Difficulty Projections; Data on Bitcoin

[cunicula]

Hi all,

I would like to contribute by projecting difficulty increases using a vector autoregression. This is just a technique which attempts to use interrelated past movements in time series such as difficulty, exchange rates, trade volumes and prices of hardware to predict future movements in these series. In math terms:

Diff[t+n] = fn(Diff[t-1,...,t-30],ExRate[t-1,...,t-30],TradeVolume[t-1,...,t-30], whatever other variables are available)
....
Diff[t] = f1(Diff[t-1,...,t-30],ExRate[t-1,...,t-30],TradeVolume[t-1,...,t-30], whatever other variables are available)

Where f1 is the difficulty projection for today, and fn is the projection n days in the future.

Are the following daily time-series for bitcoin available in excel or .csv format? (perhaps each day for the last six months?)

1) Mt. Gox Exchange Rate
2) Mt. Gox Trade Volumes
3) Difficulty
Less Essential, but useful:
4) Hashing Power per $ of hardware
5) Hashing Power per Joule
6) Average Bid-Ask Spread on Mt. Gox

If someone can send a link to the data, I will run the regressions and post the difficulty projections here.
If desired, someone who can code (not me) could automate these regressions so we can get up real-time projections via
a website.

[1715159925]

1715159925

[1715159925]

1715159925

[Vandroiy]

This method of projection might face a problematic truth:

A trend is a trend until it stops.

Just saying, external factors dominate Bitcoin at the moment. An analysis that will return some exponential uptrend... you don't need a load of numerics to retrieve that.

[cunicula]

Of Course, there would be completely unanticipated changes. However, the perfect is the enemy of the good.
The method would yield a best prediction of future trends given past ones-nothing more.

I would also aggregate other information into the formula like data on searches from google trends.