Can linear regression analysis really predict the future?

There is one issue I have involving predicting future is that my software package amibroker does not seem to handle data or graph beyond realtime(current time). Any contradiction or alternative suggestion are appreciated.

Thanks
 
Quote from TheGoonior:

As with any highly non-linear system, the further you move away from the operating point, the worse your model becomes. I'm skeptical that linear regression can consistently give you an edge (although I'm sure there are those who make it work for them...my hat's off to you, as I never found it useful).
As a school project, I did some work in recurrent neural networks as predictors for non-linear time series. These are highly adaptive (they were re-trained after every data point) and actually modeled the time series (I grabbed some speech waveforms) and I always meant to go back and see what would happen with stock data.
It was kind of a CPU hog and there was an art to training the networks, so I never pursued it.
Anybody else ever use neural networks for this type of thing? (Hope this isn't too far off topic, so OP let me know and I'll move to another thread if necessary)
More...
Yes and NNs are of no interest to me, for a variety of reasons...

The beauty of simple regression is its intuitiveness. Normally, people are happy to sacrifice some degree of robustness.
 
Quote from Martinghoul:

Yes and NNs are of no interest to me, for a variety of reasons...

The beauty of simple regression is its intuitiveness. Normally, people are happy to sacrifice some degree of robustness.
More...

Although reversion to the mean is probably one of the most fundamental and stable observations of the stock market behavior NN algorithms are not the solution. Our research in this respect shows us that an algorithm that utilizes reversion to the mean phenomenon with consistent positive expectations is best implemented with interpolators. Of course, as everything that is practical and efficient our approach is surprisingly simple and intuitively acceptable. Below is the summary of principles that we have used to develop our trading algorithms.

1. Most of significant market moves caused by the phenomenon of “Spontaneous Synchronization” where the prices move irrationally too far and too fast creating stable “panic feedbacks” that ensure that the price volatility sustains. Those moves increase “price inertia” and make position of the price center of gravity fairly stable

2. Price movements involve collectives of traders that behave like large synchronized flocks. In these flocks the average distance between the flock members and the flock’s center of gravity remains fairly stable.

3. Flock’s center of gravity normally moves in a pattern that vividly exhibits inertia properties of a flock.

4. Subconsciously observing the movements of the price center of gravity market participants synchronously anticipate its next position and place their bets with this observation in mind thus forcing the price to move towards the projected position of the flock’s center of gravity.

5. Natural cubic splines calculated on a sequence of consecutive center of gravity positions create a very reliable expectation of the next center of gravity location in terms of price/time space.

6. The distribution of the actual price fluctuations around a spline predicted point is always Gaussian with the standard deviation that is typically smaller than a price deviation over one time interval used to calculate the center of gravity itself.

7. Price reversion to the spline predicted position of the center of gravity normally has greater price differential compared to the deviation of the center of gravity itself thus ensuring stable positive expectations for the mean reversion trading algorithm.
 
Natural cubic splines calculated on a sequence of consecutive center of gravity positions create a very reliable expectation of the next center of gravity location in terms of price/time space.
More...

Are you sure you mean 'Natural cubic splines' not 'Smoothing Splines'?
 
Of course, I am using it to determining the trend, nothing will predict the exact value, no questioning of that. But LR cannot even predict the trend significantly over 50%. The best I have so far, after pretty much frying up my machine, was a 2 week (10 trading days) trend prediction of the S&P 500 index with 59% accuracy tested from 11/3/09 back to 2/4/09, using over 20 predictors and multiple lags from each predictor (yes, the matrix has hundreds of columns). I have another 3 day trend prection on S&P 500 using the same predictors plus S&P 500 's own past 3 day closing price (because for 3 days, there is autocorrelation effect), that has 62% accruacy, back tested for the same period. What I found so far is that if I drop the predictor/lags that individually have lower correlation to the S&P 500 index compared to other predictors, I actually lower the accuracy of the trend prediction, contrary to common sense. But if I add more and more even remotely related time series, accuracy creeps up. Hence do we just need a supercomputer with every single piece of data imaginable? With slightly better than 50%, one has to not pay commission and use binary options or some sort to even have a trading system.


Quote from bwolinsky:

Linear regression as I use it is for one of two things:

1)To assign a value that the market may place on 1 more dollar of revenue, assets, 1 more percent of ROE, etc

2) The other way is by regressing closing prices over recent time periods to predict tomorrows value, two days from now value, and three days from now value, not because I actually use that value, but because linear regression shows the "trend", and that's how it's meant to be used in the context of trading system development. If your predicted value is higher two days from now, the trend points higher, if your predicted value for tomorrow is lower, the trend is also lower, and you shouldn't fight it. For a history of QLD's predicted values, you can see my thread BWolinsky trading. Today, I will most likely post more predicted values, as I said, not because they mean anything to me other than that they keep you from fighting trends in the market. It's corollary is testing fitness. Using rsquared, we can see how accurately those values hold, but it's only relevant over short time periods.
More...
 
Quote from tradrejoe:

For those of you who went through the exercise of using historical data and linear regression analysis to predict the future prices of trading instruments, have you ran into situations where the best beta coefficients that generates the best curve fitting *does not* really predict the future? In fact, often times if you go back in history and pretend you were operating the prediction system in the past, the more testing the more your accuracy converge to just 50%?

What is the correlation between the ability of a set of time series data to fit a price curve and its ability to truely forecast the future with greater than 50% accuracy? Do we just pile up everything closely related to what we try to forecast (even sun spot movements) and go as far back on the time lag as we can without crashing the supercomputer? Does anyone have any experience to share? Thanks for your insights ahead of time.
More...
Nothing can predict the future.
 
Quote from crgarcia:

Nothing can predict the future.
More...

I used to believe that, I am still skeptical. That said I have seen a software program which is about 70% accurate on direction using fourier analysis and time series forecasting.
 
Quote from dandxg:

Hey nice to see a quality poster, MAESTRO, back. Nobody can stay away Gnome lasted 1 hour. :D
More...

I am an educator by nature. If I have time (and feel healthy enough) I do not mind to share the bits and pieces. I am a big believer in education and overall good of probabilities understanding. However, I am not in a position to be a part of any war of “dicks” that is why I rarely post..
 
You must log in or register to reply here.
Back
Top