Hi, I have a challange in studying statistics. What I would like is an algorithm or functions to produce a list of price levels, where the probability of prices' reversals is high. I think that has something to do with support/resistance but donot want to use Fibs but would like a calculation myself. Anyone can give me a hint please?

There are 2 main camps of statistics and the related theories and concepts are not fully compatibles. 1. The analysis of static / stable / fixed universe 2. The analysis of real series (a sub-category is time series) What you are trying to do is part of #2, it is not an easy subject and don't expect simple answers

agree, I knew that. However your answer is the one that is simple but no less instructive. It's easy to find web resources on the topic, but it would be greate if you recommend your preference?

What you are trying to do can be achieved through a logistic regression model with "price level" being "main" independent variable. By the way, I'm not a statistician.

Thank you wan2BTrader, I think its exactly what I am looking for, but can you give some direct resources as I don't know where to begin? regards, PQDzung

Who are statisticians? The old joke says, statisticians are just mathematicians broken down by age and sex. Martin

I did a series of articles sometime ago using average range off a simple moving average that shows enough moving window stability that can produce trading signals that are consistently working over many years. So such "critical points" can be included into your list of price levels in real-time. It took 3 to 4 full pages to describe how one such concept works. Thats why I said it is too complex, at least for a discussion forum. Feel more comfortable now?

I actually never used any statistical stuff. I was just thinking after read your first post. I don't really know any resource can help in this direction other than my first reaction was that Elder's tripple screen system concept is probably similar, so can be something to look at to just get some idea. If I was going to do this. I would think about a) a referrence point in time b) a referrence price level(t) c) delta time and d) delta price. I would start experiment with those two deltas given those two chosen referrence points and then experiment with new referrence points. Not a recommendation, just how I would start if I was going to work on it. This most likely is not the answer you are looking for since you want to know how to pick those points and algorithm which generates deltas.

As was suggested, a logistic regression could estimate such a probability, but really any classifier (discriminant analysis, neural network, tree induction, etc.) would do. The real trick, obviously, would be in getting the discriminatory power to a level which would make the thing profitable.