trend exist, we can't say that in any current moment, but afterwards looking back we could see - it was a trend!
Probably it is possible to be certain about a trend (or predict it) using Austrian cycle concept or law of inertia.
You are telling the truth that "trend" is an abused word and that it is not really scientifically defined, or there is no consensus definition of "trend". Without a clear definition, any discussion over this topic is probably a waste of energy.
I also ponder that whether we do need a definition of trend. The more fundamental thing, behind trend and many other concepts, is probability.
Instead of asking is the trend up or down, or is there a trend, I think it is better to ask "what is the probability the price will go down by 1%, or 2 points, in the next hour, or next day, or next week"?
Instead of telling me the trend is up, it would be more useful to tell me that "within the next hour, the probability of seeing prices 2 points higher is 67%, with an MAE of 3 points."
Probability is represented by a hard number so that thinking in probabilities really forces ourselves to focus on numbers, not abstract concepts that cannot even get a definition agreed upon.
I couldn't agree more. If you follow my trading journal you'll know that my goal has become to develop automated trading systems. My first one (BOT1) I've had running for a few weeks now works pretty well because it's looking for statistically high probability opportunities that don't happen very often, but when they do I often win. The reason I automate it is because I don't want to sit around for hours, sometimes days, waiting for these somewhat rare conditions to occur, so I have the computer do it for me. Now I'm expanding it to other markets and hopefully someday soon I'll have a bunch of bots all monitoring several markets, looking for statistically aberrant price action and taking advantage of it.
I've also been trying to develop a trading strategy that can be consistently profitable detecting, following and/or scalping trends, but, like many have said, trend is hard to detect early on, and often times you get in too late or get out too early, or you shouldn't have entered the market in the first place and you get tossed around inside a bunch of chop. At best I've gotten 55% win rates with profit factors barely over 1.0 which doesn't get me too excited. It seems the only way to get more frequent winners is to increase risk, often at the expense at overall profit.
future probabilities are not known, so at best person proposing new system should provide statement like this: this set-up is giving you this probability getting this target (of course it is not real probability, but probability based on previous occurrences). I have seen this statement only at www.elliotician.com
or for open gap plays or TRO play.
Last edited by record100; September 21st, 2009 at 04:38 PM.
There is a clear definition of trend in the statistical sense. It is defined as drift away from the series mean on a given scale. The fact that financial data contain trends is unquestionable, it is part of the reason analyzing them is so difficult because trend implies nonstationarity, ie the mean is not stationary and is constantly changing. That renders many common statistical analysis techniques invalid (including standard tests for trends in data)!
The question with trend of course becomes, what scale are you looking at? Trend is very different on a small scale of minutes compared with trend on the weekly scale or monthly, etc..
You mention also the fact that probability is represented by a hard number. This is one of the major flaws of standard probability theory, and the impetus behind the existence of things such as bayesian probability and fuzzy logic. Fuzzy set theory asserts that it is foolish to attempt to model uncertainty with a hard or 'crisp' number, rather it suggests the use of a 'fuzzy' range of values with a 'membership function' which describes how much weight to assign each value in the range.
Standard probability theory would allow you to say something like: there is 70% chance the series is trending right now based on movement within the past hour.
Fuzzy set theory would allow you to say: within the past hour, chance of membership in set 'not trending' is 0%, chance of membership in 'weak trend' is 20%, chance of membership in 'moderate trend' is 70%, chance of membership in 'strong trend' is 10%
The real beauty of fuzzy sets is that due to the way they are defined, you can easily combine them to come up with a composite forecast which is ultimately more informative than what can be obtained by the means of standard probability theory.
The following 2 users say Thank You to sefstrat for this post:
Just to come at this from a different angle, I identify trend as any identifiable action that has relatively predictable results.
Most people think a trend is defined by "is the market going up or going down"? Personally, I don't care about market direction. I only care that it changes, and it always changes.
Some people count the patterns of up and down bars to identify entry into a trade. They have identified a trend that happens when certain patterns form. Other people wait for a strong run in price and then wait for a pullback. That's simply because they've identified a trend under those particular conditions that tilt the odds in their favor.
Lets say you're watching people enter a mall through a main entrance and decide to monitor their movements. You discover 62% turn left and 38% turn right. You've discovered a trend.
The following user says Thank You to hondo69 for this post: