No offsense, but dont read the post if you dont like my form. Predicting EOD data is insane, like forecasting the weather a year in advance, unless it is something semi-periodic like El-Nino.
I'll be nice this time. 1st differencing loses all trend information, that is why it is a bad transform. It is typically done because it is easier to work with stationary data rather than non-stationary prices.
stephencrowley, You are contradicting Mandelbrot. Mandelbrot has always modelled price fluctuations as a fractal but unpredictable process. He hasn't excluded the possiblity that price changes can be predicted, but none of his models or theories even attempt to provide any directional price predictions, or to suggest that such would even be possible. If my information is out of date, then please post an exact page number and an exact quotation supporting your position. Mandelbrot has never, ever claimed that all fractal processes can be predicted using the method you describe. I challenge you to give an exact page number and an exact quotation where he says such a thing. I would go even further, to say that no reputable, respected scientist has ever made such a claim. My guess is that you don't really understand what the word "fractal" really means. You claim that the embedding theorem, due to Takens, applies to price series, but Mandelbrot never made such a claim, and neither did Takens. The embedding theorem only applies to certain types of systems. Nobody has ever demonstrated that price fluctuations fit the requirements and assumptions necessary before the Takens embedding theorem can be applied. Nobody has ever demonstrated your other claim, that price changes exhibit low-dimensional chaos. The foremost experts and founders of chaos theory made extensive efforts to demonstrate this, but then gave up on the effort. These include people like Farmer, Packard, and others at the Prediction Company. Mandlebrot has also never claimed that price changes show low-dimensional chaos. If my information is out of date, then please post an exact page number with an exact quotation. It is possible that prices do involve low-dimensional chaos, but this has never been demonstrated. If, as I and many others believe, prices are somewhat predictable, this does not imply that they also have low-dimensional chaos. It is not necessary to have low-dimensional chaos in order for prices, or any other series, to have some predictability. My suspicion is that you have no real understanding of the scientific literature you are discussing, and that you are throwing around jargon without knowing how the terms are actually defined. This leads me to doubt the validity of your claims that you are using such literature as the basis for price predictions.
Yeah I don't understand any of it, I type in random shit and somehow these accurate price predictions coming flying out of my computer. I decline your challenge, go do your own work, I'm not giving out any freebies. Also, taken's thorem cannot be applied to the raw price series, but I leave that as an experment for you.
============== Since you said forcast, not prediction; YM & ES are in nice uptrends, probabilities favor continuatuion, especially ES. YM is much more extended for sure than ES; nice thing about both those markets;usually at least one good uptrend or downtrend perday ,or more And reguardless of Monday; still true, trend =friend
Probably you wouldn't be the only person (trading or not) doing forecast on market movements. Sometimes it could take a very long time to translate a (really successful) forecasting model to real profits, if any. Have you started a journal on ET yet, as you previously mentioned on another thread? Why not?
The essence of Mandelbrot's work with regard to application in the financial markets can be summarized as "phat tails." And I don't mean the chicks at the bar with big hooters neither. And yes there is applicability. Look at this week's time-price series in the equity indexes as an example. Statistically compared with the previous n-days this run was essentially a "fat tail." It may not be the "fattest tail" that would have been 1995 through 2000, and 2000 - 2002. Or a better illustration the market implosion of 1987, yes? Now as for turning his observations and mathematical models of "Chaos Generators" into the holy grail, I don't know about that. Probably I'm not smart enough. But this stuff is interesting.
Yes, I agree, except that I do not believe you are making accurate price predictions. Sellers of worthless price prediction methods frequently make false claims that their work is somehow connected to Mandelbrot, fractal geometry, or chaos theory. It is purely a bogus marketing strategy. If you would care to provide some evidence that you are not just one of these sellers, I would read with an open mind. I will, until then, based on your incorrect descriptions of Mandelbrot's work, conclude that your claims are false.
And yet again someone accuses a poster, who has yet to post anything about selling ANYTHING , that they are a system seller. Lets jump on Crowe when he actually posts about selling something. Stephen> You will get better feedback over at NuclearPhynance than here (enjoyed your cointegration thread there btw).
Well, my previous model was not making any directional forecasts at all and I felt that I probably could be doing that as well and combine them. So I opted to do some more development before going live. I'm happy with the price forecasts now, and as you said, it is one thing to have forecasts and entirely another to turn this into trading signals, stop-losses, position sizing, etc. So, that is the last hurdle, and then of course, putting this into my trading platform rather than the hacked up mess of matlab code.