Reasons for failur in automatic trading systems
- Thread starter cohvi
- Start date
no offense intended but your five lines remind me of spam i'm currently receiving. these mails are a compilation of very interesting words and fragments - so interesting that it takes a couple of glances to find out that the whole thing does not say or mean anything.
your five lines are so general that they cannot be verified or falsified. which is fine, as long as one just small talks about markets, life or whatever. yet, if one wants to use information to dig further, this small talk is completely useless.
having said that i do not doubt at all that you know something and trade successfully upon it. your internal CPU can be in perfect shape, but it seems to me, that whenever you are broadcasting my station does not at all capture your wave length ...
ok,
What I meant was that the model was developed on a small scale when I had only 3 months information - 1st month was the pivot month from which the model learned, and then implemented its conclusions of how to trade on the 2nd & 3rd months. In the same logic I backtested a years back - a period of learning and afterwards the trading.
Correction: Pips -> I meant ticks. I'm not trading forex.
My intention was to just pass on some information on the topic.
I knew that most people who use maths to solve problems are involved in the conventional gaming theory and its branches. I was a key speaker in the field (ISAGA) in the 70's.
People who have long involvements in particular endeavors do trace a path through several levels of change.
Gaming is surrounded by technological innovation and a broadening of infomation supplies.
The power of the maths, however, may be stagnating as you see from the responses here by those who have noted the names of things. Most who have been using those names have also gone to other places with differing lists.
A Lo citation was made recently. He goes for AMH and it is in contrast to EMH.
I was taken by the threads commentary as it represented a collective group coming up against a limiting condition. Man stepped over the line somewhat (meaning he announced that better Sharpes lay elsewhere), so I sat down and tapped a spectrum of comments that characterized some considerations that either were on the mark or that represented common traps encountered by a good portion of practitioners. I passed on the airhead stuff.
The OP was was very significant and "seeking" in the brief articulation he scoped and bounded the "problem" or in a way the "opportunity". I liked it.
My tapped list also correlated with the five topics of the OP.
On a yellow pad I made a schematic of an operating system that has Sharpes that greatly exceed man's experienced observations and I knew the other contributor's Sharpes as well.
So I sat down with a yellow pad and drafted a set of paragraphs. Dictated them. Got the transcription. And posted it. It was a rough (below draft) and not paragraphed not checked for syntax or clarity.
What I didn't do was use the 28 comments as referenced footnotes by adding them to my words as direct references for your convenience. I didn't reread anything. I didn't use the schematic as an illustration. I ddn't add a list of the 70 DF's that is laying on my drafting board. I didn't add a list of the 61 signals that emerge from the DF's that is also laying on my drafting board.
This unsatisfactory effort was just a pick up job where I tucked in facts parenthetically. If I had buzzsawed the backup as the attachment; it would look like the construction drawings for a commercial building. ET can't handle infomation to any extent as we know.
I don't believe teaching works. At best learning can be supported in several ways. So I do not teach.
Here what you see is a compare and contrast where I make some points in a given OP context that has 5 elements.
The better math for making money is not Pascalian; in contrast, it is logic.
I posted a first pass and got a request for what I thought was a very brief abstract. The bullshit response I got for the effort was three words. Then it was backed up with a comparison of the five lines to the normal fare of scamming. I saw this after your evaluation of my poor exposition.
My attachment to which you refer was just a very staccatto response to a person at inquiry on the contents of a black box.
I filled the box and compared it to the boxes of the respondents here. The way boxes are now filled is to use a standard pile of hardware and code up the operation of the hardware.
If you want better results, you look in the existing box and make changes. The mete on the output is a Sharpe ratio and a clock that shows time in the market getting the Sharpe results.
Lets look.
Pascal, et al constrains unbearably and it representts a win/lose application to a full blown continuous varying opportunity. The Sharpe is _______ and the time in the market is ______.
These two values can be determined to be of a quality that is not close to what the potential of the market offers. What is the fix?
Run parallel sharpes? Yes of course and be sure capital is used alternatively. How many? With 80 choices, just go down the list (ranked by Sharpes) and keep a little space in between or use a cross over or priority strategy.
I jacked up the math and replaced it many years ago (Google papers in the early 70's, US and Europe, principally GER and SWI) and turned to modelling (logic) and "switches". This was a direct result of learning that the market operating point migrates instead of jumps and that the migration is a result of "blocking". See the "nine" comment in this thread.
The gaming goal sets were the causal problem. Extracting the potential offered is the replacement. The most significant logical hazard is that the relationships are orthagonal and not, like Boolean folk,desire, opposites. Gaming is played with opposites and the market does not operate in opposites.
Markets exhibit many parallels to electromagnetic theory (the orthagonal relationship of electric and magnetic fileds) and in particular alternating current theory.
Radar systems used quadratics (three separate radars) to send timing signals by solving two quadratics simultanoeusly (conic sections of Pascal).
Bandwidth was a key to accuracy and Braithwaite made a lot of stuff clear.
That stuff gets left behind too as better tools come into view mathematically speaking.
So a person can build on Pascal, periodic functions of field theory, bandwidth, et al. That gets us through the 50's.
Throop and MIT guys beat Las Vegas and my records there are measured in how many times an hour they cricket to change the dealer and how pissed off others at the table get with me. I learned to be on the right side of the deck. Counting an unshuffled pile of descrete elements. Don't get carried away with anything here. This is pascal and rules intersecting and the outcome is being on the right side of the action.
Human behavior is the consideration for making money. Markets connect humans each having objectives and voting with capital.
in Blackjack I watch the cards, in poker I watch the players.
So my mathematics is just an extention of how I watch the players. An input of 6 degrees of freedom gives me a very good continuous deep look.
I need about 70 clusters of functions to "see" well enough. What I watch is "when". I can only look at NOW and the past. I watch "when" come along by regarding "migration" and "blocking" of alternatives.
Everything that is done is calibrated to human levels of action and interaction. I am always at the front of the line for executing.
I feel it is logical to always stay on the right side of the market and to extract profits by giving up the spread. This is done by using conventional logic in ways that primarily mimic what is or will be going on. I look at what is going on and what has just happened. These thre items which are symmetric about the present are the sufficient setting for the future to move to the present.
My objective is to be on the right side of the market at all times and as time passes periodically extract from the market what is bing offered. This is a high Sharpe set. For me "drawdown" is on the other side of the market.
Primarily, I use vectors for signals. Adding another dimension to a slug of info is helpful.
Three differing duration MLR's have pairs of angular velocities (divergence/convergence). It would be like the red, green and blue of InvestTools index indicator display charts. The various duration colors move about and, in concert, they have meaning for practioners. Today, they say the next pitch willlet you hit a few homeruns.
My MLR's pinwheel to tell me what to do when I am at bat. An FBO is fun to watch, for example.
My lousey post, roughly stated is not a teaching deal. It is just a way of showing how using an altenative to Pacal et al stuff gives a much much higher sharpe result and much much more time in the markets making money. A common goal of traders.
It is a paradigm that addresses continually effectivenss and efficiency of extracting the continuously changing potential that the market offers.
My being from Mars ((or outer spece, more generally speaking), you can just read the top of my posts and chuck it in the recycle bin. I am just a pragmatic person now wedded to anything and I continue to enjoy improving what I do and so it is all unbelievable until you have the software and run it.
Its a good idea to pass on what I am interjecting. I just took a shot at an alternative to the convention.
As I see it the money is there; why not take it off the table as it becmes available. I just cannot imagine a person sitting in a drawdown position. For me its like looking at a position which involves risk and simply deciding that a prior action was, apparently, a mistake. The rationale for sitting in drawdown is quite clear because there may be a future payoff for doing so over the long run. My alternative view is to make what is offered as it is offered.
Natually, as you point out my view reads like a scam. Terrific, think of me as a scammer and ridiculous.
For others, they may find that they could have an opportunity to think about something.
Sorry, I forgot to keep it on your preferred scooby dooby. Would you like a bitstream version for laughs?
I love making up stuff. Its is fun to look over the forums and threads and think about a topic and then just dictate 10 or twelve pages. We almost named this "dragonflies shitting in the parsley". One of the staff looked out over the pool and was commenting on why they were different colors. I replied because the green ones were eating parsley out of the garden. You can make the connection to the title, I bet.
I really dig your journal...sometimes I feel like giving you some tips for two days out by pm. It would really set you up as a visionary type Cramer who runs ahead of the curve. We could call them the "scooby doos" for entries. And, then, for exits, the announcement would be "scooby poos". Let me know.
What color are the dragonflies in your pool? We have three colors: blue, green and bronze. The green one really loves pussy (and parsley).
Most people are automating patterns found in the market, these patterns are simply behavior patterns of human beings, behavior patterns are consistent as an individual but they change drastically when in a crowd, or when you are drunk, or in trouble, etc, behavior patterns are very easy to manipulate.
Even you yourself can detect and find behavior patterns in your Mom or dad, close friends, brothers sisters, etc. but you will also start to see bigger patterns, ones that tell you a little bit more about what is going on, I knew a guy who showed patterns of getting drunk several times a week, I did not know what day it would be but I knew their was a pattern, but over a long period of time I noticed that he would just get smashed once every couple of weeks, then I realized that he got smashed on friday of every two weeks because it was his payday, I was able to narrow down the pattern during a time their was a very high probability that he would be drunk.
Basically patterns in the market need to be exploited at certain times, the pattern needs to sync with a cycle in the market, that is what I call an edge, if you cannot find out what makes your pattern valid, it's just another random habitual pattern of the crowds.
I have a system that is focused on validating the probability of each pattern programmed, as opposed to just trading them everytime they appear. Systems work as well as the logic, and their failure rate is tied directly to the person providing the logic and the failure rate of most traders is very high..
For instance, I had a big problem with telling the computer that if it did not make it's 30 point target that it was ok to take 29 points, or 26 points as opposed to not hitting my target and turning around to break even or worse a stoploss.
How am I supposed to know if my target is going to get hit this time or not? so I had the system programmed to choose it's own exit points as soon as it entered the trade and gave it multiple options on how it can exit. The system might choose to exit at 50 points some trades and then other trades it's ten points, I don't know what it's going to be just as i did not know how many points it would travel when I used the fixed point targets. it also sets it's own stoploss points now, and they are also different every trade.
To say that Automated systems do not work seems ridiculous to me, the system needs to be adapted as time goes on, either you have to adapt it or it has to adapt itself.. The only thing constant is change, how can you use a cookie cutter method to trade?
Your brain calculates scenarios on the fly so efficiently that you can hardly remember everything you thought about to make a decision at any given point, so you take for granted what it takes to get a computer to trade a very simple pattern using logic.
If you want to expand your ideas on how to better design trading systems you should study Crowd psychology and Fuzzy logic...
I hope this wasn't to cheesy for you all..
thaks for readin this long ass REPLY!!!
Even you yourself can detect and find behavior patterns in your Mom or dad, close friends, brothers sisters, etc. but you will also start to see bigger patterns, ones that tell you a little bit more about what is going on, I knew a guy who showed patterns of getting drunk several times a week, I did not know what day it would be but I knew their was a pattern, but over a long period of time I noticed that he would just get smashed once every couple of weeks, then I realized that he got smashed on friday of every two weeks because it was his payday, I was able to narrow down the pattern during a time their was a very high probability that he would be drunk.
Basically patterns in the market need to be exploited at certain times, the pattern needs to sync with a cycle in the market, that is what I call an edge, if you cannot find out what makes your pattern valid, it's just another random habitual pattern of the crowds.
I have a system that is focused on validating the probability of each pattern programmed, as opposed to just trading them everytime they appear. Systems work as well as the logic, and their failure rate is tied directly to the person providing the logic and the failure rate of most traders is very high..
For instance, I had a big problem with telling the computer that if it did not make it's 30 point target that it was ok to take 29 points, or 26 points as opposed to not hitting my target and turning around to break even or worse a stoploss.
How am I supposed to know if my target is going to get hit this time or not? so I had the system programmed to choose it's own exit points as soon as it entered the trade and gave it multiple options on how it can exit. The system might choose to exit at 50 points some trades and then other trades it's ten points, I don't know what it's going to be just as i did not know how many points it would travel when I used the fixed point targets. it also sets it's own stoploss points now, and they are also different every trade.
To say that Automated systems do not work seems ridiculous to me, the system needs to be adapted as time goes on, either you have to adapt it or it has to adapt itself.. The only thing constant is change, how can you use a cookie cutter method to trade?
Your brain calculates scenarios on the fly so efficiently that you can hardly remember everything you thought about to make a decision at any given point, so you take for granted what it takes to get a computer to trade a very simple pattern using logic.
If you want to expand your ideas on how to better design trading systems you should study Crowd psychology and Fuzzy logic...
I hope this wasn't to cheesy for you all..
thaks for readin this long ass REPLY!!!