On Nov 14, 2014, at 10:54 AM, Viktor Dukhovni wrote:

> On Fri, Nov 14, 2014 at 10:26:29AM -0600, Edgar Pettijohn wrote:

>

>>> On January 15th each year Wietse sets a counter for the following

>>> year's release to zero. Each day after that he rolls a 6 sided

>>> dice, and adds the value to the running total. When the total

>>> reaches 1278, a new release is cut. :-)

>>

>> So around August?

>

> [ Off topic alert, move along... ]

>

> Your arithmetic is different than mine.

>

> $ echo "2k 1278 3.5 / p" | dc

> 365.14

>

> Your task is to compute the variance, it is easy to compute the

> variance of total after 365 days. I have not thought about how to

> correctly compute the variance of the number of days needed to

> reach a target total. A naive order of magnitude guess is to take

> the variance of the expected total after 365 days and divice by

> the mean increment per day. That gives a guestimated standard

> deviation of ~sqrt(365 * 35/12)/3.5 or 9.5 days. Replace the dice

> with a coin toss, how does that change the standard deviation? :-)

>

> --

> Viktor.

>

It would be a minimum of 213 days which is around august or a maximum of 3 1/2 years which would be around august. Plus there are probably unwritten rules. For all we know he re-rolls all 3's.