Haven't checked the code but my assumption is that crafting is conditional double roll of a dice. First roll includes success rate and if that returns False second roll determines 'level' of failure.
I highly doubt first roll isn't correct. But as with any atomised* RNG, to really reverse results to 0.85 (85%) you would need to roll the dice infinite number of times. Would be just as possible to get small number (80 is small) of attempts and get something higher than actual success rate.
Of course, whether 85% SR at target level is fair, is another topic.
But I would definitely argue that having 5%+ failures resulting in >40 levels lower result @85% SR isn't.
---------------------
* RNG series where each 'roll' is completely independent of any and all previous rolls.
Typical 'easy-way-out' of game design. Properly one would design sliding RNG where chance rate starts at advertised and then gets adjusted slightly after each roll to nudge average towards advertised result.
Trivial example:
Premise is 0.8 (aka 80%)
So if one gets into slump, rolling below, the longer this goes on the better effective chance one gets. And vice versa with rolls above.
In the long run, it settles into cA but the benefit is that player sees far less 'bad luck black holes' and thus assumes fairness.
I highly doubt first roll isn't correct. But as with any atomised* RNG, to really reverse results to 0.85 (85%) you would need to roll the dice infinite number of times. Would be just as possible to get small number (80 is small) of attempts and get something higher than actual success rate.
Of course, whether 85% SR at target level is fair, is another topic.
But I would definitely argue that having 5%+ failures resulting in >40 levels lower result @85% SR isn't.
---------------------
* RNG series where each 'roll' is completely independent of any and all previous rolls.
Typical 'easy-way-out' of game design. Properly one would design sliding RNG where chance rate starts at advertised and then gets adjusted slightly after each roll to nudge average towards advertised result.
Trivial example:
Premise is 0.8 (aka 80%)
# pseudo code
k = 0.02 # k << 1, w/o it would oscillate too wildly
cA = 0.8 # advertised chance
cE = cA # effective chance
while < attempts
y = RNG(cE) # some function that returns min < cE < max
cE += (cA - y) * k
So if one gets into slump, rolling below, the longer this goes on the better effective chance one gets. And vice versa with rolls above.
In the long run, it settles into cA but the benefit is that player sees far less 'bad luck black holes' and thus assumes fairness.