I was on a voyage for scientific discovery just now, and failed the 71%-68% checks 20 TIMES IN A ROW (not to mention that I succeeded a 10% check in ‘A state of some confusion’ 9 times in a row too)
1 Like
Consider it the Correspondence reminding you that even probabilities may burn.
Seriously, although it is extremely rare, it’s not outright impossible. Maybe there are other shadow-mechanics in action.
Failbetter stated at some point, that they are using an industry standard random number generation. On the other hand, there was a time, when other systems than the random number generation itself messed with the experienced randomness. I believe it had to do with the opportunity deck and cards that required certain qualities.
That being said, your experience is anecdotal evidence at best. Streaks are part of randomness, so one might even claim that your experience is actually evidence of the RNG being completely fine.
1 Like
On the other hand, using the correct tool does not mean you use it technically correctly [1]
and even using it technically correctly does not mean the end experience is a fair randomness. [2]
[1] off the side of my head …
You might have a thread-unsafe code used in multiple threads without synchronization - or you have synchronization but it creates a performance bottleneck.
Or you mix output of multiple rng instances in a chaotic way, which destroys statistical properties.
Or you seed it too often (explicitly or implicitly).
Or you throw away the most random bits and use less random bits instead.
[2] I have heard about a case when a game was shuffling an array of results, the code looked fine and fair at first sight … but it produced certain orderings way more often than it should have, and others (that it should have produced sometimes) never.
You can also introduce subtle unfairnesses by making the effect (and number) of random rolls depend on a previous random roll.
Or when you have a series of equally fair rolls, but one of them has larger effect on the total result (maybe even simply because it’s the first (or last) one) - and that one also happens to be influenced the most by a factor biased towards one of the results.
Or many other subtle things that you won’t notice, and then you will reject all observations of unfairness with “but we use an industry-standard RNG!”.
This is all just to illustrate that randomness is not something as simple as pressing a button on a black box.
Well, nobody rejected any observations. I actually chipped in on the stated observation (though I am no expert in implementation of random number generation, nor do I have deeper insight in the intricacies of the engine). But it’s not likely that we will get a more nuanced answer from Failbetter. Getting into technical discussions over the implementation with the playerbase whenever someone experiences streakiness is not an efficient allocation of ressources. The case I alluded to was discovered by a player who made an observation and backed it up with statistical data and Failbetter reacted to it. I am sure any further backed up claims about problems with the output randomness will be addressed in the same way.
1 Like
It’s as random as can be, and we haven’t done anything recently that would change it, which I think is pertinent.
2 Likes