Forecasting prices, and Mandelbrot is right.

[stephencrowley]

Without giving away specifics, do you forecast price changes, price direction, horizons, etc? I used to think this was impractical but I've changed my mind.

Taking first differences of prices is a very high frequency transform which removes lots of valuable information, as a test I used a moving-detrend by taking differences from a 5-minute moving average and then forecasting those.

Here is an OUT-OF-SAMPLE forecast, this forecast was generated from data previously unseen and is unbiased. Not shown here is the increasing error bounds as the horizon increases. Also, no, I didn't just pick a particular starting point that had a good fit, the results are robust from most any starting point.

How did I do it? All I can say is that Mandelbrot is right.

 

[kut2k2]

Quote from stephencrowley:

Without giving away specifics, do you forecast price changes, price direction, horizons, etc? I used to think this was impractical but I've changed my mind.

Taking first differences of prices is a very high frequency transform which removes lots of valuable information, as a test I used a moving-detrend by taking differences from a 5-minute moving average and then forecasting those.

Here is an OUT-OF-SAMPLE forecast, this forecast was generated from data previously unseen and is unbiased. Not shown here is the increasing error bounds as the horizon increases. Also, no, I didn't just pick a particular starting point that had a good fit, the results are robust from most any starting point.

How did I do it? All I can say is that Mandelbrot is right.

:eek:

Intriguing

But is being able to forecast the smoothed change, as opposed to the actual change, of practical value? I'm not sure.

Why do you say first differencing loses information? Unless a trader is doing the ol' support/resistance dance, price change matters more than price level IMO.

My personal goal is forecasting price direction. That seems simplest and gets right to the question of buy or sell. Good trading

 

[bgp]

i would like to see monday forcast of e-mini.

thanks,
bgp

 

[OddTrader]

Quote from stephencrowley:

Without giving away specifics, do you forecast price changes, price direction, horizons, etc? I used to think this was impractical but I've changed my mind.

Taking first differences of prices is a very high frequency transform which removes lots of valuable information, as a test I used a moving-detrend by taking differences from a 5-minute moving average and then forecasting those.

Here is an OUT-OF-SAMPLE forecast, this forecast was generated from data previously unseen and is unbiased. Not shown here is the increasing error bounds as the horizon increases. Also, no, I didn't just pick a particular starting point that had a good fit, the results are robust from most any starting point.

How did I do it? All I can say is that Mandelbrot is right.

So you call that a forecasting. Then show us what's not a forecasting.

 

[jimrockford]

Quote from stephencrowley:

Without giving away specifics, do you forecast price changes, price direction, horizons, etc? I used to think this was impractical but I've changed my mind.

Taking first differences of prices is a very high frequency transform which removes lots of valuable information, as a test I used a moving-detrend by taking differences from a 5-minute moving average and then forecasting those.

Here is an OUT-OF-SAMPLE forecast, this forecast was generated from data previously unseen and is unbiased. Not shown here is the increasing error bounds as the horizon increases. Also, no, I didn't just pick a particular starting point that had a good fit, the results are robust from most any starting point.

How did I do it? All I can say is that Mandelbrot is right.

Mandelbrot has written, for decades, that price changes are unpredictable. You are contradicting Mandelbrot, by claiming you can predict prices. Yet you say "Mandelbrot was right." Could it be that you misunderstood Mandelbrot? If you did correctly understand Mandelbrot, would you please cite the Mandelbrot publication which confirms what you are saying?

 

[stephencrowley]

Yah of course its useful.. when you get down to the highest frequency of raw ticks, it's not that useful because you are then trying to forecast the mostly-unpredicble actions of a few people/bots/whatever. The aggregate behavior us much easier to see.

The weatherman doesn't tell you the exact forecast down to sub-degree accuracy a week in advance does he? No.. he gives you a range and some confidence interval.. same thing here, chaos theory and all.

Quote from kut2k2:

:eek:

Intriguing

But is being able to forecast the smoothed change, as opposed to the actual change, of practical value? I'm not sure.

Why do you say first differencing loses information? Unless a trader is doing the ol' support/resistance dance, price change matters more than price level IMO.

My personal goal is forecasting price direction. That seems simplest and gets right to the question of buy or sell. Good trading

 

[stephencrowley]

I don't have any futures feeds yet, and I still need to port this over to my realtime platform from matlab..

Quote from bgp:

i would like to see monday forcast of e-mini.

thanks,
bgp

 

[stephencrowley]

I don't follow. A non-forecasting type of strategy would be basket spread trading where you dont bet on the individual price movement of a single holding in the basket.

Quote from OddTrader:

So you call that a forecasting. Then show us what's not a forecasting.

 

[stephencrowley]

Where did you get this idea?

"Gaussian Self-Affinity and Fractals: Globality, the Earth, 1/F Noise, and R/S"

Fractal/chaotic processes are clearly predictible in the short term if you have enough data, find the proper embedding and then find the appropriate lower-dimensional phase space representation.

See 'Taken's theorem'.

Price series can be realized as the output of a high-dimensional but semi-deterministic process, reconstructing this process is an ill-posed inverse problem where you have to reverse engineer the actual/hidden process generating this data.

Quote from jimrockford:

Mandelbrot has written, for decades, that price changes are unpredictable. You are contradicting Mandelbrot, by claiming you can predict prices. Yet you say "Mandelbrot was right." Could it be that you misunderstood Mandelbrot? If you did correctly understand Mandelbrot, would you please cite the Mandelbrot publication which confirms what you are saying?

 

[kut2k2]

Quote from stephencrowley:

Yah of course its useful.. when you get down to the highest frequency of raw ticks, it's not that useful because you are then trying to forecast the mostly-unpredicble actions of a few people/bots/whatever. The aggregate behavior us much easier to see.

The weatherman doesn't tell you the exact forecast down to sub-degree accuracy a week in advance does he? No.. he gives you a range and some confidence interval.. same thing here, chaos theory and all.

Stephen, top-posting is bad form.

My EOD indicator says GOOG was a buy on Tuesday the 15th; confirmation uptick on Friday. What does your forecaster say (GOOG, EOD basis)?

Also, (unanswered question) why do you say first differencing loses information? Thanks.