I get that.
I cannot calculate the likelihood of this because I don't know the bet size distribution. It would take a large number of high bets (0.5-1% of the betting pool) for this to be at all probable.
It seems weird given the fact that there have been so many bets that you would not be closer to a normal distribution. However, maybe there is a large number of big bets and the rest of the bets are very tiny.
The luck is due to 2 people running the extreme ends of low probability wins, and winning early.
Estimations, but I believe 2 wins of a 1 in 1,000,000 chance happened at 750,000 rolls cumulatively? That is where the luck came from.
One thing that is on the table (doog has no shortage of feature requests) is to make luck scores based on time or bet # bands. Those spam bets ended a few days ago and the luck has dropped 40-50% since then, and didn't yield much loss for the site. So the luck for the past 2 days is vastly different than the overall luck, etc.