would love to see this on a log scale, starting at zero.
What do you mean by this?
Another possible assumption is that the top 100 days is somehow different than the top 101,102, etc days. Why stop at 100? If you plot all the days then the log curve definitely does not fit. This is originally why I asked what assumptions were usually made with the king effect model. The outliers are only outliers in that they don't fit the proposed model (price=log(rank)). Its not clear from that wikipedia page why we should expect such a relationship between rank and price.
the wikipedia page has a good graph of population-ranking data behaving in a similar way. the idea is that the few very large data points represent statistical aberrations, as evidenced by that the rest of the price data correlate (with an extremely good r- value) on that scale.
from this, one assumes: The price events at the top of the list represent behavior that is abnormal in relation to the rest of the [top 100]* trading days.
*disclaimer: this data only regards the top 100 trading days and makes no prediction about the behavior of any other data set
from this assumption, we can reason from history that these outlier price events all relate to an event generally regarded as a bubble (i.e. high prices, low stability). the full number of predictions extendable from this assumption was made clear in the OP:
- Price remains below ~$25, with stability
- Price rises above ~$25, then quickly collapses below ~$25, with some stability
- Price rises above ~$25, then continues rising to around $40 before returning to the $25 to $30 level, with little stability
- Price rises above ~$25, then continues rising to far above $25 before collapsing but remaining above $25, with almost no stability
The king effect means that it is unlikely we will see both new highs and stability. One has to go.
Ok, but why does the price=log(rank) (of top 100 prices) imply stability? What is the mechanism behind this? Stability is something that is assessed over time, a factor which the above graph ignores. I'm not trying to say its wrong, just that I don't know if I follow the assumptions that need to be made to draw inferences from it.
