|
PureBytes Links
Trading Reference Links
|
Bruce Babcock's book "Profitable Commodity Futures
Trading from A to Z" contains a snippet of an
interview with Fred Gehm. (Fred wrote the EXCELLENT
book _Quantitative_Trading_&_Money_Management_ which
is available from Amazon, Traders Press, etc).
On p. 178 of Babcock's book, Gehm describes his
"torture test" for futures trading systems.
...by subjecting it to the worst possible conditions.
For example, he always assumes he is filled at the
high of the day for buys, and the low of the day
for sells. "If a trading system survives that, it
should be pretty resilient to other market adversities."
This is a GREAT idea. It imposes the MAXIMUM SLIPPAGE
on a trade, both at the entry AND at the exit.
Of course, this is hard to do using Omega Research
products. (It's a built-in capability of TTChartbook
and Trading Recipes, though.) To stimulate discussion,
I offer a prize of ten dollars American money to
the first person who proposes a feasible way to
implement Gehm's torture test in TradeStation, and
then implements the proposal. Here are two approaches
that work but are clumsy:
(1) create another data series that has the
dates advanced by 1 day. Then buy at high of
data2, sell at high of data2, etc.
(2) modify the volume and/or open interest fields
so that they encode tomorrows high and low.
Resubmission of (1) or (2) will not result in a prize :-).
I implemented Gehm's TORTURE TEST in other, non-Omega,
software, and applied it to a common, well-known
trading system. I simultaneously traded twenty
futures markets out of the same account. These are
markets that I actually trade myself in my own
real-money account. They come from four commodity
groups: the energies, the softs, the currencies,
and the interest rates. I don't trade the meats,
the stock indexes, the grains, or the precious metals.
** Surprisingly, this trading system "WKPD" was very
** profitable even when every entry and every exit
** had MAXIMUM SLIPPAGE. It looks amazingly robust.
TEST METHODOLOGY
markets: BP, CD, CT, CL, DM, DX, FV, HG,
HO, HU, JO, JY, KC, LB, MB, NG,
SF, TU, TY, US
Calculations are performed at night after the
market closes. They determine whether or not
to make a trade the next day. Usually, the
trade would be placed "at market on the open",
but in this test, ORDERS WERE ALWAYS FILLED AT
THE WORST POSSIBLE PRICE: buy orders were filled
at the high of the day, sell orders were filled
at the low of the day.
Besides this MAXIMUM SLIPPAGE, a commission fee
of $25.00 per contract per round trip trade was
also assessed.
I tested the common, well-known, public domain
system this way. And for comparison, I also
tested a vendor system (Aberration) the same way,
using Gehm's torture test. Surprisingly, the
"free" system (it comes with TradeStation), did
about as well as the vendor system, on a one-lot
basis. {but not as well when using geometric betsizing}
RESULTS WITH MAXIMUM SLIPPAGE AND $25.00 COMMISSION, TRADING ONELOTS
====================================================================
WKPD* Ntrades=743 NetProfit=$339879 AvgTrade=$457.44
Aberration Ntrades=632 NetProfit=$331012 AvgTrade=$523.75
Take a look at that. Even with MAXIMUM SLIPPAGE, i.e. the worst
possible fills, the Well Known Public Domain ("WKPD*") system
averaged a profit of more than $450 per contract per trade. Wow.
The vendor system did a little better but not much. (Its strength
is in non-onelot trading, i.e. trading with geometric betsizing).
Fred Gehm is right: these suckers ARE resilient!
As long as I was running the test software, I indulged myself and
did a test using minimum slippage, i.e. fills occurred EXACTLY at
the price the system hoped for. In both of these systems, all orders
are placed market-on-the-open, and so I rigged the test software
to fill the orders exactly at the reported open price of the
historical dataset.
RESULTS WITH MINIMUM SLIPPAGE AND $25.00 COMMISSION, TRADING ONELOTS
====================================================================
WKPD Ntrades=743 NetProfit=$918556 AvgTrade=$1236.28
Aberration Ntrades=632 NetProfit=$781624 AvgTrade=$1236.75
Apparently the days on which the WKPD system trades, are a little
more volatile than the days on which Aberration trades. The average
difference between the desired fill, and the worst-possible-fill, was
a little bigger with WKPD. But not very much.
DISCUSSION
==========
I was surprised. This well known public domain system has been around
a long time. I don't have a huge private library; the earliest book
*I* own that mentions this system is Jack Schwager's "A Complete Guide
To The Futures Markets", copyright 1984. So the system is AT LEAST
fifteen years old. And Schwager's book contains a footnote referring
to a paper published in 1980 (19 years ago) that discusses the system,
but I haven't seen the paper myself.
A version of the system ships "free" with TradeStation, at least,
I got it with my copy of TS version 4.04, build 19.6. It's so
tiny it can be written in the "Quick Editor".
Despite being old, well known, public domain, and widely available,
this system gives great test results. I wonder how the Efficient
Market Hypothesis people feel about that :-).
I have attached an ELA to this message, so that you can play with
the system if you like. But remember, Omega Research products can't
do the MAXIMUM SLIPPAGE test, or at least, nobody has claimed my
$10 wager saying so :-). Thus you won't really be able to apply
Fred Gehm's torture test easily, I'm sorry. The ELA is cloaked;
I'll uncloak it in a few days if anybody really wants me to.
Meanwhile, think it over and ponder the question: what well
known, public domain, 19-year-old system could possibly give
an average of $457 profit per one-lot trade with MAXIMUM
SLIPPAGE (and $1225 per onelot without slippage)? It's a fun
little problem to work on. Hint: this is an indicator-based
system, not a chart-pattern-based system. Murray Ruggerio's
articles in _Futures_ magazine claim this is not possible :-).
The system is "ZZZ_WKPD", a name chosen so that it will be
alphabetically last on your list of installed systems in your
Omega menus. That ought to make it especially easy to find.
Attachment Converted: "c:\eudora\attach\Zzz_wkpd.ela"
--
Mark Johnson Silicon Valley, California mark@xxxxxxxxxxxx
"... The world will little note, nor long remember, what we
say here..." -Abraham Lincoln, "The Gettysburg Address"
|