wolfpack0530 wrote:You assume waaay too much sir. First you assume that i care alot about winning and losing, and you keep using the word rigged , which implies certain connotations of cheating and deception. Both are quite inaccurate, and unwarranted.
I said nothing regarding cheating or deception. My point was quite straightforward - if you object to streaks, the only logical alternative is to "fix" the dice so that they don't happen.
I dont need any dice to be rigged so that 3v2 wins 37.17% of the time. I KNOW that they WILL win 37.17% of the time with a very large sample size approaching infinity. Quit talking semantics and wasting our time citing things we both already know
I dont doubt that the numbers themselves are perfectly random, or that random.org is unfair. I started this by just posting my personal observations. Perhaps random.org's numbers, and tests were performed on large sample sizes. I am assuming that the more rolls one takes, the more "even" the dice become. After 20,000,000 rolls you see pretty clearly that you have pretty much rolled each number 16.67% of the time, and so has the defender.
But what about the short term? a sample size of 30 rolls, or 15, or even 100. perhaps, random.org's or whatever manipulation CC uses, doesnt stack up so "fairly" over such small sample sizes as the previous manipulation did.
Perhaps, changing the dice was a huge mistake?
This is an illogical conclusion. We're using the same RNG that we used to (all numbers previously came, and still do come, from random.org). We've only changed the method we use to obtain
the numbers from that site. Therefore, unless random.org has changed its algorithm around the same time we changed our method of obtaining the dice from them, then nothing has changed, and you can't say the dice have changed at all.
How could you be so obtuse as to assume that?? I cant speak for others, but i expect an event with a 1% chance of occurring, to occur 1% (plus or minus the standard deviation) of the time.
Well, I don't want to be rude, but I assume things like that because of incorrect statements like this. Probability only meaningfully deals with the proportions of events occurring over a large sample size, as you point out in the next line - it has nothing to do with how often
an event occurs in time. If the only test for fairness is that the event occurs about 100 times in 10,000 rolls, then the test is still being upheld if all 100 times happen in the first 200 rolls of that sample. You're being disingenuous below when you say that this proves there's a problem with the dice - either you believe that probability is only meaningful over large sample sizes, or you don't. If you do, you can't point to a small sample (~1000 rolls) and use it to say the dice aren't random.
Of course this would be over a larger sample size. Over the course of 10,000 rolls, I expect to see between 80-120 "unlikely events" defined by me as events likely to occur less than 1% of the time. Now if I see 50 "unlikely events" in 700 rolls, then i start to think that is very odd. then i see another 50 unlikely events in the next 1300 rolls, my doubts start to become confirmed in my mind. I have just seen 100 rare occurrances in 2000 rolls, this is far exceeding the 1% of the time probablility i was expecting. I doubt that true randomness is occuring, and i highly doubt that i will see only 0-20 rare occurances in the next 8000 rolls.
Again, the only way that the dice could be non-random is if random.org has a flawed algorithm (which they don't). That being said, carl and I have also said that one way that streaks could occur is if the sequences of numbers we get are somehow skewed toward changing the probabilities of a given 3v2 roll, for example. If the numbers were purely random, then this would not be an issue, but I suppose it's conceivably within the realm of possibilities in the Universe that there's enough of a bias in the distribution such that these patterns could occur. This is why I've submitted an eTicket for some of the old dice lists to be released, and we'll do some analysis on those lists to see whether they substantially deviate from expected outcomes for large samples. I would hope that if we find that the results are suitably random, that you would admit that there really isn't a problem with the dice.