|
PureBytes Links
Trading Reference Links
|
Hi Frank,
I agree that 10 breakouts would be so limited in the assessment of a system,
regardless of the world events, that is would be useless. One of my research
lines led me to a conclusion that different markets have their own set of
patterns. No system I've found to date will work on all markets. Possibly
there is one, but I've not seen or heard of it. However, it is a good
validation process to see if the system works in other markets but only on a
limited basis. By the way I do trade Euro, Swiss Franc and Yen. The system
works in those markets as well, which triples the number of trades in the
data base (approximately 450 to 500 trades).
Anyhow I guess it comes down to certain systems work in certain markets for
a finite length of time and one has to have some stop loss on the system
when it ceases to function. Another possible addition to try to handle the
uncertainty would be to trade a portfolio of systems, each having multiple
issues for the separate systems and preferably all non-correlated.
Regards,
Michael
******--------***********---------********
Michael,
Sunday, February 23, 2003, 7:59:33 AM, you wrote:
MM> Hi Frank,
MM> I like your opinion. A small data set has always been a problem with
swing
MM> systems that may average 20-30 days in the trade. So you're right,
that's
MM> why I brought up the point of having 3,000 trades. The problem with that
is,
MM> only the Dow goes back to 1870 or so and I'm not trading that. Almost
any
MM> other market is going to have a maximum of 30 to 60 years of data that's
MM> usable (I know some older commodity charts but they only have the close,
as
MM> does the Dow once you go back past about 1924). Any idea how to get
around
MM> that?
It depends on what these systems are looking for. Suppose one looks
for a very specific breakout pattern. Perhaps you can look at the
Dow for this pattern, even though you are not trading it. People will
say this is crazy, but I think it's crazier basing a system's validity on 10
breakouts that happened in the past 30 years in the market you are
trading in. You need more data to validate this pattern, imo.
Breakout patterns, and trends, can be treated as generic, and
shouldn't be limited to the market you trade in. Look for enough
uncorrelated markets, add up all of the specific
breakouts you can find and you may get a more respectable number that
can reveal whether or not these patterns have any validity.
"Uncorrelated" markets is important because breakouts and trends are
notorious
for looking similar everywhere at the same time. I would treat that
as one event. You want to see that this specific breakout pattern is a
consistent phenomena that takes place when any event happens that
attracts people to any market.
However, there may be system ideas that shouldn't be tested in such a
"horizontal" manner because they are non-generic and very specific to the
nature of the
market you are trading in. Perhaps, for example, cyclical patterns,
or repetitive associations between your market and another.
Another approach is to look at systems that trade more frequently.
But if you have confidence in the past 10 breakouts in 30 years
because of their association to the same type of world events, then you are
attempting to validate your systems with speculation.
-F
MM> Regards,
MM> Michael
MM> ****************************
MM> Michael,
MM> 175 trades in 31 years does not give me a lot of confidence. I would
MM> then focus on the shape of the equity curve, which is essential.
MM> That's why the Sharpe Ratio is an important effort, imo.
MM> A sloppy equity curve constructed from a small number of datapoints is
MM> a random event, in my opinion, whether it represents 5 months of 5
MM> minute data or 31 years of daily data. Regardless, of your system
MM> ideas, the types of market personalities you're
MM> trading, or the perfect alignment of the planets on a Tuesday, you
MM> need a "tight" equity curve filled with several data points (trades)
that
MM> are
MM> based on realistic trades and trading (i.e. reasonable liquidity,
MM> slippage, etc.), because it will tell you sooner, rather than later,
MM> that the non-random profitable phenomena you extracted from the market
MM> "ether" is now meandering into no man's land, which is your cue to take
MM> your money and run.
MM> Conversely, if your equity curve bounces all over the place toward the
MM> current net profit of $248,391, after 176 trades in 31 years, you
MM> won't know that the ship sunk before you're underwater.
MM> Any cleaver speculation or market theory cannot save a lousy equity
MM> curve.
MM> Saturday, February 22, 2003, 7:58:53 AM, you wrote:
MM>> Hi Frank,
MM>> Good question. All of the systems were break out or trend following
MM> swing
MM>> systems (no day trades). We went through 5 systems like Aberration and
MM>> Dollar Trader. A good example would have been a single contract on
MM> Dollar
MM>> Trader with the following quantity of trades.
MM>> 1972-1980 = 45 (gain $97,187 with one bad year of negative $1500)
MM>> 1980-1990 = 73 (gain $140,137 with on bad year of negative $12,427)-
MM> that's
MM>> a hell of a nice 18 year run.
MM>> 1996-present = 58 (11,067 with four bad years - two back to back
MM> $15,927)
MM>> Obviously there are a lot of indicators missing like sharp ratio, max
DD
MM> per
MM>> trade, max DD, # days in DD, etc. It sure was a severe reality
MM> adjustment. I
MM>> had been going for longer lengths of time so as to make a system go
MM> through
MM>> different markets to increase probability of gain. What I didn't put on
MM> the
MM>> chat was that this was more a study of markets than systems as I
noticed
MM>> that no matter the system, they would lose money on the BP. The person
I
MM> was
MM>> working with had a long term background in the futures market and
MM> mentioned
MM>> that the BP had been a darling market in the 80s so we took it from
MM> there to
MM>> see the contrast. There are obviously quite a few systems that have
very
MM>> good long term gains (10 years plus). All of the systems had very good
MM> gains
MM>> for the BP on the time segments mentioned earlier. Also the losses for
MM> the
MM>> last 6 plus years did vary by a good percent.
MM>> I like to use Futures Truth as a comparison because they only list
MM> systems
MM>> they have been able to paper trade to see actual results including
MM>> commission and slippage (and they track about 200 systems). They don't
MM> list
MM>> anything that they haven't personally tracked with the exception if the
MM>> person has actual brokerage statements for a couple of years. So you
MM> know
MM>> you're looking at an apples to apples comparison. There is the caveat
MM> that
MM>> if the systems trades a thin market you can't possibly use the system
MM>> results but when you're into S&P, Yen etc., then you have a pretty
MM> liquid
MM>> market (assuming you trade the currencies through the banks and not on
MM> the
MM>> floor) and the numbers should be close over a longer period of time.
MM>> One thing I have not reconciled is the comparison of number of trades
MM> rather
MM>> than over a period of time. A day trade system may make 3,000 trades
in
MM> a 5
MM>> year period whereas a swing system would have to go 300 years to make
MM> that
MM>> many trades and how would it do over that period of trades rather than
MM>> looking at time. Although, for application purposes its no doable as
MM> you're
MM>> not going to trade for the next 300 years unless you're a vampire and
MM> plan
MM>> on living longer than that (I have heard there are some vampires in
this
MM>> business).
MM>> OK, that's the long answer.
MM>> Regards,
MM>> Michael
MM>> PS - MARK WROTE ABOUT ABERRATION (what every market does have is a
MM>> personality and like a human these
MM>> traits can be identified and psychology applied to the subject to
MM>> obtain the maximum output of production like in a workplace
MM>> environment for example.)
MM>> He's right about every market having a personality except "psychology"
MM> will
MM>> do nothing but get you into trouble - systems testing is a much better
MM>> answer. Aberration is a good example where it does well in some markets
MM> and
MM>> terrible in others.
MM>> *********************
MM>> Michael,
MM>> how many trades did these systems generate (on average) between
MM>> 1980-1990? How many trades did they generate since 1996?
MM>> Friday, February 21, 2003, 11:33:10 AM, you wrote:
MM>>> I wrote the statement below a few days ago and although no one on this
MM>> forum
MM>>> replied I did come up with some additional data that might be of
MM>> interest.
MM>>> We were researching what tends to be a very bad market on any system
MM>> that is
MM>>> tested on it - the British Pound. From 1996 to present, no matter what
MM>>> system tested (unless over-optimized), they all lost money. I'm
talking
MM>>> about well known and very good performing systems. However, ALL of the
MM>> same
MM>>> systems made money (and nicely too, I might add) from 1980 to 1990.
MM>> These
MM>>> are all good systems that presently make consistent returns year in
and
MM>> year
MM>>> out in other futures markets.
MM>>> Thought it might be of interest.
MM>>> Regards,
MM>>> Michael McGahee
MM>>> **************************
MM>>> I my investigation of systems and statistical probability I've gone
MM> down
MM>> the
MM>>> road that essentially says the longer the system survives with the
MM> least
MM>>> amount of draw down and the highest return (by tracking actual trades
MM> in
MM>>> very very liquid markets so as to minimize the possibility of over
MM>>> optimization) the higher the probability that I can rely on that
system
MM>>> continuing to perform. As some of you know I've stated similar
comments
MM>> in
MM>>> previous strings. This, of course, comes with the same caveat that
past
MM>> or
MM>>> present performance is not indicative of future performance (which
MM>> applies
MM>>> to any theoretical or practical system).
MM>>> The above simple philosophy put into practice has led me to look at
MM>> certain
MM>>> data and systems from that point of view. The only limiting viewpoint
MM>> that
MM>>> I ask is to put forth ideas or postulates based on applicable material
MM>> and
MM>>> not to stray too far into the "theoretical" as that tends to muddy the
MM>>> water. The floor is open. Any takers?
MM>> --
MM> --
--
|