What are pseudo-random numbers? What’s wrong with them?
In: 3
Lets imagine you have a situation where you have different kinds of variables for a model of a complex system. Something you might not be able to just mathematically solve. Of course you could iterate every single possible variation, but that can be insanely massive task. Or you could try to solve it, but it doesn’t necessarily give you an actual useful answer like tests would. Example you might be testing an actual phyiscal piece of hardware or other equipment.
So what do you do? Well you have random numbers, you feed in to one variable totally random numbers and see how the system behaves.
Example if you want to filter noise out of a signal, a image or whatever, you can not test ALL the noise that there can be in the world. So you generate random noise based on random numbers.
Issue with pseudo-random is that it isn’t random, it appears to us random but there are clear patterns and limitations to it. Let imagine something like *chaos* as a game mechanic: Roll a D10 dice and if you get more than 2 you perform an action. Sounds simple enough right? Well just like anyone who has ever played dice games knows some sessions or even dices are cursed. In a truly random system there is no reason why you should ever roll anything but 1 and 2. You can bang on about how unlikely it is to roll n- times and never get more than 2, but it doesn’t matter because the rolls are independent from eachother, the result of last roll doesn’t affect the next. So to avoid this, in video games you generally use pseudo-random, meaning that it is random, but the last result affects the next or some other factor affects the result. Like if an ability has a 10% chance to do something, then it is coded so that every 10 uses it does it at least once. Anyone who played Chaos Knight or Ogre Magi in Dota before they were made pseudo-random, knows the pain that is to play a 1hr game and never has “the thing” happened – it made for a shit gaming experience, absolutely unreliable hero to play as, and totally unusable in a tactics game. So they changed it to pseudo-random so that those abilities “did the thing” regularly, but you couldn’t be sure when. Generally how this works is so that every time you use it and the thing doesn’t happen, it add a modifier to the next roll, meaning that as the failures accumulate at some point is has to happen. Like lets say you have to roll 100 sided dice and get 100, meaning 1% chance. Every time you roll less than 100, you get +1 to your next roll. Meaning that after 99 rolls, you will get 100. Or another method is to use the last roll and add it on top of the new roll, or use it as a multiplier. This is how pseudo random-works. You take something and do something to it to get a result that is different enough from the last – as long as someone doesn’t know what you did it is random to them; but given enough results they can map out what you are doing.
Random numbers are useful to testing things. Whether it is to test how it handles randomness, how randomness contributes to it, or just how meaningful something is to the overall result. You can’t predict the random nature of random numbers, therefor your results wont be biased because of HOW you generated those numbers. If you test a pseudo-random system long enough, you will start to see patterns in it, because it has one.
Latest Answers