Understanding Random Number Generation and Its Impact on Game Probability Testing

Players often hear mystical claims about luck and probability in games, but QA engineers need factual verification to ensure probability‑related quality.

Random number quality depends on three properties: randomness, unpredictability, and irreproducibility. True random numbers satisfy all three by measuring physical phenomena such as coin flips.

Programmatic pseudo‑random numbers usually only satisfy randomness because they are produced by deterministic algorithms that repeat after a long sequence.

x = seed
R_1 = f(x)
R_2 = f(R_1)
R_3 = f(R_2)
...
R_n = f(R_{n-1})

In Lua, setting the same seed reproduces the same sequence:

math.randomseed(1)
math.random()
>>>0.56356811523438
math.randomseed(1)
math.random()
>>>0.56356811523438

Common algorithms include the linear congruential generator and the Mersenne Twister.

Many “mystical” observations stem from bugs, such as repeatedly setting the random seed using the current OS time (only precise to seconds), causing identical loading tips for players who load within the same second.

function LoadingCtrl:onCreate(view)
    uiCtrl.UICtrl.onCreate(self, view)
    math.randomseed(toString(timeUtils.getOsTime()):reverse():sub(1,6))
    self.startTime = Time.realtimeSinceStartup
    self:initTips()
    self:initLoadingPic()
end

General seed guidelines: set the seed only once (e.g., at server start), use high‑entropy sources (prefer millisecond‑precision timestamps or better), and ensure thread‑safe random generation in concurrent environments; QA should simulate probabilities under realistic conditions.

Understanding these fundamentals helps debunk false luck myths and prevents critical probability bugs in both client and server code.