![]() Something noteworthy is that even the most harmless of missions result in a death sometimes. For other seeds they could easily be more forgiving. In my case, the riskier mission are slightly more risky than stated, but well within bounds. The biggest deviations are for small risks with a low number of deaths, lower numbers have a higher tolerance before being significantly off, so the -30.82% are nothing odd. Slight deviations are to be expected since it is a RNG after all. The calculated number of deaths based on the sample values is basically as expected. I also counted the number of values where a mission would result in a death compared to the given chance. I had no case in which a calculated death didn't occur or in which a merc died that should have survived based on the odds. I tested a few high risk missions based on the given values to see if the stated chances are accurate, I also killed a few mercs with low risk missions when the values would allow it. This doesn't translate linearly to bigger sample sizes, but I estimate that the +8% deviation in one interval is nothing special for a uniform distribution. The chance for at least one interval being significantly off is almost twice as high as the chance for this not happening, therefore some bigger deviations are normal. Since there are 20 intervals, the chance for all intervals to lie inside of that region is the chance for a single interval raised to the power of 20, resulting in only a 36.92% chance. In this case, there is a 95.14% chance for the interval to be hit between 2 and 10 times (inclusive), with 10 hits being +100% from the ideal size. The lower and upper 2.5% form a significant deviation. The probability distribution forms a bell curve, a normal result lies within the 95% center region. The actual chance for it to be hit exactly 5 times is only 18%, still the highest probability for any exact number of hits. But with 100 samples, that doesn't mean that it will be hit exactly 5 times. Since my spreadsheet calculator can't handle the ridiculous ranges of numbers that occur when caclulating odds for 10000 samples, I have to use 100 samples as an example. Unfortunately in my case, the two upper intervals are hit more than average which would lead to more deaths, but is +8.0% really significantly off? If the percentage value is bigger than the survival chance (100% minus risk of death for the mission), the merc dies. So the actual number of hits per interval ranges from 468 to 540, that is -6.4% to +8.0% deviation relative to the ideal number. The number of hits in my distribution was as following, each lower bound is inclusive, each upper bound except for 100% is exclusive. The actual number of hits should be somewhere around that ideal number. In a random distribution, exactly 500 hits in all intervals are possible, but highly unlikely. In a perfect distribution, each interval would be hit 500 times. I divided the 10000 samples ranging from 0% to 100% into 20 intervals of 5% each. I generated 2000 odds for each of my 5 mercenaries, so I got 10000 values in total. ![]() But it was perfect for calculating lots of odds for analysis. ![]() I do not recomment using this calculator for actual gameplay since it lets you peek into the cards, which is basically cheating. Since it would take too long to get a big amount of success/fail numbers by just playing, I looked up the mercenary calculator that calculates your survival odds based on your current savegame. Short results are at the end for those that are not interested in the math. My goal was to find out if that is the case. While I don't believe it is rigged on purpose, it would certainly be possible for the RNG having a bug and distributing random numbers wrong, leading to wrong chances. Lots of "risk of death is rigged/bugged/wrong" threads since the introduction of mercenaries compelled me to analyze the numbers with a decent sample size and without any bias.
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |