE 2.0 for several weeks. It creates the absolute
best expert system equations I've seen. It takes all four opted
indicators and combines them into one equation that in most cases is
really unbelievably good.
That being said, users have to learn to use ICE 2.0 effectively. I
can create long only or short only equations without any problem. On
some symbols like SPY or QQQ, I optimize long first and then short.
Once I have the optimal long equations and optimal short equations
and I've backtested them in-sample and out-of-sample I combine them
together into one expert. I use a set of one size fits all screens
like the ones from Elder, Pring or Clayburg. I have a custom ADX
equation that I use to create highlights and a set of ribbon
equations for scaling into trades that are based on fib numbers. I
also have a custom volume indicator.
When the ICE 2.0 equations long only and short only are combined with
the other stuff, it produces better trading results than I've seen on
any other system, and I've studied several of them.
I've never seen a better equation writer than the one in ICE 2.0.
I did run into some of the problems you mentioned. Most were caused
by not having enough bars on the chart, optimizing with the wrong
settings and not picking the right set of 4 optimized indicators.
Even if a particular indicator yields the highest total in one of the
4 V categories that doesn't mean it's the best indicator for use in
the combination of 4 indicators. It's an experience thing, but it can
be learned fairly quickly with trial and error.
I've learned how to create equations without having to have a volume
indicator for situations where volume is not a good indicator.
Adaptick tells me they've got a slight modification that I can use to
run experts on mutual funds which only have closing price
informaiton. I'm waiting on that modification.
For QQQ, SPY, IWM, DIA and MDY, I haven't seen anything better. The
hourly, daily and weekly equations are second to none. It's difficult
to get enough bars from esignal to do opts in hourly or lower time
frames. I've collected the data for about a year from the data on
demand cache file and combined it into one file with a year of
history. I've been waiting on esignal for two years to supply
intraday history longer than two months--they say they're in the
process of compiling the tick data, but so far not much. I've found a
source of tick data history, but it's about $50 a symbol or $36,000
for all the symbols--not very affordable for making intraday experts.
I've optimized a group of Dow symbols and some from the S&P 500
picked at random. On about 75% of the symbols, the results were as
good as any thing possible with any trading system.
I optimized HOV and then went over the entire history with a firm
that supplies quant data to mutual funds. They've been doing research
on a few symbols using TA with their quant models, and they agree
that the signals ICE 2.0 created for HOV were some of the best that
they had seen. They picked HOV and asked me to run it for them. One
of the executives had read about ICE, did TA using quant screening
and was curious how it would work. The choice of HOV was not mine.
Once again I combined it with my custom stuff, and the results were
almost unbelievable.
Depending on how you structure the systems tester, it takes me
anywhere from 2 hours to 4 hours to create the experts. It's really a
no brainer. I used to spend many hours writing equations with filters
and many, many tricks to get them to work well on a chart. ICE 2.0
gives me far better equations with little or no work. It's amazing.
Basically, I think it's going to give the newbie's a huge leg up on
getting a system to trade with without having to write one equation.
I also think when the results get out from people who are trading off
of systems generated with ICE 2.0, it's going to increase the sales
of MS.
It took me around 20 hours to figure out the best set of conditions
to use with it and how to get it to give me the results I wanted.
That's not bad. It beats every other MS plug-in hands down.
I just started trading the SPY and DIA with ICE experts on my charts.
It's too soon to tell how the experts are going to impact my total
P&L but adding the ICE experts to my previous system has increased my
confidence in the signals I'm seeing. It took me a couple of years
and a lot of trail and error to develop my original systems. It took
me 6 hours to create a system with similar backtest results using
ICE. What a difference!
I agree with you that some people shouldn't use ICE because they just
won't get it. The manual is very basic and can't cover every
contingency like the correct number of bars on your charts. Every
expert has a different requirement. If you run the trade verifier and
the have the system tester show you the signals on the chart where
trades were made, you'll see all of the siganls are there. That's a
prime indication the problem is your chart setup. Once you have the
right chart setup all of the signals will appear where they're
suppose to be.
Every user has a different setup regarding MS and their system. You
have to get all that insync with ICE to get the best results
including setting up the systems tester. I've found that points only
tests with no commisions and slippage work best in general. I know
that sounds counter to what everyone learns about systems tests but
you are creating an expert based on 4 optimal indicators and an
optimal stop loss. I've found that the less restrictions you put on
the system the better the weighting works which produces a better
equation across various markets. Other people may get different
results and I'm sure it will be the subject of much debate.
The more time you spend the better the equations you can generate.
For example, if you optimize over three different time periods and
find the 4V indicator that on average works best on all three
periods, you'll get an expert that produces the best results in any
market conditions. However, that kind of optimization takes 12 or
more hours per expert per symbol. I've found that I don't need that
precision because I use the other stuff I mentioned that validates
the signals and almost totally eliminates bad trades. (every system
makes losing trades)
I've also heard from early ICE users that you can create experts for
groups of stocks. That baloney. Of course you can optimize the
indicators over a group of stocks. Any group. However, like all
systems that work well with a group of stocks, it's not going to be
as precise as one that's optimized for only one stock. In addition
the expert won't work equally as well on all stocks.
Another trick I've found is once I optimize a symbol like GM, I run
the exploration equation ICE creates on a pool of 1700 high volume
stocks that I sometime trade from. Each expert will turn out to work
well on 20 to 50 stocks from that pool. I screen those to make sure
there are no out lying data like huge gap days or unadjusted splits
that cause any expert to work well on that symbol. Once that's done,
I have 10 to 25 symbols where that particular expert produces results
well in excess of buy and hold.
The point is there are many, many ways to use ICE and most of those
applications aren't in the manual.
If I were buying MS today, I wouldn't buy it without ICE. If I have
one complaint about EOD versus intraday MS, it's the lack of data
history for intraday symbols. Without that history, it's very hard to
make reliable experts and backtest them properly. If you're going to
sell an intraday version of a program and it's shell program like MS
where you have to develop your own trading systems before it has
value, at least the history should be there to develop the system
with.
I've enstalled the 9.0 upgrade and it's so, so. There's one or two
things in it I might use. I already had the sector stuff and the
broad market indicators that I had programmed myself. I guess for a
lot of people this will be something new. However, I use Rydex fund
flows more than any of that other stuff. MS should have had sectors
in it since day one so that's a little late.
Point & figure charts might trip a few triggers but not many. Most of
the aggravating bugs are still in MS. I could write a list a couple
of pages long about problems they know they have and haven't fixed.
This upgrade should have been version 8.5 instead of 9.0--my opinion.
I've asked them a dozen times for a bug fix list for 9.0, which I got
for 8.01. Of course, there's stillno list, so that's an indication
that not many bugs got fixed.
The biggest news is ICE 2.0. I used 1.0 but had to write my own
equations. What a fantastic addition. I think it should have been ICE
4.0 given the huge difference between 1.0 and 2.0.
I like MS more than the other TA programs out there. Equis has the
best customer support in the business and now they've got the best
system development tool in ICE. I wish they would spend more time
tuning up MS. They could easily match Trade Sim for systems testing.
With ICE MS needs a queing method for systems test, or the ability to
select no averaging so you could run a group of symbols all at once
and get the optimal results for each one instead of the group as a
whole. ICE 2.0 is going to put pressure on Equis to provide this
feature and the sooner they get it done the better.
Okay, that's my 2 cents worth in 10,000 words or less. I don't care
if you use ICE or you don't. I don't care if you can trade quarters
with your grandmother, and I sure don't care if you go broke, so
don't send me a bunch of emails (or posts) with all the parnoid
conspiracy theories or whinning and moaning about whatever. I have no
idea where you can find something like ICE for free to go with your
FREE data. It makes my trading systems better and that's all I care
about. It might make yours better and it might not. Equis gives a
money back guarantee and there are no yearly fees for the equation
generation--at least not yet.
I'd make a significant bet that this plug-in is going to out sell
every other MS plug-in by 2 or 3 to one. I also guess that next year
or the year after it will show up in S&C as the plug-in of the year.
Who knows--maybe I'm the only one that thinks that! Oh, well.
--- In equismetastock@xxxxxxxxxxxxxxx, "metastockuser"
<metastockuser@xxxx> wrote:
> I installed the pre release of 2.0 and was somewhat sorry I did.
> Numerous issues.
>
> As far as I can tell you can't get it to make a long only system or
a
> short only system. The final product has to be optimized long and
> short. When I try a long only system the buys show up visually on
the
> expert system but not the sells.
>
> I don't particulary like the results of the optimized long and short
> systems. I prefer making long only and short only systems.
>
> I'm not clear on what the final product is going to do or when it's
> coming out. A certain amount of frustration here as the older
version
> was a pain for me as far as working out forulas but at least I could
> use it to make a good long trading system.
Yahoo! Groups Links
<*> To visit your group on the web, go to:
http://groups.yahoo.com/group/equismetastock/
<*> 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/
|