Historical Bitcoin Price and Prediction

This article covers a source of historical bitcoin price data and a simple prediction based on that data.

One place to find historical bitcoin price data would be from <a href="http://www.quandl.com/BITCOIN-Bitcoin-Charts/MTGOXUSD-Bitcoin-Markets-mtgoxUSD">quandl. With an "L," as in "candle." It's a finance thing.

I grabbed two data sets and performed separate regressions using Gnumeric. One set of data ran from October 30, 2011 until October 30, 2013. Data points are daily weighted averages which helps to avoid cyclical effects from day trading. All data is based on the Mt Gox exchange. I ran a total of 8 regressions. Results are summarized below and the spreadsheet with nitty gritty details is <a href="https://mega.co.nz/#!qJYgRCBS!P1aoHG8fo2h5IhFbbof2_jWw_t0s5EkfDlYSIH9t40I">available here.

In contrast to what I say in the image, there are two possible best models. Regardless, as you can see I found a logarithmic model with exceptional predictive power. I used model #6 to forecast that the price of 1 bitcoin on January 1, 2014 will be between about $115 and $150. This indicates one of two facts and both seem quite possible. In the first case the model is suggesting that we are currently in a bitcoin price bubble. In the second case the reality of the fact that bitcoin is over $200 today suggests that the model is flawed. Thirdly, of course, it is possible that each effect is true to some intermediary degree.

Let's assume the model is garbage. Despite the statistical validity of the model, it is not altogether unreasonable to suggest that forecasting from that data would be extrapolation due to the fact that so very many factors have changed within the world of bitcoin between October 2011 and today.

It could be argued that things are still changing so rapidly that even a model using relatively recent data might be considered extrapolation, but I went ahead and ran regressions on data from June 1, 2013 until October 30, 2013. The results are available here and summarized below.

The regression which used more recent data provided what I think to be a more realistic projection. The minimum forecast was for around $175 and the maximum was for around $250. I am very comfortable forecasting that the January 1, 2014 bitcoin price will fall somewhere in that range and my intuition is that it will fall in the $200 - $225 with breakout potential outside of those two psychological barrier numbers.

I would issue a hold for pricing above $200, a sell at $238 and a buy at $182. The 2-offs may or may not be a big deal. The point is simply to get act before the psychological bitcoin price correction points arrive at $240 or $180. Remember, these are all Mt Gox prices. You will need to make some small adjustments to apply these numbers elsewhere.