Chestno govoria, mne len’ perepisyvat’ po-russki to, chto ia uje napisal po-angliiski na http://forums.immigration.com/showthread.php?323624-DV-2012-was-a-SCAM-!!!/page7
Ne budu.
Let’s assume we have 3030 entries sumbitted throughout the whole entry period.
Let’s divide it to 10 intervals (not necessarily contigous) where we have 303 entries submitted during each time interval. So, equal number of entries per interval.
First we use 1/10 probability to select an interval randomly (we need first random number for that), and then we will select one winning entry from the interval (second random number for that). Simple enough. Probabilities to select the intervals are all equal to 1/10. Columns C, D and E show that the probability of each entry to be selected will be finally the same = 1/3030. The law about equal chances for entries is followed - look at column E.
Now we calculate expected number of wins (G12 = 1) and standard deviation (I12 = 0)
Now, let’s assume the length of all intervals is in column J and is not the same. The middleman would have a uniform distribution throughout the entry period, and columns K-P show that and expectation and standard deviation of the number of wins registered in the niddlman’s bucket.
The expectation is still the same (N12 = 1), but standard deviation is different (P12 = 0.2667)
So, if previously the middleman had a guaranteed 1 win with 0 standard deviation, now he would have still 1 win in average, but 0.2667 standard deviation.
So, if the middleman charged $100 for the win before and always got $1000, now he would still get an average of $1000, but there is certain probability he would get less and certain probability he would get more. For instance, with about 30% probability the win would be 730$ isntead of $1000, and about 30% probability it would be $1270.
If the cost to enter is $800, and the middleman expected to get $1000 in return and get $200 profit, he would be bankrupt with probability more than 30% (because he spent $800, but got only $730 in return)
The numbers are all samples, and in reality the numbers could be different, but the idea is clear.