Board Thread:Game Discussion/@comment-13368701-20130905202323/@comment-11490363-20130907053258

Andrewds1021 wrote: @4.71.20.114

Not neccissarily, I am just saying it isn't too far fetched.

@Ranoake

I'm not sure I follow what you are saying about the bias balancing out. I'll try to explain in more detail now and hope I don't confuse you more.

Assume there IS a bias, in that only the 'hard luck' players complain and not as many of the 'average' outcome players. so that in a group of 100 players, the 10 outliers post results because they are outraged at the unfairness and only 10 'average ones' because they are just average and don't care as much since they are relatively happy with their pulls. That is clearly a biased sample.

BUT, the 10 outliers are likely evenly split between people with alt streaks and people with bot streaks. This would be true ONLY if the odds were 50/50. For 50/50 there would be the same number of bot streaks as alts, so adding the numbers individually would not stray that far from 50/50 in that case. So the bias cancels out. Let's say 5 outliers report alt streaks of 4, 5, 6, 7, and 8. and 5 outliers report bot streaks of 5, 6, 6, 6, and 7. This cancels out and points to a 50/50 base rate.

If the odds were not 50/50 then there would be a bias overall towards one side of the other. But that is OK, because whatever bias there is would reflect the actual percentage. If the percentage were 90/10 for example, with 90 being for bots. 9 of the 10 bias samples would be for bots, and 1 would be for alts. Also the 'average' samples would not really exist because it is so skewed, so tallying the numbers would actually show a 9:1 ratio for total bots vs alts. Which would indicate the percentage is NOT 50/50.

So all we have to do is get as many people to report as possible, and even with a biased sample, it is likely that the biases cancel out and if there are roughly equal numbers of bots and alts TOTAL (forget streaks, that is too hard to track) then that indicates a likely 50/50 base rate. The more posts we have with results, the more confident we are in the total. 1000 total pulls would be decent, and I would expect it to be about 500 of each overall. And if the base rate is not 50/50 AND the results are biased or not, the totals would not be roughly equal, they would reflect whatever the percentage actually is. Standard stats. Take a bunch of events and tally them and over time you see what the percentage is. However It would NOT reveal if the cyberdex is checked AFTER the first pull since the pulss after the first are just a duplication of a previous event and that would be eenly split between bots and alts so the average would still be 50/50.

I am not sure yet if we should include the 100 MobaCoin bridge though, in theory it should be the same, but it is always best to remove possible unknowns.

Make sense?