Excerpt from Deming's SOPK http://www.maaw.info/ArtSumDeming93.htm "The Theory of Knowledge The theory of knowledge includes an understanding that management in any form is prediction. A statement, if it conveys knowledge, predicts a future outcome including the risk of being wrong. Prediction requires theory. Without theory, experience has no meaning and there is no learning. Copying examples without understanding the underlying theory may lead to disaster. Any number of examples cannot establish a theory. "
I must agree that failure is richly rewarded on Wall Street, although failure is a relative term since someone is making moeny, just not investors. The car analogy is flawed because you can see ahead quite clearly. The proper analogy would be driving down a cliff-side highway in total darkness with no lights. It might a very good idea to watch your wheels carefully. Certainly looking in the rear view mirror and extrapolating ahead would be disastrous.
This can give an idea of the model results on stock : this was the kind of analysis I made 2 years ago. Stock is far from as precise as indice which is statistically understandable (since the standard deviation of an indice is about to be proportional to those of a stock by the squareroot of number of stocks) . <IMG SRC=http://www.elitetrader.com/vb/attachment.php?s=&postid=245180>
Well said. This guy makes money from writing not trading. So he will write more.. and more... and more...
Regardless of true/right and false/wrong... it was interesting reading... Let's try to keep an open mind...
Is DS full of BS? I REMEMBER DISTINCTLY and I even have it printed out over the summer that there will be this HUGE drop in teh SP500 starting in Sept and Oct! I swear I saw it on his homepage. Then when you check it now, he took out that graph! WTF?! And then he extended the prediction til 2004. And if 2004 doesn't have a big drop he'll extend it to 2005. haha! I mean he might be right eventually that between now and 2004 there will be a huge drop in the SP500. But long term forecasting can have error terms of a FEW MONTHS or QUARTERS! And that would be disastrous trading off of that variance! Imagine him advise some institutional/hf managers to put on a huge short position in Aug. They would have been fried with this up move even though he might be right eventually... I think simple TA will tell you when and IF the big move comes. When it comes you'll see. Predictions are useless. With all that effor and analysis expended on figuring th next big move, he's missing the up move in the last few months. Just follow the market. What do you think harrytrader? I know you are a fan of predictions...
I thought I saw a nice little graph with a rally and then a big dump in the fall too. If he did remove the graph and rewrite his dates and I were an academic I would make sure this intellectual dishonesty was well known. Of course he will talk about how his model had bad data and upon data scrubbing the model spit out this new graph. Really all he needs is some good prechterizations.
What I think is that stochastic modelling of stock market are too poor - including the last co-integration type of model. They are adapted for global risk assessment of investors type profile or arbitrage / volatility strategies but not for some type of speculative trading with huge leverage where precision is critical. The stock market must have a very weird correlation structure as would say Mandelbrott - since he said that for the Nile - that stochastic models have difficulty to capture, I have explained globally why here : http://www.elitetrader.com/vb/showthread.php?s=&threadid=23134&perpage=6&pagenumber=21 "In econometric modelling - as in other fields - there are normally two class of modelisation : - stochastic (from the simplest like ARMA, to more complex like Box & jenkins that are classically used in short term forecasting of products like champaign - I said that because it is on that that I learned them - and in financial modelling they have been extended with Garch and so like models - see the link I about Jean Philippe Bouchaud audio conference) and - deterministic. The research on deterministic models is much recent in forecasting techniques and according to a statistical book "Methods for short term forecasting" (translated from french) it has only begun since the 1990 and at the moment nobody has never found something very convincing in financial modelling and one of the reason is that tools for detecting deterministic models are not very accurate. For example in another book I already quoted which is "modelisation of stock market returns" (sorry also translated from french) they show that the market must have some deterministic components but that the lyapounov exponents - which characterise such process - are too weak to be able to assess the hypothetical model. So by default they try to model long term memory effect with stochastics models that extend the classical ones above (farima, wavelet decomposition). Is it astonishing ? No because to get a deterministic model you must know the real factors that is difficult since it would be like trying to read in the mind of God's market . And even using more sophisticated stochastic approach won't change the instrinsic wall of uncertainty attached to stochastic model not because of absolute impossibilty but because of the structural form of the class solution. Of course it will be possible to detect interesting things but it will stay shadows compared to a true deterministic model."
Harry- while I still do not completely understand what people say when they throw the word around deterministic, at least now I know what they think they are saying ior n the rare cases what they are saying. Harry you are either getting better at writing in English or my filter has adapted well, because your signal to noise ratio has improved dramatically. thanks