Can this formula be applied to charts? Mathematicians...please help

[kartthik]

Hi,
I was searching for an equation or a formula where in path of a moving object can be predicted. I found many articles on the web and here is one example http://repository.cmu.edu/cgi/viewcontent.cgi?article=1336&context=robotics. I couldn't decipher the formulas however learned that they must be applied to a smoothed data. Now, considering price as a moving object and a moving average as a smoothed data curve is it possible to predict the course of the price? any suggestions from mathematicians?

Thank you

 

[kut2k2]

My crappy tablet can't read .CGI but just based on the key words "robotics" and "moving object", I can make some guesses.

Real physical objects have inertia. Price doesn't have inertia. Noise or no noise, price can and often does "turn on a dime". If the article makes any assumption at all of inertia, it will be useless to you.

And if the method of smoothing is a conventional moving average, there is a built-in lag which must be overcome so your "prediction" may only serve to bring you back to the present, with no future insights at all.

Jm2¢

 

[vicirek]

This paper is not about finding single prediction "formula". Smoothing is just one minor ingredient in it because of the known concept of noise that authors are concerned with and they are using Gauss (bell curve window) to smooth data which is conceptually more like exponential moving average than simple moving average. This is prediction model based on clustering (machine learning) and probability models using curve segments. Segments are normalized and scaled for the purpose of analysis. First they use clustering which is part of machine learning to classify segments and then use probability of transition between states using Bayes and Markov model to predict next segment based on derived parameters. All hinges on the hypothesis that current segment occurrence depends on previous state/s (segment/s) and the transition probabilities can be applied here. This is probabilistic model and definitely can be tried on time series. It cannot predict future price movement but can find probability of next possible price movement. The proxy for such model are thousands of "strategies" being discussed on this forum and elsewhere because they are based on parameters derived from short price movement segments. This should give you some idea about possibility of applicability of such models to achieve consistent market performance. However, this is relatively simple model and if you have time experiment with it or improve it.