Metsfanmax wrote:degaston wrote:Metsfanmax wrote:We were told that the list is replaced every hour. So one possible failure mode is if, for some reason, the list is no longer updating and got stuck on a list that was particularly non-uniform. But if the lists are still updating once per hour, then your explanation would require the sum total of those lists to be non-uniform.
The results I've seen are exactly what I would expect if the list was not being replaced. A small bias in a list 50,000 numbers would be reinforced every time the list was re-used. It would take just 139 fewer 1's, 66 extra 2's, etc. to achieve this result. Well within what you could expect from a random list of 50,000.
As you posted in the dice thread back in 2010, "we'll believe that there's a problem if you show us the evidence. ...demonstrate that there's some sort of bias in the dice. If you do that, we're not going to ignore the evidence". (http://www.conquerclub.com/forum/viewtopic.php?f=634&t=120373&start=270#p2748566) I realize that you were talking to people who felt that the dice were biased against them personally, but I think I've provided plenty of evidence that the dice have a small, but consistent bias against rolling 1's.
You can trust that I have taken notice, and I hope it's evident that I'm not ignoring your results. But this result is still not enough to prove that the dice are non-uniform when considered sitewide, since there could be some hidden selection bias in your sample (which is small compared to the site as a whole). More importantly, since this effect is relatively small, if you want this to be a site priority then you need to collect the data suggesting that the dice are not fair.
Did you read my post above? I pulled 50 random players across many ranks and found clear evidence of dice bias. 48 out of 50 players had more 6's than 1's. Only 1 player had a dice average below 3.5 but many had an average above. How many players would I need to check to provide you evidence? My evidence is on top of Degaston's.