Talk:Fall 2019 Event/@comment-31894513-20200110022541/@comment-26809641-20200111173419

I guess I'm not really understanding what you're claiming, because "close to the expected value, but with some variance" is also precisely what we'd expect to see if it were truly random. To use your coin-tossing example, the odds of getting 90 or more of the same result is something like 3*10^-17, or so small that if everyone in the world were to toss a hundred coins a million times, we still wouldn't expect to ever see it.

Further, I'm not sure why the developers would go through all that hassle. To guarantee a result, they would either need to be working off of a pre-made list (which would have to be absolutely massive to make it seem random), or else track the historic drop data and modify the probability accordingly. And even then, if it's shared among all players, you would still expect to see outliers where some players got a lot of a drop and some players got very few/none.

Why not just go the simpler route of choosing a random integer and then choosing a girl based on the number? Say, pick a number between 1 and 10,000, and if it's 0-249, go with a girl with a 2.5% chance. If it's 250-499, go with a different girl with 2.5% chance. If it's 500-599, a girl with 1% chance, etc. This is simpler, doesn't require tracking historic data, and also prevents a race condition of multiple players looking for a number at the same time, as each person's random number is independent of the others. Beyond that, the law of averages will handle smoothing things and preventing extreme outliers.