Since things are quiet here this evening, I thought I'd post the
monthly recap of our neural trend's performance for August. It's
actually a newsletter we send to subscribers, but I've removed
all the links to our web site so as to avoid the appearance
of solicitation..that is not my intention here.
Allan
============================
The S&P futures market in August was characterized for most of the month by
a tight horizontal trading range. This range was broken below in dramatic
fashion, however, in the final few days of the month. This early range was
de-marked by a support line which could be drawn exactly through the spike
intraday lows of four evenly spaced days. The break came on a day in which
the S&P lost 51 points, with the two following days seeing losses of 5
points and 77 points, respectively. At the end of August, the S&P futures
had closed at levels last seen in mid January, erasing all of the nearly 20%
gains of 1998.
All of Trendy's models performed very well in August, capturing around two
and a half times their 2-year historical average returns. Further, our good
fortune was not based on one big trade, as our models engaged in an above
average 9 to 12 trades during the month.
Finally, we made some last minute refinements to our new Trendy Millennium
model, in preparation for its September release.
Trendy's Performance for August
===============================
Commentary:
Our models had an excellent August by any measure. Over the 9 to 12 trades,
our hypothetical equity curves rose in mostly a straight line. We connected
with a Short trade on Thursday, Aug 27th, riding the 55-point decline. We
subsequently gave back part of this win by being incorrectly Long for the
Monday, August 31st collapse.
That Monday evening our net had come in bullish (for Tuesday, Sept 1st), but
we elected to publicly force the net to neutral due to the huge volatility
of Monday's market. Tuesday was indeed a powerful 45 point up day, but early
negative swings would in fact have stopped out our models for losses long
before the rally began in earnest. Thus, our neural net was correct on that
day, as was our decision to forego the signal due to the expected volatility.
During August our S&P models moved into all-time equity high territory, and
our DowJones model registered an all-time high, relative to the point at
which that model was released.
Performance Summary:
Points Average Total Winning Points Points
Returned Points Trades Trades Returned Returned
in the per in the in the In Last In Last
Month* Month* Month Month** 3 months* 12 Months*
-------- -------- -------- -------- --------- ---------
S&P Hybrid +67 +24 9 66% +102 +246
S&P DayTrade +46 +18 12 50% +73 +231
DowJones DayTrade +408 +160 12 58% +350 +1738
* slippage/commission not deducted; simulated trading
** excluding breakeven trades (winners/losers under 1-sp/5-dj points)
*** trade still open; figures as of this writing
Trendy Millennium; Refinements & Commentary
===========================================
Refinements prior to September release:
Over the month of August, we made a few changes to Millennium in preparation
for its September 1st release. First, we replaced Cocoa with Sugar, because
we decided that the historical profile of Sugar was superior. The system now
trades Wheat, T-Notes, Crude, Copper, and Sugar.
A second change that we made was to equalize the relative margin
requirements among the various commodities, by trading different numbers of
contracts. The table below shows the contract lots being traded, and the
resulting Combined Margin requirements. Note that these Combined Margins are
roughly evenly balanced among the five commodities.
Commodity NumContracts 1-lot Margin Combined Margin
--------- ------------ ------------ ---------------
WHEAT 3 $675 $2,025
T-NOTE 1 $1,360 $1,360
CRUDE 1 $2,025 $2,025
COPPER 1 $1,620 $1,620
SUGAR 3 $560 $1,680
Results from our "Paper & Pencil" Backtesting Period:
As we have stated, we first investigated the concept which has evolved into
Trendy Millennium by means of paper & pencil testing. We investigated a 21
to 33 month period in this manner, in order to convince ourselves that the
system concept was worth pursuing. We then wrote custom software to actually
trade the system, and to more accurately model performance. Due to data
constraints having to do with threshold level generation, the database from
which the computerized model runs begins with January 1998. However,
presented below is a summary of the results of our complete paper & pencil
backtesting period. The ending date for the comparisons presented below is
July 1998 (which represents a several month overlap with our computerized
model's data).
RtnOn Ann
Contract Mrgn Mnths Win Lose Win% Win$$ Loss$$ Profit Mrgn Rtn
-------- ------- ----- --- ---- ---- -------- ------- ------- ----- -----
Wheat(3) $2,025 24 22 10 69% $33,000 $17,370 $15,630 772% 386%
T-Note(1) $1,350 21 16 7 70% $12,000 $3,600 $8,400 622% 356%
Crude(1) $2,025 33 11 22 33% $22,000 $8,000 $14,000 691% 251%
Copper(1) $1,620 33 30 17 64% $26,250 $12,500 $13,750 849% 309%
Sugar(3) $1,680 21 20 11 65% $13,440 $6,180 $7,260 432% 247%
-------- ------- ----- --- ---- ---- -------- ------- ------- ----- -----
OVERALL 99 67 60% $106,690 $47,650 $59,040 310%
Our inspection of the Win Percentages and Annualized Return on Margin of
these data, relative to the Descriptive Statistics data generated by our
computerized model implementation (found in the Full Report), suggests to us
a general similarity and agreement.
System STOPs:
Note that, because of the nature of our system, for the remainder of the
first day in which a new trade is entered there is no corresponding
Protection STOP. It is not until we run that evening's numbers that we are
able to compute a new Reversal Threshold level, which functions as a
traditional STOP loss level. It has been our experience in backtesting that
it is unlikely that a contract would hit a trade entry threshold, and then
on the same day reverse direction and move sufficiently in the wrong
direction to expose that trade to significant losses. However, since
anything can happen in futures trading, individuals are encouraged to be
aware of their current market liabilities, and to implement their own
exposure-limiting techniques for the first day of each new trade, as they
see fit.
This is a BETA Test:
We are considering the next few months to be a "Public, Real Time, Real Money
BETA Test" for Trendy Millennium. By this, we mean that it is being offered
as a system which we believe to be finished and complete, but which --
reasonably -- should go through further testing before its effectiveness is
fully known. We are therefore now trading Millennium in one of our private
company futures accounts. As this is one of the two accounts for which we
make public the equity balances and monthly statements, all interested
individuals will be able to follow Millennium's actual trading results on
our web site going forward. We have no opinion as to the advisability of our
subscribers trading Millennium now, during this BETA test period. |
