Nemer, Tom and whoever interested, here is something about my system.
Date SPX M:spx m:a-d S:spx Singal Profit
===== ===== =====
02/27/97 795.07 down down turn back down SHRT(larger & longer play)
03/05/97 801.99 UP oversold LONG(small & brief play) -6.92
03/13/97 789.56 down heading down SHRT(larger & longer play) -12.43
04/15/97 754.72 UP UP heading UP LONG(larger & longer play) 34.84
06/18/97 889.06 down overbought SHRT(small & brief play) 134.34
07/01/97 891.03 UP oversold LONG(larger & longer play) -1.97
07/18/97 915.30 down heading down SHRT(small & brief play) 24.27
07/23/97 936.56 UP heading up LONG(larger & longer play) -21.26
08/08/97 933.54 down down turn back down SHRT(larger & longer play) -3.02
09/02/97 927.58 UP UP oversold LONG(larger & longer play) 5.96
09/11/97 912.59 down oversold already long
09/16/97 945.64 UP heading up already long
10/08/97 973.84 down overbought SHRT(small & brief play) 28.20
10/16/97 955.23 down down oversold wait if not short yet 18.61
The way the signals are generated are summarized below.
M:spx m:A-D S:spx Signal
===== ===== =====
UP UP UP (oversold or heading up) LONG (larger & longer play)
down down down (overbot or heading down) SHRT (larger & longer play)
down UP UP (oversold or heading up) LONG (small & brief play)
UP down down (overbot or heading down) SHRT (small & brief play)
UP UP down (overbot or heading down) wait for pull back if not in yet
(may miss part of a strong move)
down down UP (oversold or heading up) wait if not in yet
(may miss big-K)
UP down UP (oversold or heading up) None
down UP down (overbot or heading down) None
There are three separate sub-systems:
M:spx Intermediate-term market direction, based on spx;
m:a-d Intermediate-term market direction (usually shorter than M:spx),
based on market breadth;
S:spx Short term market situation, based on spx.
Table on the top lists the actual signals generated by my system since I
used it for real trading in February. M:spx and S:spx are nothing special.
So I am going to talk about m:a-d, which is my favorite.
Market breadth is calculated daily:
(NYSE number of advance issues + NASDAQ number of advance issues) -
(NYSE number of decline issues + NASDAQ number of decline issues)
These numbers can be obtained from most quote machines. (Originally,
volumes are part of the formula, I dropped them later because they don't
make too much sense to me.) Then I smooth the result by calculating
exponential moving average. So I have one number every market day, I call
it a-d pointer, which becomes the basic number for m:a-d sub-system. The
pointers then are smoothed one more time to produce x days simple MA
line, along with two pairs of bollinger Bands (also based on x days), so
that
(MA+2*std) >= HBB1 > HBB2 >= MA >= LBB2 > LBB1 >= (MA-2*std).
Well, these lines are similar to BBs but they are really not. Why? In
the case of BB, the number of std (standard deviation) is fixed in any
single chart. If you choose 2, then the distance between HBB and MA
(and LBB and MA) is 2*std everywhere in that chart. The value of std
changes every day, but the multiplier (2 in the example) never changes.
In my system, however, multipliers are constantly changing. In fact,
they are the function of market strength a while ago. If one applies BB
to a strong stock, say DELL, one can see price sometimes penetrates HBB,
then falls below HBB or MA pretty soon (depends on number of days and
number of stds chosen). Experienced eyes can see the up trend is still
intact because previous move was so strong. However, computer recognizes
only numbers in most cases, so you have to quantify everything. Back to
my system, when the movement of my a-d pointer a while ago was strong,
the value of multiplier for HBB1 and LBB1 increases, but the value of
multiplier for HBB2 and LBB2 decreases, so that two zones formed by
HBB1 and HBB2 (high zone) , and LBB1 and LBB2 (low zone), become wider.
A sell signal is generated when a-d pointer moves across high zone from
above, a buy signal when pointer moves across low zone from below. It
is more difficult for a-d pointer to cross a wider zone. Meaning, it
is not easy to trigger a sell signal if the previous up move was strong.
I can usually correctly predict in the early morning that a new signal
will be generated by day end. Take the case of 10/7, a-d pointer was 370,
HBB1 397, HBB2 334. In order to trigger a sell signal (pointer moves
from above HBB1 to below HBB2), declining issues needs to outnumber
advancing issues by 370 or more. Early next morning, that number was
about 900. So I reversed the positions.
There is at least one fundamental drawback for all std based systems, I
think. Some statisticians claim that market prices have unlimited
variances (hence stds), and I tend to agree with them. One implication
of that is that std based systems can't predict market crushes.
Hopefully, however, those systems will generate a sell signal before
big K actually occurs.
I do believe market breadth contains a lot prediction power. The
question then is how to use it. The m:a-d sub-system alone produces
pretty nice results for me. It is my primary signal, while M:spx and
S:spx are secondary. From the summary table above, you can see that
one of my rule is never go against m:a-d signal.
The market is such a wonderful place for me to learn something new every
day. As you can see from my previous posts, I bought puts in the
morning of 10/8. I sold part of October puts two days later, and I
still think I did the right thing because time value was decreasing
rapidly, and near term options are for very quick play anyway. In early
morning of 10/16, market was up, so I closed all puts. Since both M:spx
and m:a-d were down then, I should have at least kept two January puts.
I also bought 3 Sep. 925 calls the same morning, on the anticipation
that a buy signal would be generated next day. My thought was getting
a small position just to test the water. The test costs me $2800.
(Overall, that buy signal of 10/8 gave me around $4100 profit, not too
bad.) Now you see how costly it is for me not to follow the system.
Most moneys on OEX index options are made by options sellers because
time value works for them. Trading future is a much better alternative,
in most cases. Because of the nature of time value, I make a trade only
when I think the odd is absolutely on my side and the move is going to
be strong and quick. Even in the case of buying longer term options, I
assume that the trade is going to last maximum of two to three days.
Overtrading in the OEX market is nothing but suicide, pure and simple.
When play relatively large, it is critical that market goes my direction
at the very beginning. On average, my system generates one signal every
month, which is about right for me.
Hope all these make some sense to you. Comments are very welcome.
After all, we are here to learn from each other.
Regards,
DC |
