I think both the proponents as well as detracters of trend-following on this thread have focused too much on the "science" or "math" behind trend-following. It is obvious that if their was a 'holy grail' such as trend-following or any other method for that matter, it would have been found and applied in such as size as to overwhelm it's own effectiveness. Obviously, there is more to successful trend-following than having the correct algorithm to ride trends or the monetary staying power to hold on. Just as anyone can be taught to paint, anyone can be shown the trend. Has anyone looked at a 3-yr daily bar chart of the Euro? How about the $ in general? With that said, while anyone can paint, there is only one Picasso or Renoir. Just as in trend-following, there are only a handful of true masters. Trend following is more of an art than a science. The fact is that markets spend a great deal of time moving sideways and it is those sideways or choppy markets that a trend-follower (no matter how skilled) will lose money. Knowing when to apply trend-following and to which markets is a subjective call which is often the difference between successs and failure. My point is that there are many 'non-mathematical' aspects to trend following that need to be considered and addressed. It is understandable that a simple mathematical test of trend-following applied to all markets over numerous time frames would not yield any siginficant results, just as in Poker, a statistical model will likely not produce positive returns when playing against the top players in the world. With this said, trend-following has become so popular in the last 10-20 years that one needs to consider the 'edges' that are being thrown off as a result of the huge amounts of capital utilizing the same methodology.
trend following delusion shattered
- Thread starter hank rollins
- Start date
- Status
I think we're just guilty of 128 pages of misinterpretations. This quote is not what I, or the numerous other posters have been arguing here for days; rather we have been arguing this idiotic idea of predicting movement with some jackass equation of |p1-p0| >= d.
I disagree with your statement that trend entries are 50/50, as I have done a fair amount of personal research that proves this wrong. But saying this is defiantely not the same as saying a trend approach has to predict a set movement, or that trends are not tradeable at all.
Today has seen some reasonably intellegent posts on this subject, rather than the ignorant spamming of an equation with no practical meaning.
Regardless, good trading to all regardless of your methodology.
P.S. I apologize for typos in this post.....it's hard to type when you're staring at a T&S display.
i'll be happy to speak to mr. henry.
I actually think NickelScalper asks a valid question. It may be a question that is formulated/executed incorrectly, but it does seem to be a valid question at least what he is trying to ask if I understand correctly. Basically he is just saying show me one trend-following system that works at least to some degree.
Now first let me address the problems that I have with NickelScalper's question. There are at least seven:
1) He is only looking to work off of trend following systems that make use of pre-determined targets. In practice, while there are some of these, many (most?) don't use targets; they exit based off of the price beginning to stall in its trend or other factors that make use of bar-to-bar conditions, as opposed to pre-determined conditions. Obviously, people would not be trading them if there was not some expectation of consistent follow-through. But in many cases what we are looking at is variable "targets" if you will, and to analyze performance requires looking at the distribution of returns and applying some sort of confidence test, perhaps similar to the one EricP has proposed that factors both # of trades, average returns, and standard deviation. Perhaps if NickelScalper can modify his definition from profit target to the average expected return that would be sufficient, as opposed to a hard profit target.
2) Even if some people post live trades, as he is requiring for his demonstration, there are issues with ET's clock and other factors that cause some people to question the validity of the trades. One solution would be for one of us to host a new application that is like an (un?)official ET trade server, that gives a simple Buy / Sell interface with a username. If I had time I could program such a setup that works off my IB feed. Even if the program isn't smart enough to simulate partial fills / missed fills / slippage 100% correctly, it would still be a vast improvement to posting on the boards, I think.
3) The main problem I have with his question is even if we do fulfill his conditions, what does it prove? In fact NickelScalper himself has even said something along the lines of "for every good trend-follower there are 19 to 20 who fail". That begs the question, even if we post 1 functional trend following system, could he (rightfully or wrongfully) say something like "there are over a thousand people on this board and all you can post is 1 system?". So from a philosophical point of view (higher time-frame if you will), there may be issues with the challenge.
4) Similar to #3, we really would need at least 20 trades (preferably 100) to have statistical significance. Thus, it would be possible for an excellent trend-following trader to do a live demonstration, and say net lose money on 10 trades altogether but still be a great trader as that isn't enough samples.
5) Some participants on this board have posted potentially valid trend-following systems or parts of them already. In some cases, these systems may be too difficult for NickelScalper and others to understand and/or forward-test. So an implied additional requirement of NickelScalper's system appears to be that the system is easily understandable and/or makes use of simple calculations.
6) There is no mention of scalability, so I can only assume we are operating under the most favorable (1-2 contracts) conditions, as opposed to 1000 contracts or so (ie a scalable system). This should probably be made clear just so we are all comparing apples to apples so to speak.
7) As trade-ya was saying, can one really say a price is moving "in trend" in an overall objective sense when there are so many timeframes? Price is always in multiple trends at the same time, depending on the time-frame. That is not to say one cannot have a profitable system by focusing on trends on a particular time-frame(s) to the exclusion of others. This is more of a remark than a problem per se.
So those are some of the issues I see with the challenge. Despite the problems, I think there is some usefulness in his challenge, at least from a learning point of view. If nobody can meet it, that gives some information (perhaps a small amount), namely that either 1) successful trend-following systems may be too complex or time-consuming to describe or 2) successful trend-following systems, even marginal ones that are only slightly profitable, may be too valuable to release to the public. But from a logical point of view, I'm not certain we can say that a *lack* of anyone meeting the trend-following challenge means that there are no trend following systems that work well. This is because we have no way of knowing whether the reason is #1 or #2 above, versus truly a lack. So logically it may be difficult to make a conclusion, and in any case this is only for the set of people that comprise of ET's membership, which is a small subset of the worldwide population of traders. Still I think there is some information to be gleamed from this exercise.
-Taric
Now first let me address the problems that I have with NickelScalper's question. There are at least seven:
1) He is only looking to work off of trend following systems that make use of pre-determined targets. In practice, while there are some of these, many (most?) don't use targets; they exit based off of the price beginning to stall in its trend or other factors that make use of bar-to-bar conditions, as opposed to pre-determined conditions. Obviously, people would not be trading them if there was not some expectation of consistent follow-through. But in many cases what we are looking at is variable "targets" if you will, and to analyze performance requires looking at the distribution of returns and applying some sort of confidence test, perhaps similar to the one EricP has proposed that factors both # of trades, average returns, and standard deviation. Perhaps if NickelScalper can modify his definition from profit target to the average expected return that would be sufficient, as opposed to a hard profit target.
2) Even if some people post live trades, as he is requiring for his demonstration, there are issues with ET's clock and other factors that cause some people to question the validity of the trades. One solution would be for one of us to host a new application that is like an (un?)official ET trade server, that gives a simple Buy / Sell interface with a username. If I had time I could program such a setup that works off my IB feed. Even if the program isn't smart enough to simulate partial fills / missed fills / slippage 100% correctly, it would still be a vast improvement to posting on the boards, I think.
3) The main problem I have with his question is even if we do fulfill his conditions, what does it prove? In fact NickelScalper himself has even said something along the lines of "for every good trend-follower there are 19 to 20 who fail". That begs the question, even if we post 1 functional trend following system, could he (rightfully or wrongfully) say something like "there are over a thousand people on this board and all you can post is 1 system?". So from a philosophical point of view (higher time-frame if you will
), there may be issues with the challenge.4) Similar to #3, we really would need at least 20 trades (preferably 100) to have statistical significance. Thus, it would be possible for an excellent trend-following trader to do a live demonstration, and say net lose money on 10 trades altogether but still be a great trader as that isn't enough samples.
5) Some participants on this board have posted potentially valid trend-following systems or parts of them already. In some cases, these systems may be too difficult for NickelScalper and others to understand and/or forward-test. So an implied additional requirement of NickelScalper's system appears to be that the system is easily understandable and/or makes use of simple calculations.
6) There is no mention of scalability, so I can only assume we are operating under the most favorable (1-2 contracts) conditions, as opposed to 1000 contracts or so (ie a scalable system). This should probably be made clear just so we are all comparing apples to apples so to speak.
7) As trade-ya was saying, can one really say a price is moving "in trend" in an overall objective sense when there are so many timeframes? Price is always in multiple trends at the same time, depending on the time-frame. That is not to say one cannot have a profitable system by focusing on trends on a particular time-frame(s) to the exclusion of others. This is more of a remark than a problem per se.
So those are some of the issues I see with the challenge. Despite the problems, I think there is some usefulness in his challenge, at least from a learning point of view. If nobody can meet it, that gives some information (perhaps a small amount), namely that either 1) successful trend-following systems may be too complex or time-consuming to describe or 2) successful trend-following systems, even marginal ones that are only slightly profitable, may be too valuable to release to the public. But from a logical point of view, I'm not certain we can say that a *lack* of anyone meeting the trend-following challenge means that there are no trend following systems that work well. This is because we have no way of knowing whether the reason is #1 or #2 above, versus truly a lack. So logically it may be difficult to make a conclusion, and in any case this is only for the set of people that comprise of ET's membership, which is a small subset of the worldwide population of traders. Still I think there is some information to be gleamed from this exercise.
-Taric
great ! i think he would have much to add to this conversation.
thanks !
A voice of reason.
Thank you for your post. Much has been written in this thread but little has been said (especially in the last 50-60 pages).
Here is my offer: I will take a system from my "archives" that I have no real interest in trading and post backtested results here. I will post Easylanguage code, as well as plain english pseudo-code. All who are interested can track the system in real time if they like. Perhaps this should even become a journal thread.
However, I will not contribute heavily in the posting. We are looking forward to the birth of our second daughter on Monday, and posting signals of a system on here for the glorification of a pissing match is not high on my list of priorities. Trading will be difficult, let alone frivilous thread posting.
I am sitting on my ass tomorrow, so perhaps I can do it then. Again, I am not going to stand by the results or signals. But it will be a trend-based system that will be make publicly available to those who care.
Again, thanks for the always welcomed voice of reason, Ptunic.
Edit: I just noticed this one thread is responsible for around a 30% increase in my total post count.....good lord that's sad.
- Status