Dukasaur wrote:Degaston, your own data shows the 50K file has been replaced at least a year ago, possibly more. It was the 3.51 averages that led you to uncover the problem with the 50K file, and 3.51 averages are becoming steadily less common, which implies that the problem was solved. Obviously the distortion caused by six years of the 50K file is not going to disappear from people's dice stats overnight, and the longer they've been members the slower the correction will be, but as 3.51 get less and less common there is obviously a correction taking place.
Two years ago, when you posted your research, more than 99% of members had a 3.51 for all three dice stats. Today I randomly chose a starting point on page 3 of the scoreboard and checked 25 members in sequence. I found six with all 3.51s, twelve with a mixture of 3.51s and 3.50s, and seven with all 3.50s. You can do a larger sample if you'd like, but I think that it's fairly indicative that half or less than half of all active members have 3.51s now.
As a second check, I resorted the list by Games Completed. Then I moved down the list and started at 200 completed games (so only players with a reasonably large data pool would be included) and started upward, looking only at the dice stats of those who have joined in the last 18 months. Looking at the first 25 members who fit that description, I found eleven that had all 3.50s and fourteen that had something else. Among those "something else" members, however, there were more 3.49s than 3.51s. There were also a few 3.47s and a 3.52. Overall, an average somewhere below 3.50. Clearly, members who have joined in the last 18 months do not have a tendency toward 3.51s like the older members do.
I really hope you take another look at the data. I'm following your own reasoning. The overwhelming prevalence of 3.51s in the past proved the problem with the 50K file, and the fact that 3.51 are steadily declining in frequency proves that the problem has been fixed.
I'm aware that they replaced the data file, and believe that the bias was (mostly) removed, but that doesn't make the dice rolls random.
Randomness is about every possible event having a specific mathematical probability of occurring, and when you're reusing a 50k file, certain events are going to have either a higher or lower probability than they should.
For example, if the rolls were truly random, then someone auto-assaulting from a territory with 13 to a territory with 1 should expect to fail once every 195,505 attempts. With only 50,000 possible starting points in the file, if a sequence that loses 12 in a row exists, then players will hit it about four times more often than they should (assuming that it only occurs once). If it does not exist anywhere in the file, then players will never see this happen when sometimes they should.
Rolling five of the same number should occur once every 1,296 rolls for a 3v2 attack. So in a 50,000 roll file, there should be 6.43 sequences of 5 in a row for each digit. Has someone checked to make sure that they're there? Even if they are, the best you can do is to have the occurrences be either 8.8% high, or 6.7% low.
Rolling the same number seven times in a row should happen only once every 46,656 attempts. Does this occur at all in the data file? If it does, which number is it? Wouldn't it be odd if every time you saw seven in a row, it was always 4's?
I don't know what's in this data file, but I can assure you that there are some rare sequences that are happening too often, and others that are not happening at all, but should. So when the next person starts a "dice suck" thread, how do you know that they don't have a legitimate complaint?
[/quote]
