There's a lot of confusion about this and people are really going crazy after the COLDCARD issue, I just wanted to share with you the discussion I had with some people on Telegram
On Telegram, as I was saying earlier, there's a lot of excitement in the serious Bitcoin groups and all they do is talk about dice and bullshit like that... always related to seed generation.
There are people who are sick and overdo it and recommend the AAA grade dice from the casino because they are perfect and balanced.
but could something like this ever be true?If they ask you, the answer is no.
I had it calculated and the deviation would be 1% in the case of cheap commercial dice... that is to say a die that has a percentage on one face not of 16,..% but of 17/19% maximum... (otherwise it is rigged)
Another user went to look through the statistical mathematics literature
I found a Harvard University study published in 1971, when Bitcoin didn't even exist. The researchers analyzed 219 commercial dice from four different brands, rolling each die 20,000 times, for a total of 4,380,000 rolls.
The study found small statistical biases in some commercial dice, but if we translate those results into the entropy of a 24-word BIP39 seed, the loss is on the order of a few tenths of a bit, at most about 1 bit under the most pessimistic assumptions.
In practice, a 256-bit seed would still be a seed with approximately 255-256 bits of effective entropy.
This difference is theoretically measurable, but completely irrelevant from a security perspective: the search space remains astronomical and impossible to explore with any existing technology.
so commercial dice are more than fine, this is the conclusion about dice
