Quote from: rpietila on April 26, 2014, 10:05:14 AM

Quote from: boumalo on April 26, 2014, 10:01:39 AM

Quote from: rpietila on April 26, 2014, 07:20:01 AM

Quote from: jdough on April 25, 2014, 10:44:15 PM

Quote from: rpietila on April 22, 2014, 12:24:00 PM

The new information is in the distribution parameters that were updated based on the Mt.Gox data.

The Mt.Gox leak was the data after accounts were frozen right? Considering there were many hints of Mt.Gox's impending failure well in advance (goxing was a verb a long time ago), I expect a large percentage of people withdrew their funds either completely or mostly (my case). Therefore, I am not sure how useful raw Mt.Gox numbers would be for finding the overall distribution.

20 pages back in this thread, when the methodology was created, we used the distribution data from a certain company called Silvervault as a basis for some assumptions of this model. Even impaired Mt.Gox data is an improvement. An example problem where we have better answer now, is:

1.0 million coins belong to people with 100-1000 coins
There are 45,000 people with 10-100 coins.

How many people own 100-1000 coins?

1million coins belong to people with 100-1000coins :

if they each have 100coins, they are 10,000
if they each have 1000coins, they are 1,000 so people having 100-1000coins are between 1000 and 10,000

Yet you wrote : 13k BTC100-1k that would mean each person has less than 100coins

Without fudging the numbers, you would guess too easily Cheesy

The problem above can actually be answered statistically, and much better with Mt.Gox data than without it.

Hmm

Do you mean you are fudging the numbers?

I just applied basic logic; would you elaborate? Maybe all numbers are approximative and you prefer to keep the numbers you find instead of changing them just to make them work together better