Volatility puts algo trading under pressure

[tortoise]

Quote from QuantPlus:

Yes... what I have been trying to explain repeatedly...
But a better way to frame it is...

Pro Trader with 10 years Experience Intelligence

VERSUS

Computer Algorithm "Intelligence".

Everybody here is looking for a short-cut to free money...
So the incredible Learning Curve a successful trader goes through in 10 years is trivialized and discounted.

I think this is key: There are no shortcuts. Those who are willing to put in the time/effort/toil to obtain mastery AND have the gift for this sort of thing will outdistance any A.I. configuration out there at the present time and, very likely, for the forseeable future. Everyone else is better off with an automated approach.

I took exception to the poster who asserted that ANY sound methodology could be coded, implying that said methodology could not be sound, if it defied a coder's expertise. That strikes me as a statement of faith as ungrounded, in its own way, as the guesswork of newbie trader/gambler.

 

[tortoise]

Quote from rosy2:

the calculation on your part could be automated if its consistent. anyway,
maybe there is a guy who can "read" the market and has a feel but when you look at the super traders their winning/losing ratio is about 50/50 so they dont even know winners from losers.

i am sure most people on this board will agree with you because the market feel approach to trading requires no work. its just fun.

I just saw this. Here's the disconnect:

Define "consistent."

Returning to my workplace analogy...my sense of when to break the "rule" and joke up the boss may seem to be entirely random, inconsistent in terms of mood "quanta," time of day, seasonality, earnings cycle, number of employees out sick, etc., etc.

But it's not.

It's entirely consistent with my read on his personality, based on thousands of hours of observation of this individual, and the accumulated wisdom of a lifetime of observation. And that is why I can explain to you just why I decided to tell him that particular joke on this particular day: It's not merely some spontaneous, random act. It's a function of "EQ." social intelligence. A calculation.

Again, the fact that I can do this, not just once, but again, and again, knowing when to follow the rules, when to break them, how to break them (albeit breaking rules within a larger set of rules which are NEVER broken -- e.g., never piss on the boss's shoes, never tell him he's an ugly pig, etc.) suggests to me that these are not random events. Rather, they are consistent in their application of the existing data to existing circumstance, and they are consisten in the result they produce.

Could these interactions be automated? Could code be written for a computer to generate, in real time, the words I'd speak and the things I'd do to curry favor with my boss, based on the same inputs I'm using to make my decisions?

If the answer is "yes" then please tell me where this has been done.

If the answer is "no" then please tell me this:

On what basis have you come to believe that the markets are any less complex?

 

[rosy2]

Quote from tortoise:

I just saw this. Here's the disconnect:

Define "consistent."

On what basis have you come to believe that the markets are any less complex?

i define consistent as one does the same thing under certain circumstances. you said your acts are not random so then they are consistent. that fact that you have accumulated thousands of observations to come to your conclusion on what to do is precisely what programs can do...take thousands (or millions) of observations and come to some conclusion that will probably occur.

i believe the markets are complex and that any consistent strategy can be automated. Unless you have a background in programming and basic math then you might not even know what is possible.

 

[maxpi]

Quote from FAST.AM:

Just wait until the machines take over.
One day they will create an artificial panic and crash or markets.. I think this can also exploited if these machines use past trend lines this can be predicable behavor

There was a great Sci Fi short story decades ago about just that!! If somebody can tell me the story title or the author I would be very happy. I have wanted to re-read it ever since, never ran across it...

 

[tortoise]

Quote from rosy2:

i define consistent as one does the same thing under certain circumstances. you said your acts are not random so then they are consistent. that fact that you have accumulated thousands of observations to come to your conclusion on what to do is precisely what programs can do...take thousands (or millions) of observations and come to some conclusion that will probably occur.

i believe the markets are complex and that any consistent strategy can be automated. Unless you have a background in programming and basic math then you might not even know what is possible.

i hold an undergraduate degree in applied mathematics from the university of chicago. apparently, one requires more than a background in programming and basic math to know what is possible.

 

[gaidaros]

way too many variables involved to be put in a computer algorithm
maybe the future holds other spooky solutions:
brain cells cultured on a dish and networked to the market

when they perform well they get, say, glucose or opiates, when they perform
bad they get nothing. then you let them explore and learn from the market. i heard it has been done with brain cell culture and flight simulators

http://iwarrior.uwaterloo.ca/?module=displaystory&story_id=1652&format=html&edition_id=35

 

[NYOBScalper]

Quote from rosy2:

i define consistent as one does the same thing under certain circumstances. you said your acts are not random so then they are consistent. that fact that you have accumulated thousands of observations to come to your conclusion on what to do is precisely what programs can do...take thousands (or millions) of observations and come to some conclusion that will probably occur.

i believe the markets are complex and that any consistent strategy can be automated. Unless you have a background in programming and basic math then you might not even know what is possible.

Many strategies can be automated, but to say that my market feel can be automated shows me that you are the kind of guy who doesn't have a feel for price action and thus assumes it acts in a way that can be analyzed by the scientific method. Computer programs are always changing... my trading edge is based off exploiting the stupid things masses of algorithms do. Everytime a new algorithm for working orders comes out, it changes the way the market has to be traded, and there is no way you can program a computer to predict the way the futures will be and thus react to its change. If you can do that, you could make more money by getting that kind of advanced AI applied to other purposes in life or selling ot to a company. By "do that," I mean create a strategy that is consistently profitable and as adaptive as a wise, experienced, consistent trader. That can only be proven with time, of course, which is the fundamental problem behind most these computer strategies in the first place. The more people have computer programs chasing the same money and the same fills the more you have computer programs that overpay and do dumb things. I would say 60% of my trades are algorithms, but generally for me, every trade is unique. I don't have defined "setups," I let setups develop and react according to my ability to manage risk. It's much easier to give computer black and white setups and have them look for thousands of repetitions throughout the day in a universe of stocks. When a box's edge goes away, though, it had better be able to get out and not give away all its profits.

 

[maxpi]

I have very well defined parameters for entering a trade but when it comes to the decision, it's not exactly gut feel but the decision between conflicting indications comes from experience and nothing more... it bothers me because my goal starting decades ago was to have automation and i never could achieve a fully automated strategy, currently I am happy with semi-automation, some of the trade management can be done by the computer better than by hand...

 

[nitro]

Quote from tortoise:

i hold an undergraduate degree in applied mathematics from the university of chicago. apparently, one requires more than a background in programming and basic math to know what is possible.

You hold an applied mathematics degree from the university of Chicago? This is rather surprising, as the university of Chicago has zero interest in anything "applied" except maybe on the medical side - it prides itself on theory.

But to answer your statement, in my experience, is that the answer is yes and no, simultaneously. The market is generally covariant to coordinates used to describe it's price on your manifold of choice, and therefore one description does not invalidate the other. It is akin to saying, I use a thermometer to tell me it is hot, and you use your skin to tell you it's hot. The temperature is the temperature, but you are using a human instrument and I am using a quantitative one. The temperature is the temperature regardless of instruments used measure it.

Your questions comes down to: what is the tensor (or fibration or some mathematical structure we have not invented yet) that gives the map from the intuition/experience coordinates to quantitative methods coordinates. Both methods ultimately exploit structure. If people knew that map, they would be more than rich - they would have won the Nobel Prize in multiple disciplines because it would unify in a single stroke many mysteries of the human brain.

nitro

 

[mingsphinx]

A mathematical definition of structure always implies that certain relationships must hold. Even though we can build in a certain randomness or fuzziness, models being models, are essentially expressions of the associations between variables. The problem is that markets, as one of the most tangible manifestation of the collective human mind, are as prone to sharp reversals as human beings are to changing their minds. We can analogize the mind as a machine, but the analogy will never be perfect and the flaws can lead to serious mistakes. What it boils down to is the free will and consciousness. No machine can replicate what the mind can do. Indeed, if you are so inclined, there are serious philosophical doubts that have been raised about whether it is even possible for something like the mind to arise from inanimate matter for nothing else we see in the physical universe is capable of causation. But algorithms are useful in that large amounts of information can be handily looked at and conclusions reached. The human mind is no match for the computational power of the machine in such instances. Nonetheless, as the machine is unconscious and the intelligence but artificial, the computational strength of the machine must still be guided by the judgment of man who understands mind. To do otherwise invites auto-destruction as the machine is lead by data to commit hara-kiri as lemmings jump off a cliff. And I seriously doubt if there are any luddites amongst us who will not use some kind of machine aided system in their trading/investing pursuits.