Care to provide an explanation of how a 2,000% increase is supposed to occur within the next 6 months?
Good question. My analysis has its obvious limitations. It extrapolates, and it does so based on only 6 previous crashes. If you are willing to assume that the past is a guide to the future, and that the next cycle will have some similarity to the previous 6, then we can continue talking.
I very much appreciate all constructive criticism. That's why I posted this here. If you have ideas for improving the analysis, let me know!
The prediction of the next peak price is based on the peak to peak ratio. In the past, this has ranged from 3 all the way to 28. The geometric average of the past ratios is 6.6. Since this current peak is 1,130, the next peak could be around 7,500. However, the ratio has a pretty wide range -- 7,500 is not definite. I think the peak can reasonably be anywhere between 4,400 and 13,000.
The durations I am less confident about than the prices. And of the 2 prices that I mentioned, I am a lot more confident not in the peak price but in the trough following the next peak, which I see as being around 2,300, based on another ratio which has a lot tighter range.
In the past, the peak to peak duration has ranged from 99 days all the way to 664 days. I do not know the best way to construct a predictive distribution based on that. Do you have any ideas?
In the original post, I assumed a log-normal distribution. Really, I should have used the truncated log-normal, not the regular log-normal. Unless I am making an error in calculation, conditional on the next peak occuring more than 167 days from the previous peak (which have already elapsed), the 25/50/75 percentiles for the peak to peak duration are 226, 305, 413. Thus means the inter-quartile for the next peak is 2014-07-13 to 2015-01-16, with the median being on 2014-10-01. [This calculation, done on the back of the proverbial envelope, is not exact. Suggestions for improving it are welcome.]
Now, to your question -- is this pace realistic? In another thread, I describe
a very basic long-term model for price. It's basically just an improvement of the log-linear trend that people like to draw. Here is the prediction from that model:
n.fut date p_5 p_50 p_95
1 0 2014-05-14 NA 443 NA
2 1 2014-05-15 400 446 496
3 7 2014-05-21 349 463 615
4 30 2014-06-13 298 539 974
5 60 2014-07-13 281 655 1,530
6 91 2014-08-13 280 802 2,300
7 140 2014-10-01 293 1,100 4,170
8 247 2015-01-16 358 2,220 13,800
9 365 2015-05-14 483 4,800 47,700
10 731 2016-05-14 1,470 52,400 1,860,000
According to the basic long-term model (BLTM), by 2014-07-13, we will not hit the peak predicted by the peak/trough model (PTM). According to BLTM, by 2014-10-01, it is possible, though pretty unlikely, that we could be at above 4,000. And by 2015-01-16, the peak predicted by PTM is easily within the price range predicted by BLTM.
Based on this, I revise my prediction for the timing of the peak to be closer to 2015-01-16 rather than 2014-10-01.
