Topic: Predicting future Bitcoin Price - page 3. (Read 10662 times)
Bro, do you even imgur.com? I can't see your pictures.
Care to backtest it? (That is doing the predictions at a already past point of time so you can see how useful the model is.)
How about this method:
1)Get the gox historical daily prices
2)Calculate the day-to-day percent changes, ie (Price[n+1] - Price[n])/Price[n]
3)Create a markov chain starting at the most recent price
4)Sample a price change from the distribution of differences in the historical data (see topright panel below)
5)Update Price by one day in future
6)Goto 4
Example outputs:
Overall Posterior Distributions:
Interpretation:
30 days from today there 95% probability the price will be between 62.5 and 320.7 usd, 67% probability it is between 105.7 and 219.6 USD, and the expected value is either 181 or 169 usd depending on whether you prefer the mean or median as your estimator.
1)Get the gox historical daily prices
2)Calculate the day-to-day percent changes, ie (Price[n+1] - Price[n])/Price[n]
3)Create a markov chain starting at the most recent price
4)Sample a price change from the distribution of differences in the historical data (see topright panel below)
5)Update Price by one day in future
6)Goto 4
Example outputs:
Overall Posterior Distributions:
Interpretation:
30 days from today there 95% probability the price will be between 62.5 and 320.7 usd, 67% probability it is between 105.7 and 219.6 USD, and the expected value is either 181 or 169 usd depending on whether you prefer the mean or median as your estimator.
Jump to: