What is Bitcoin Correlated With? [PART 2/2]
I will reveal here the data that is best correlated with the price of Bitcoin and can be used to forecast the price! If you have missed the previous article, read it here, so that you know what I am talking about:
ARIMAX ( 1 0 1 ) {Bitcointalk NewTopics}
If we use the Nr. of New Topics/Day at the Bitcointalk forum as our regressor, then we get this result. Pretty good, with only an error of 83.104, way better than all our other variables. And this should not be a surprise, all other variables could be heavily manipulated by bots & shills, but the number of topics is usually restricted, so it measures activity better. After all the activity of Bitcoin should determine the price in our theory.
ARIMAX ( 1 0 1 ) {All variables}
I have added all variables as regressors, and this is the result, literally 2x better than all others individually. So by using all variables, it models the price much better, even the inadequate ones like the Pageviews & Online users. It has an error of only 42.854.
ARIMAX ( 1 0 1 ) {New Blockchain.info Users}
The number of new Blockchain.info users literally trumps alone all other variables. It is probably the most accurate variable alone, to measure the activity of users. Well yes, it's registration based, less spam, less shills, less bots, so it is an accurate representation, and it is useful as a regressor. It has an error of only 41.882.
THE BEST MODEL
ARIMAX ( 1 0 1 ) {All variables - excluding Bitcointalk Posts}
If we take all the variables, but exclude the Posts on Bitcointalk, then we get this. Why the Posts? Probably because the signature spamming on that forum distorts the statistics. Therefore this is the best model I have found, but I haven't checked all combinations yet, so that may be for a future article. This model has an error of only 40.685.
The residuals are normally distributed: Chi-square(2) = 7312.285 with p-value 0.00000, they don't correlate with their lags, and fit well around the data, so this suggests that this is a potentially very accurate model, by the research done so far. I will explore this more in the future though.
Forecast into the Future
So now that we have a decent model, the only thing remains is to forecast something into the real future:
We forecast 300 days into the future, and this is what we get. Based on the best model, it suggests that the price in 300 days will be:
957.27$
95% chance between 730.02$ - 1184.52$
This is all for now, hope you enjoyed my research, I will explore this more deeply in the future, but for now this is what I have discovered.
Disclaimer: The information provided on this page might be incorrect. I am not responsible if you lose money using the information on this page! This is not an investment advice, just my opinion and analysis for educational purposes.
I agree with your price target. My Elliott wave analysis has a similar target price.