This assumption is wrong. The median corresponds exactly to at least one witness-provided feed value. So for 25 witnesses the raw probability that the median is yours is 4%.
even the probility is 4%, when it become 4%^10, it's still a task impossible.
and you have give your answer for a wrong question.
the pattern what crazybit did is that he just follow the median price exactely instead of given another number which became a new median price.
let me show you what's the different, when there are 3 numbers, (0.271,0.272,0.273), 0.272 is the median number from others, 0.273 is your current number
you can change your number to any between 0.271 and 0.272 which could become a median number.
but for crazybit, he just change to 0.272 instead of any others between 0.271 and 0.272, so the psossibility is much less than your answer.
This raw probability is in reality increased significantly by at least two more aspects:
* When a witness publishes a new feed value, his feed is the most recent and therefore more likely to be close to x than anyone else's.
you should see other's line chat from my previous post.
https://bitsharestalk.org/index.php?topic=27687.msg326833#msg326833it's unfortunately that the data seems not available after 2018.12.26 13:00, or you will see it more clearly.
* Some witnesses publish a new feed only if that would change the median. (This is a simple optimization which is perfectly valid and acceptable if the feed is checked often.)
IMO if you publish a price only if it changes the median, and your price is likely to be close to the median, the conclusion is that your feed value will be chosen as the median quite often.
exactely the opposite, crazybit never change the median, he always publish a new feed price which already in the median price, this means all his data is totally useless.