There remain two ranges in particular that are intriguing, at least some. The 90-95% one, as ~10% below the expected (though I was 9% above the expected in my 81% hit chance test -- came out at 90% after 50 rolls). And the 50-55% one. You personally also rarely come out above the expected in your bigger samples, whilst for me it was a mixed experience of being above and below. Excluding the damage roll and AI vs player bias tests I did:
The 90-95 range was even worse for a time with around 14% behind the expected hit rate. Even though my numbers are lacking a bit behind I think it is fine over all. The game won't have "true" random generation, as this is quite the challenge in computer science. I assume there are no correction mechanisms implemented to force expected hit ratios. If so, I would say a sufficient number of rolls for a certain percentage range would need to be around 1000 (just an arbitrarily chosen high number). Bad/Good streaks are comon with rolling dice (simulated and irl). Thus, those streaks can easily ruin expected ratios. It is also important to note, that a difference of 2.5% in my analysis should be considered okay, because so far I do not account for the spread of expected values within a certain percentage range.
Originally Posted by Sven_
NOt that'd convince anybody who isn't convinced yet -- I actually kind of find this fun.
Same here^^
Also, before recording my rolls I was convinced the rolls are incorrect (especially one encounter where all my 70% failed and the enemies below 50% [according to the log] hit me). But now I believe it is mostly alright. There might still be issues in some places (for instance, I don't know if chances with the lucky trait are displayed correctly) but overall I am satisfied with the RNG so far.
