Hydecka, you can use my links with Wolfram Alpha to plug in your numbers. If FD is truly 5%, then the chance of getting 3 out of 9 fails is slightly less than 1%. This sounds random, but since there are thousands of monks on the server using it constantly, it's actually almost guaranteed to happen.
Speaking of which, I promised to update my numbers there. I have 410,000 new log lines since 9/28:
Code:
zonko@vanderboom$ tail -n 410000 Logs/eqlog_Raev_project1999.txt | grep "Raev has fallen to the ground" | wc -l
59
zonko@vanderboom:$ tail -n 410000 Logs/eqlog_Raev_project1999.txt | grep "Talking to yourself again" | wc -l
990
59/990 = 6.0% (not 4.8%)
Chance of this happening according to Wolfram Alpha: 5.5%. So I could ALMOST write a scientific paper about how FD is broken and get it into a journal (usually they want 95% confidence). Of course, this is also why so much crap is published in medical journals: you do 20 experiments and one works, so you can publish that one
[You must be logged in to view images. Log in or Register.]
Note to Hydecka: the reason my 5.5% is so much more interesting than your 1%, nominally a smaller number, is that my sample size is 100x larger. But still, if we suppose FD is implemented correctly and there are 100 high level monks, then 5 of them will experience my level of success or less. Or with 25 2 week periods per year, each monk should get 1 period with this level of failures or more.
I have a healthy amount of respect for the streakiness of random number generators, but its hard for me to understand how that could be happening here. As I posted on page 1, even if the RNG was "return counter++ % 100" I would still expect it to look pretty random here because there would be so many calls between my FD attempts.
verdict: if I had to guess, I would say maybe the chance is more like 6-7% instead of 4.8. But I am not sure, and Haynar has a lot of other things to do.