Automating the Hershey System
James Fowler & Harla Yesner
Okay here is my understanding:
1. There are five distinct trading modes:
a) End effects (beginning of day)
b) AM trading (until 11:15 EST)
c) mid-day
d) PM trading
e) End effects (end of day)
Questions:
Are these fixed (e.g. b ends at 11:15) or are there signals for when
they begin/end? For example, does the third critical point end a)?
How do these modes differ from A-E considerations, and how do we know
when the mode takes precedence over the signal and vice versa?
2. Before sync (spot-futures index, MACD sign, signal, and equilibrium
differences all reach some fixed tolerance) the program does not allow
trades.
3. Somehow the progam discerns the "sentiment" of the market.
Questions:
Is sentiment determined by the cell of the Hershey matrix that uses EOD
and 30 min fractals? If not, how?
4. The program has a lookup table to determine the optimal value for
away values of the 1min MACD.
Question:
Am I looking at sentiment to determine which cell I am at in this lookup
table? If not, what?
5. The program has a lookup table to determine the optimal frequency to
update stops.
Question:
Am I looking at sentiment to determine which cell I am at in this lookup
table? If not, what?
6. The program discerns the first TA formation based on the first three
critical points.
Do you use the 1min MACD to determine the first three critical points?
If not, how?
7. If the third critical point is less extreme than the first, a market
order is sent (buy if a higher low, sell if a lower high).
Question:
If not, should we send a market order on the fourth critical point?
(sell on the high after a lower low, buy on the low after a higher high)
8. After a trade is initiated, stops are based on the right trend line
and the frequency with which we update them is based on the time
distance
between the troughs/peaks.
9. The scalp range added to the stop is based on another lookup table
which is based on the channel width, determination of trend, volatility,
and
absolute volume.
Question:
How exactly does this work? Is it a four dimensional matrix of some
kind? Or are all the lookup tables symmetric, with an appropriate
critical value, stop frequency, and scalp range in each cell of a single
matrix?
10. For simplicity, there are five possible settings in these lookup
tables (your p,q,r,s,t).
11. We change lookup tables when there is a pace change. ( "you would
have changed lookup tables to tighten stops
because the second day there was a pace incerease in the market." )
Question:
Huh? Now I am really confused. Do you mean that the pace change sends
us to a different cell in the same lookup table?
12. On rare occasions the program will hold an overnight postion. This
will be determined by sentiment.
Question:
Can you generalize this condition a bit more? I am unclear about what
variables we are looking to in order to make this determination.
13. Volatility compression and climax run routines will be used to
optimize exits.
Comment: We need to talk about these more in detail, but we should get
the other stuff down first.
Thanks Jack!
------------------------------
James Fowler
Littauer Center
Harvard University
Cambridge, MA 02138
James Fowler & Harla Yesner
Here is the circumstances that dictated my response. Every day you start
out with the "end effects" nature of the market. As sync occurs you are now
turned loose to appraise the day. You are as a first order of business
looking for the TA formation that begins the day and you particularly
regard the slower fractals guiding the position traders and finally the
sentiment of the market. All of these items form trend containers for each
other in turn. Our intraday focus allows us the possibility of a slow paced
positon multiday trend (rarely) to a tight low volume low volatility
Greenspan congestion. On the 15th an entry after sync appeared tight off
thus dictating an exit at the next peak if nothing else enterred the
picture. See my snip on how the day was unfolding to set up the right line
on the 5 min bar. The exit out of the first trade was on the decond
formation data point and it helped to define a potential channel tipped at
the angle to next be determined by the up coming trough. If the trough were
lower than the first point we would have a short trend and the right line
would be drawn parallel to the two point line using the single point at
10:00. As it turned out the trough was higher so a long trend was
discernable using the first and third points. Now gerome see the pace is
slow and he goes in on the trough setting a trailing stop below the scalping
( your soft ware program auto matically picks off the three points and sets
up an offset trailing stop sequence updatable at the frequency of estimated
trough time separations the buy order you automatically send to the broker
is a "market" order in response to the trend being established. Prior to
this your software was successsively passing through setting up sync by
getting the INDU/DJ00M spread to a constant as well as the MACD on both INDU
and DJ00M at same sign, value and equilibrium MLR. the second thing that
happened software wise was that you set the days volatility tolerance range
and tweeked the 10 point away to a good value like 8.3 from a look up table
and the 7 1/2 min nominal to 7.13 from a coresponding look up table. Your
first order to the broker would have bettered jermone's by two minutes and
7 points earlier. The software would have been tuned to "end effects
(beginning) EXIT signal" subroutine as the formation package was getting
data the cancel the "end effects (beginning) EXIT signal" modus either one
was going to prevail up to the point that the third data point arose. It
turned out quite nicely that the peak was a double peak that allow the
broker "sell market" mesage you would have sent on the peak that came in
prior to the third point to hit as the second peak was in play all normal
timing considered.
Your software would have kept you out for the short that ensued ( didn't
trigger because of insufficient "away" value on MACD and time out on the
time criteria. and it would have traiggered a long market entry on the way
out of the trough just after the end effects priority was shut down and you
went into the "AM trading mode" that is always available daily up to 11:15
when it is stopped or over ridden by an A through E choice of what is going
on. The choice your software will make is that you are in a trend...its a
channel ...the volatility is low, volume strong and to go to scalper offset
p of p, q, r, s, or t, and finally up date the broker set stop at time
intervals of whatever the table look up tells it. the software would have
persisteed in this mode until you hit the "end effects(end of day)" mode and
the broker would have sent you a settlement notice to reset the software for
tomorrow with all the slow fractal up dates in place for the "sentimant"
cookbooks tables preselected. The next day you would have enetered on the
three point trend and dona baout the same for the day. this would have been
good for about 250 points the first day and 400 the second. you would have
been adding contracts to settelment both days as a subroutine effort and one
in the two days you would have changed lookup tables to tighten stops
because the second day there was a pace incerease in the market. in any
event you would have been driven to an exit as an endof day trade
settlement.
You can see by these notes that the FDates and WDates approach is missing a
few beats here and there.
Iwould add a volatilty compression routine in there and a climax run
subroutine as well; both of these will nicely take you over to the high side
of the channel mathematically and set you up for end of trend (channel)
early exits and not reversing. a soon as we get to a comprehensive use of 1
through 4 stuff and A through E stuff we will take alook at how the P,V
realtion allows you to optimize trading. If 30 points a day gets you 75
grand a year subsistence living, then it is important to think about what
working half days might yield (out west we think of the New York NYSE Day
as a half day enterprise that starts late and ends early..we fill our
afternoons with other things. Gary Smith loud and clearly says how to take
income form your work effort and put it to work effectively where it needs
little or rare adjustment as it beats the standards of the industry. We
want to stay focused on trading and software that appreciates a nice slug of
money as above as we add contracts and settle before we go to lunch. I got
into a tangle with Eddie Robots because I had to sideline on a Gary Smith
type investment. It took 31 transactions to exit 100,000 shares in the 25 to
30 range where the profits were only 25 to 35%. These kinds of problems can
occur as a direct consequence of doing the DJ00M kind of stuff manually.
When you slip into an automatic software approach that you will have soon,
then the days will be less ornery detail wise. But you will have to be
coaching brokers that constantly give you feedback that leads to their
further training.
this is the snip that was in my reply to Jerome's posting that lies between
your snips. It shows the on the other hand thing reasoning I did as it
occurred and finally allowed the day to proceed.
"If you had seen that using the first three points that determine the
channel of the trend 9:45, 10:00, and 10:20's. you would have rode that
single trend all day long"
I am changing my lingo a little to add in the software overlay that you will
want to produce. There is a point when the software construct manifests
just like the point you reach doing a xword puzzle. At that point you "just
do it" as nike says and it is right.
to connect to brokers the Streetsmart top version , i. e., Schwabb are
penetrateable. Unfortunately there is weak link in people at the top of the
software pile (and it is remote from operations as well)at Schwabb.
James Fowler & Harla Yesner
I reposted Jack's original message before his hiatus so you can see it. It is
to this message I was responding. Your answers are all correct for the 30
points a day plan, but I think Jack was beginning to broaden the picture
somewhat.
I am not quite comfortable with the 30 points a day plan because it lacks
precision on the exit strategy (or I have not yet fully understood it, anyway).
My main difficulty is with automating the number of stops to stay back. It
appears that "slow" markets may actually be more or less steep in price, so I am
having a hard time understanding exactly analytically what a slow market is so
that I can tell the program to stay 3 stops back (or 2 or 1 for medium and fast
markets). I am also having difficulty with objectively determining when to do a
market order at a trough. Finally, the reversal strategy is not completely
clear to me, since on sharp troughs sometimes the 1min line crosses over but the
price continues to drop (opposite logic for peaks). I guess you just scratch on
these when they hit the wrong stop, but I haven't worked it out in my mind quite
yet.
In the next 2-3 weeks I will be testing the 30 point system to see if
historically it produces enough good trades to justify the scratches. There are
probably some people here who have already tried it and had success, but I am
still gunshy. When I get some simulations in hand that are statistically
significant, then I will go to market.
"=(:b)illiam" wrote:
> Okay, we are talking about the same thing, and I see what you are trying to
> do - but I don't know of a "safe" way to do it. You'll probably encounter a
> coding challenge that I have been tinkering on for a while - how to develop
> an algorithm that plots a simple support or resistance line based on your
> "critical points" (an optimization similar to the Travelling Salesman
> problem).
>
> > 8. After a trade is initiated, stops are based on the right trend line
>
> Yeah, let me know when you have an automated way of drawing that trend line.
>
YES! That has been bugging me, too. I don't know if you saw my posts on
automated trend lines to get the "right" line that Jack always talks about. I
have a simple set up in Excel that I believe is getting me close to a
solution--let me know if you have time because I could send it to you and I
would appreciate some input.
Sorry this is so stream-of-consciousness.
James
James Fowler & Harla Yesner
> >Thanks Jack for the detailed first cut at an automated system.
> >
> >Okay here is my understanding:
> >
> >1. There are five distinct trading modes:
> >a) End effects (beginning of day)
> >b) AM trading (until 11:15 EST)
> >c) mid-day
> >d) PM trading
> >e) End effects (end of day)
> >
>
> Have any idea when this will be completed? LOL! My system does not discern
> the sentiment of the market which it could really use. Another drawdown
> like the last and --- let's not talk about it (arrrgh).
I'm sure a lot of people in MIF think it will *never* be completed!
If Jak gets back on line here, we may be able to pick his brain about how to do
this. For now, my hypothesis is that it has to do with the dual relationship
between an EOD MACD and a 30min bar MACD. E.g. maybe if the fast lines of both
are diverging, this indicates a bullish sentiment.
Making all this stuff objective is quite difficult. I really respect traders
who don't need any of this stuff--their subconscious just processes everything.
James
------------------------------
James Fowler
Department of Government
Don Cameron
<james_...@harvard.edu> wrote:
>
>In the next 2-3 weeks I will be testing the 30 point system to see if
>historically it produces enough good trades to justify the scratches.
A week or so ago I posted a message asking Jack what are the specific
criteria for deciding to scratch a 30-point trade. He seems to be
incommunicado these days. Perhaps you would like to give me your
interpretation of the criteria, as they would be essential to any
automation of his approach.
Troutman, Defender of Sticks
<james_...@harvard.edu> wrote:
>I'm sure a lot of people in MIF think it will *never* be completed!
James -
FYI, be careful of who you're responding to. If you look at the
poster's id, it's somebody spoofing me. It's no big deal, but worth
mentioning.
Sticks
--
Troutman, Defender of Sticks http://www.defendercapital.com/
Jonathan Matte, President trou...@defendercapital.com
Defender Capital Management, Inc. (IB, CTA)
There is a significant risk of loss in futures trading.
James Fowler & Harla Yesner
> On Fri, 07 Apr 2000 07:50:40 GMT, James Fowler & Harla Yesner
> <james_...@harvard.edu> wrote:
>
> >
> >In the next 2-3 weeks I will be testing the 30 point system to see if
> >historically it produces enough good trades to justify the scratches.
>
> A week or so ago I posted a message asking Jack what are the specific
> criteria for deciding to scratch a 30-point trade. He seems to be
> incommunicado these days. Perhaps you would like to give me your
> interpretation of the criteria, as they would be essential to any
> automation of his approach.
This is exactly my problem right now, as you can see from my last two
posts. I see the problem as follows. We know that entry occurs on a
crossover of the 1min MACD fast and slow lines, and it is confirmed by the
fast line crossing 0 within 7 and 1/2 minutes. But for exit we have two
options. We can either exit on a trough that nets us 30 points or more,
or we "hang loose". Jack clearly has expressed opinions about when each
is appropriate in specific situations, but as yet I have not been able to
generalize it. To "hang loose" means alternatively, a) staying 3 stops
back, b) staying beyond the right trend line (which I am also having
problems automating), and/or c) staying in the trade as long as the 5min
MACD fast line does not cross the slow line (or maybe 0).
The real problem I am having is that the way he describes reversing in and
out of trades requires that entry and exit guidelines be symmetric, which
clearly they are not except in special circumstances (like 1min MACD
absolute equilibrium values of 10 or more). But then we should be going
for the trough, and Jack specifically told me this analysis was wrong when
I asked him about one of his trade descriptions.
When he comes back on line we should really try to get him to nail it down
for us.
Gary Smith
Better yet, maybe John Lothian can give us some insight with Jack's
permission. Jack made an inquiry on this forum several weeks ago about
opening a trading account at John's firm to show how his methodology works
with real money.
Gary Smith
James Fowler & Harla Yesner <james_...@harvard.edu> wrote in article
<38EE29B9...@harvard.edu>...
Don Cameron
<james_...@harvard.edu> wrote:
Clearly there is lots of scope for misunderstanding - see below.
>Don Cameron wrote:
>
>> On Fri, 07 Apr 2000 07:50:40 GMT, James Fowler & Harla Yesner
>> <james_...@harvard.edu> wrote:
>>
>> >
>> >In the next 2-3 weeks I will be testing the 30 point system to see if
>> >historically it produces enough good trades to justify the scratches.
>>
>> A week or so ago I posted a message asking Jack what are the specific
>> criteria for deciding to scratch a 30-point trade. He seems to be
>> incommunicado these days. Perhaps you would like to give me your
>> interpretation of the criteria, as they would be essential to any
>> automation of his approach.
>
>This is exactly my problem right now, as you can see from my last two
>posts. I see the problem as follows. We know that entry occurs on a
>crossover of the 1min MACD fast and slow lines, and it is confirmed by the
>fast line crossing 0 within 7 and 1/2 minutes.
Not the way I see it. I think the clock starts when the 1 min MACD
peaks. These quotes are from Jack's posts.
"All you will do is take fast signal line "away"peaks/troughs on the 1
min bars to trade the trends of the DJ00H that appear neatly on the 5
min bars."
" Here is the emergence of the signal. After the peak/trough the fast
signal line curves to the "0" crossing the slow signal line on the
way. When it approaches the "0" axis, pick up the phone, place two
orders as stops, the first with the trend of the fast line, the second
the first listed trailing stop. By the time you have done both orders
the fast line will cross "0". "
> But for exit we have two
>options. We can either exit on a trough that nets us 30 points or more,
>or we "hang loose". Jack clearly has expressed opinions about when each
>is appropriate in specific situations, but as yet I have not been able to
>generalize it. To "hang loose" means alternatively, a) staying 3 stops
>back, b) staying beyond the right trend line (which I am also having
>problems automating), and/or c) staying in the trade as long as the 5min
>MACD fast line does not cross the slow line (or maybe 0).
Quotes bearing on this issue include:
"N.B. Each time there is no action in 7 1/2 minutes exit the market;
there is not trend for you now."
"If the trend fails, i. e., the 5-minute bar curves don't continue,
you will know from the lack of action on the 1-minute bars. "
"What is clear is how to scratch on entries that go to non-trends on
the peak/trough that follows that is "near" and not "away" ."
"it was a good day to see how the 7 1/2 minute time Out (TO) worked
and also to see that when you pulled the trigger to enter the fives
times possible you did exit within minutes on a subsequent peak to
complete a scratch trade. "
I take these comments to mean there is another alternative for
exiting, i.e. that if the trend stalls soon after the 1 min crosses
zero you exit on a near-zero 1 min peak. But I am not sure - hence my
question.
>
>The real problem I am having is that the way he describes reversing in and
>out of trades requires that entry and exit guidelines be symmetric, which
>clearly they are not except in special circumstances (like 1min MACD
>absolute equilibrium values of 10 or more). But then we should be going
>for the trough, and Jack specifically told me this analysis was wrong when
>I asked him about one of his trade descriptions.
>
It seems to me that in a number of posts he has used the term "C & R"
(cover and reverse) when he really means just cover to close the
position. I think C&R only applies to a far away peak after a
substantial move, when the heavily overbought or oversold condition
would suggest a bounce.
>When he comes back on line we should really try to get him to nail it down
>for us.
>
>James
>
I suspect that there are lots of nuances which are perhaps
unconciously applied based on experience and which would be very
difficult to automate.
vladimir
I think Your summarization is correct and simple to
understand. Do You think that it should be backtested the way
You posted it? I can see it tradable with addition of some
confirming(independent) indicator.Do You know somebody who could
backtest it?
Cordially,
Vladimir
* Sent from RemarQ http://www.remarq.com The Internet's Discussion Network *
The fastest and easiest way to search and participate in Usenet - Free!
Don Cameron
Is it the summary of his stock trading approach that I put on the
bulletin board or my comments on futures trading here?
Kate Ewald
many serious doubts. Real money is the only valid test, IMHO.
Kate
Gary Smith wrote:
> Better yet, maybe John Lothian can give us some insight with Jack's
> permission. Jack made an inquiry on this forum several weeks ago about
> opening a trading account at John's firm to show how his methodology works
> with real money.
>
> Gary Smith
>
> James Fowler & Harla Yesner <james_...@harvard.edu> wrote in article
> <38EE29B9...@harvard.edu>...
> > Don Cameron wrote:
> >
> > > On Fri, 07 Apr 2000 07:50:40 GMT, James Fowler & Harla Yesner
> > > <james_...@harvard.edu> wrote:
> > >
> > > >
> > > >In the next 2-3 weeks I will be testing the 30 point system to see if
> > > >historically it produces enough good trades to justify the scratches.
> > >
> > > A week or so ago I posted a message asking Jack what are the specific
> > > criteria for deciding to scratch a 30-point trade. He seems to be
> > > incommunicado these days. Perhaps you would like to give me your
> > > interpretation of the criteria, as they would be essential to any
> > > automation of his approach.
> >
> > This is exactly my problem right now, as you can see from my last two
> > posts. I see the problem as follows. We know that entry occurs on a
> > crossover of the 1min MACD fast and slow lines, and it is confirmed by
> the
> > fast line crossing 0 within 7 and 1/2 minutes. But for exit we have two
> > options. We can either exit on a trough that nets us 30 points or more,
> > or we "hang loose". Jack clearly has expressed opinions about when each
> > is appropriate in specific situations, but as yet I have not been able to
> > generalize it. To "hang loose" means alternatively, a) staying 3 stops
> > back, b) staying beyond the right trend line (which I am also having
> > problems automating), and/or c) staying in the trade as long as the 5min
> > MACD fast line does not cross the slow line (or maybe 0).
> >
> > The real problem I am having is that the way he describes reversing in
> and
> > out of trades requires that entry and exit guidelines be symmetric, which
> > clearly they are not except in special circumstances (like 1min MACD
> > absolute equilibrium values of 10 or more). But then we should be going
> > for the trough, and Jack specifically told me this analysis was wrong
> when
> > I asked him about one of his trade descriptions.
> >
> > When he comes back on line we should really try to get him to nail it
> down
> > for us.
> >
> > James
> >
vladimir
I was referring futures trading. I think somebody might backtest
it the way it is now,without "scratch trade rules" just to see
if it will produce decent percentage of profitable trades.
Don Cameron
able to make money trading the DJ futures based on his experience.
The secondary question is whether he does it according to a clear set
of quantifiable rules that could also be applied by others. This
question may never be answered as he may not want to reveal exact
details and who could blame him.
I may be completely wrong, but I wonder whether his attempts to
quantify his methods here are perhaps a simplification of what he does
intuitively based on his experience and may be an incomplete
description. This is not to suggest he is being misleading. Just the
contrary, I believe he is trying to be helpful. However, when I
review his writings I find inconsistencies which make me wonder
whether his rules have really been cast in stone for a long time or
are a latter-day attempt to transfer knowledge. I will give one
example.
When he describes his "divergence" type of trade he talks about the
fast and slow lines of the 1 and 5 minute MACD diverging and
steepening and says to jump into the trade when all four lines pass
through zero. However, unless I am doing something very wrong, in the
example he gives (Feb 23 just after "sync") of a divergence trade, the
move appears to be over by the time the 5-minute slow line crosses
zero. He may very well be able to spot a fast, strong momentum move
and trade it, but it seems to me that the rule he suggests to trade
such a move may be an after-the-fact effort to quantify his intuition
and in this instance, at least, potentially misleading.
Jack Hershey
Terrific thread. I apologize for being inattentive for so long; there are
reasons as you would suppose.
I have spent a lot of time reading and rereading the comments here. So, as
far as I am able, I will respond in four subparts to try to serve the
interests of all of you as selective readers so you don't have to sort
through stuff not of interest to you. The four sub groups are entitiled: I
the validity and transferance of the approach; II the thirteen points and
Q's; III the computer automated paradym for both stocks and commodities; and
IV additional thread comments.
Thanks to all of you who have been working with each other. Most
unforntunately I have had the experience of learning that the 911 system is
not integrated across the country and I had the difficult experience of
enduring the sounds of my fiancee being attacked by a drunken person over
the phone that was snatched from her while I was1700 miles away. She had
just taken up residence in our new home (to us) and we were saying good
evening. She is well again and everything is well under control. I
appreciate your kind thoughts.
regards
Jack
Jack Hershey
used for the stocks approach, i. e., a person that i haven't met should be
able to computerize the method, trade using the computer, and appreciate
capital to definitely determine transferance is possible. secondly if it is
done manually instead, I feel the transference is achieved and achieved more
broadly. both of these are done for stocks and only the manual is being done
for commodity futures by my standard. the commodities effort is not
sufficient duration wise to be significant by the standards of most well
written successful authors and practicitoners.
At this point, i have asked for several packets from John Lotharian to get
set up under NFA 2-8 (E) item (a) with him and we have not completed our
calls and mailings. Although he is getting some direct independant requests
as i have been informed by others. i have established an additional
residence as of this month in Illinois and my intended is there now so I am
commuting. So some additional real money tests will be under way soon with
John and i will drop into the chicago area from southern Illinois or St.
Louis occasionally.
I agree with all of you that validity is most easily measured with financial
success. I am hoping that the financial success part of it is really
transferable to either those who operate as amateurs individually and
manually as well as to those who are computerized to some extent. There are
ways to use this subjectively I believe; I hope this can show through in the
sub parts III and IV.
The sub thread of Kate and Gary nicely points out all of these concerns we
have about trading successfully.
Jack Hershey
The source of the thirteen comments is a post made by me on 20 MAR 00 in the
evening in response to James and Jerome. i appreciate the comments already
made on the points and I will try to post here any insights that need to be
made to clarify upon the additions as well.
Going below now.
"James Fowler & Harla Yesner" <james_...@harvard.edu> wrote in message
news:38ED5443...@harvard.edu...
> Thanks Jack for the detailed first cut at an automated system.
>
> Okay here is my understanding 9 and my parens ( ):
>
> 1. There are five distinct trading modes:
> a) End effects (beginning of day) (up to sync)
> b) AM trading (until 11:15 EST) (where hi/lo for day usually sets up),
(active P,V)
> c) mid-day (flat , drift period)
> d) PM trading (active P, V)
> e) End effects (end of day), (last 20 to 30 min)
>
> Questions:
> Are these fixed (e.g. b ends at 11:15) or are there signals for when
> they begin/end?
Nothing is fixed on this. A condition establishes the boundary.
a) ends with sync (See 2 below) usually by first 15 minutes.
b) ends with hi/lo set OR congestion commences on low spotty volume ( 3 to 5
trades a min on DJXX) I look for it starting around 11:00 EST.
c) ends with a BO (break out) The easiest trade of the day.
e. begins during last half hour typified by fast pace and erratic MACD. i
always sideline unless strong trend is overriding into next day.
For example, does the third critical point end a)?
Not really
> How do these modes differ from A-E considerations, and how do we know
> when the mode takes precedence over the signal and vice versa?
A-E are "trading modes" that i shall connect to the "market sentiment" to
simplify as much as possible. A-E occur during AM and PM trading. i will
try to redefine A-E based on signals that are used to trade A-E.
>
> 2. Before sync (spot-futures index, MACD sign, signal, and equilibrium
> differences all reach some fixed tolerance) the program does not allow
> trades.
No trading before sync. This is to eliminate unnecessary risk. CBOT opens
(8:20EST) on low volume and the index moves to take overnight stuff into
account. Many first market reports dwell on this action with S&P most often
mentioned. It is a weak reporting technique by lazy predictors. NYSE
opens at 9:30EST and index does two things: resetablishes the normal point
spread offset for the portion of the future index time left before rollover
and then they move together using this offset; and secondly, the INDU and
DJXX MACD's come into the same approximate value. This takes about 15
minutes. Now we can make some money ......
>
> 3. Somehow the progam discerns the "sentiment" of the market.
Yes. Wizzes and scalpers do this in many ways that work for them. you can
use a yellow underliner when you read stuff and you will see what I mean.
Noise is a typical item in old writeups. Sentiment will be defined on the
computer by: scalping; volatility; "pace"; entwining or not; and the value
of MACD and the slope of the trend and volume. this allows us to go to look
up tables for stuff. the hershey matrix for copmmodities indexes is used to
get the pace but we can omit its use in the computer program as a KISS
thing. See part III for how. for now use the A-E categories as sentinals.
the p, q, r, s and t as i mentioned it.
>
> Questions:
> Is sentiment determined by the cell of the Hershey matrix that uses EOD
> and 30 min fractals? If not, how?
See part III
>
> 4. The program has a lookup table to determine the optimal value for
> away values of the 1min MACD.
>
> Question:
> Am I looking at sentiment to determine which cell I am at in this lookup
>
> table? If not, what?
Yes, we need this to convert my hard +/- 10 MACD rule to something the
computer can adapt to as the volatility, congestion width and pace vary.
I think we can use MACD and volume. I am going to have a recommendation
that Price, MACD (2 signals), and volume be the only computer RT values
used. This reduces to P, V upon which the market relation is based. Some
hard assed computer buffs will defeat themselves yet again at this point.
>
> 5. The program has a lookup table to determine the optimal frequency to
> update stops.
>
> Question:
> Am I looking at sentiment to determine which cell I am at in this lookup
>
> table? If not, what?
Yes See 4. the "away" and the "optimal frequency of update" are mutually
dependant.
>
> 6. The program discerns the first TA formation based on the first three
> critical points.
>
> Do you use the 1min MACD to determine the first three critical points?
No.
> If not, how?
The trading trend (which is on 5 minute bars since we trade on the 5 min and
anticipate on the 1 minute) has a roughness caused by swings on the 1 min
bars. the first three 1 min moves do determine these 1st, 2nd and 3rd
points on the five minute bars. Pints 1 and 2 are on opposite sides of the
width (noise ) in the trend. point 3 is on the opposite side from point 2.
we can now set up the right side trend line. either through the odd points
or through the ewven point parallel to the odd two points line. Assume a
channel always and measure the "noise" as the channel width. the "noise"
includes the scalping fringe which is usually predominant on one side. We
need the noise level updated regularly and we need the math representation
of the line on hand as well slpe wise.
> 7. If the third critical point is less extreme than the first, a market
> order is sent (buy if a higher low, sell if a lower high).
I am not always going to wait this long to trade. The when to start a trade
isn't on the list but it will be taken care of.
> Question:
> If not, should we send a market order on the fourth critical point?
> (sell on the high after a lower low, buy on the low after a higher high)
See part III where its comprehensive and in one place.
>
> 8. After a trade is initiated, stops are based on the right trend line
> and the frequency with which we update them is based on the time
> distance
> between the troughs/peaks.
Yes very good rough understanding. Very important optimizing stuff involved
to max ROI.
>
> 9. The scalp range added to the stop is based on another lookup table
> which is based on the channel width, determination of trend, volatility,
>
> and
> absolute volume.
YES.
>
> Question:
> How exactly does this work?
See part III where I outline the computer program.
Is it a four dimensional matrix of some
> kind?
No we can keep it simple; bunch of two's.
Or are all the lookup tables symmetric, with an appropriate
> critical value, stop frequency, and scalp range in each cell of a single
>
> matrix?
YES, we can get a value of some sort, then we use it as the one dimension of
all the look up tables. Nice breakthrough we have diagnostically preserved
our options to tune the tables independantly by back testing. Again this is
where computer guys go complex and defeat themselves by intuitive human
intervention to try to cover their rears. Our job was just to see the
variables and to be able to use them to make money.
>
> 10. For simplicity, there are five possible settings in these lookup
> tables (your p,q,r,s,t).
Always a good place to start. We will go from discrete to continuous very
nicely soon. we won't need a lot to start and we can interpolate if
necessary to make it a continuous functionality relative to the market pace.
Once you can see the roiling effect of the indu, DJ, S&P's NAS, and russell
stuff and how it drives ticks, you get to KISS thinking. We will do it on
the computer focussed on a given index.
>
> 11. We change lookup tables when there is a pace change. ( "you would
> have changed lookup tables to tighten stops
> because the second day there was a pace incerease in the market." )
I would prefer to start with one table at first and build families later.
Laffler really screwed that stuff up vis a vis Reagen tax stuff; they were
on the wrong side of the max and didn't know it
>
> Question:
> Huh? Now I am really confused. Do you mean that the pace change sends
> us to a different cell in the same lookup table?
for now it does...you aren't confused It is easier. Use few look up tables
and use 9 c as the op pt change trigger. we are getting a little elegant
soon.
>
> 12. On rare occasions the program will hold an overnight postion. This
> will be determined by sentiment.
>
> Question:
> Can you generalize this condition a bit more? I am unclear about what
> variables we are looking to in order to make this determination.
Yes, cool understanding. Strong sentiment means slow steady enduring pace
(orderly market). Say like Greenspan is at Martha's Vineyard. We look at
the slower fractals thay are all "away" or have same sign and first
derivatives. Someone recently saw four fractals peak or trough
simultaneously here a while back (near end of a day i think); its that kind
of phenomena that wraps up a position trade over time. We are talking about
before that occurrance, i. e. , during the money making time over several
days.
>
> 13. Volatility compression and climax run routines will be used to
> optimize exits.
>
> Comment: We need to talk about these more in detail, but we should get
> the other stuff down first.
Okay, its your call. But i must say you have almost everything quantified
and qualified to a good extent. very shortly we will be able to express
what "subjectivity" really is with the KISS traders. we will notice also
there is quite a range for what is considered success in ROI. 13 actually
just deals with a few rapid shift phenomena that occur in the lookup arena.
Its second derivative stuff.
Look at the first hour of tuesday to see a little of it. That level of ROI
is from outer space by some standards.
The next one (part III) is good. I probably will ship a flow chart to Don
if I can to back it up.
>
> Thanks Jack!
>
> ------------------------------
> James Fowler
Jack Hershey
and the manual alternatives. The computer has to perform a lot of stuff in
detail that may be omitted in a manual approach although are both ways are
largely mechanical. how detailed the trading effort is is up to the
practitioner. I notice there is a direct corrolation between effort and
ROI.
The stock paradym operates at about 10% ROI per each 6 to 8 day half cycle,
cycle after cycle. use 50 cyles of capital appreciation over three years as
a nominal rate of use of market time available. For the futures index use
the 30 points per day or 10% on margin for a daily ROI as a minimal effort
target. If day trading is envisioned using an automated computer model,
then there is a multiplier of up to 5. another view is to divide the the
daily hi/lo point range by the margin points to get a simple nominal
potential.
for both stocks and commodities these approaches represent only the most
actively monitored and analyzed portion of the portfolio, i. e., that part
related to trends rather than investment positions that are sometimes held
in adversity. The surpluses in the high ROI attended to sectors are often
moved to other sectors to preserve the size of the active segments. An
alternative is giving away money.
The heirarchy of investment to me is:
stocks
commodities
1. captital appreciation slow
fast
2. market selection equities
futures indexes
3. the cycle 6 to 8 days
intraday
4. the trend, non trend continuous 1/2
the time
5. Sentiment flawless trend
A- E
6. Trade Market entry, exits
MKT, stops
7. Trade protection stops
stops
for stocks and commodities the basic software covers 3 through 7 and for
stocks there is also a computer analysisfocus at level 2 to obtain daily
lists of ripe stocks (scoring stuff, 7 -0). What follows is a listing of
the levels where data is processed to the levels and the output of levels
are used as gating signals that enable level 6 trades and level 7 risk
minimization.
I'll make two passes (III and IIIa) to get an overview and second, to
elucidate details. Lastly (IV) I'll try to relate the automated computer
paradym to a reasonably simple manual system that has a high dialt ROI yield
and broad transference potential.
Early levels ( 3 and 4) generate gate signals, intermediate levels (5)
generate go/no go conditions and the last levels (6 and 7) display decisions
and values. In sub part IIIa, I will list inputs to to each routine in
detail to round it out. you can also see that there are several
conventional indicators that may be used.
Level 3 The cycle. Determines whether trading is possible or that we are
sidelined. No trading for a or e and possible trading for b, c, and d.
Output signals are trade or sideline.
Sideline starts a clock that expires and causes the test to be repeated.
Trade signal triggers the start of 4 trend routine and 5 sentiment routine.
Level 3 routine checks: sync and time < 15:30 EST. Sync is determined by
spread of DJXX- INDU is within +/- 0.1 * MA10 of difference times ratio of
(90 - days)/ 90 and (absolute value of difference of MACD's of INDU and
DJXX) <3 * (volatility factor) where vol factor is centered on 1 nominally.
Level 4 Trend. right side line determines presence of trend as a gate that
has to be opened by a test for failure of a non trend. There are two
outputs: non trend and failure of a non trend. Failure of a non trend goes
to level 6 Trade prep routine and right side line routine also in level 6.
Non trend signal goes to level 6 MACD "0" test.
the level 4 test establishes the price envelope slopes at any time when the
5 min slow MACD signal absolute value is > 3 to 5.
Level 5 Sentiment. This is where the entry value of the look up tables
comes from. use thes inputs from associated subroutines that periodically
logs the subrouting values:
a. price volatilitiy ( lots of choices to choose from) (value)
b. MACD hi /lo for day and for recent past (3cycles min) (value)
c. MACD offset 9 use MLR and determine equilibrium too (slope) (value)
d. status of congestion, convergence, and centering This is a price hi/lo
status where congestion has nominal unchanging value times volatility
factor; where convergence has a value rate of change decrease; and where
centering is a relatively minumum value for up to several minutes. (three
siganls indicating status)
e. pace. several component routines: Average (3) peak to peak time; MACD
hi/lo peak to peak ; i min slope test; 1 min retracement test. (three
signals: fast, normal , or slow)
These five will give you a choice of A-E and a p, q, r, s, or t signal for
look up.
Look ups are required for:
a. MACD "away" +/- 10 nominal (value and yes/no signal)
b. TO (time out value) 7 min nominal (value and yes/nosignal)
c. stop offset (value)
d. scalp offset (value)
e. specific design stop (value from c, d and price)
Level 6 Trade this involves several "test" routines.
they include:
A. Trade Prep
1. peak/trough test (retracement %, yes/no signal)
2. MACD "0" test (yes/no)
3. MACD "away" test (yes/no)
4. no trade test (yes/no)
5. Reversal peak/trough test (yes/no)
B. MACD Status
1. Divergence (yes/no)
2. Entwined (yes/no)
C. Volume status (adequate, increasing, decreasing, low)
D. Order/ C&R signal
1. countdown routine (pending/wait/stop)
2. action routine (go)
E. Exit /Reverse signal
1. right line (value)
2. BO reversal (yes/no)
3. trend end (yes/no)
Level 7 Stop Design. This sector generates values as appropriate or action
signals
a. Order/Bracket routine (buy, sell values)
b. market entry/exit override routine (go)
c. trailing protection routine (stall, retrace) (value)
d. C&R frequency/trigger routine (next time)
The above is the overview of levels and signals and the values generation.
it involves a bunch of routines that are gated. In IIIa I'll connect it
together a little more carefully now that we can see what signals are
involved. To make it work best, it will iteratively adjust its self as the
market changes its nature because of news, climate caused by slower fractals
(Bull bear correction stuff) and periodic phases like quarterly reports and
year end stuff. Back testing will allow us to slip in a few coefficients
as you would expect. I have it flow charted signal wise and there is no
particular frequency of operation that is constraining; this will work on
very slow computers. I just want it to continually display what is going on
and announce the script for phoning in an order a few minutes ahead of time
for the market to come to the order/stop.. It, after all, will require
price and volume inputs from INDU and DJXX or similar pairs.
In sub part IV I'll set up the manual version in detail using the two
monitoring charts I have presented. I average four or five pages on each
log daily and that will be about the amount of display text that will emerge
as the computer announcements form the automated log for the day.