Implement a true RNG as an OPTION

[Kevin]

How about adding an option for an truly random RNG? Could be a button you select in the options screen that actually allows die rolls to average out to an even probablilty of rolling each number 1 through 6.

Just played another game where all but 1 misfortune I landed on rolled a 1. If you don't know, low rolls are bad in this game. According to the statistics screen after this latest game, I rolled 8 "1's", 6 "2's", 2 "3's", 3 "4's", 3 "5's" and 1 "6". This is the second game in a row that had such very skewed rolls toward low numbers.

It's a great game, and I get that artifically low die rolls may be part of the "Gremlins" aspect of the game, but many posters hate the rigged RNG (actually NG, as there's not much randomness to it) and it detracts from an otherwise fun game. How about implementing a good, fair RNG as an option button to remove all die roll "modifications"?

[Samantha Prezombie]

The dice rolls are fair.

Randomness doesn't mean "roll 6 times get each number once" or "roll 12 times get each number twice". It means "roll 6000 times and you'll get each number approximately 1000 times."

Randomness tests require large pools. Don't just count 2 games. Collect the numbers from a dozen games, and don't just look at your numbers, collect the numbers of everyone. Then you'll see that the numbers come out at similar frequencies.

What you're experiencing is clumping, which simply happens when looking at sequences of random numbers, and the human brain is very good at assigning meaning or intent to what is simply random. Some games you'll seem to have terrible luck, others you'll have great luck, and most will be fairly normal.

So yeah, relax, take a break, and watch a youtube lesson on into to statistics before claiming RNGs are rigged and many posters are angry, when the first has been tested and disproved multiple times, and the second is simply false.

[Name Lips]

You'll see this in real life games too, when a player in a game of Monopoly seems to keep rolling doubles, or a D&D player seems to always roll crits. If you pay attention and track the rolls, you'll find that over hundreds of rolls, it averages out. In the short term, with just a dozen or so rolls, you'll see definite "clumping" or seemingly obvious patterns.

Your brain wants "random" to mean "spread out evenly." But real randomization doesn't work that way.

[[PHNX]Kordarus]

RNG doesn't means you will get an even number of each die face. On the long run over many games, it may be tru, but RNG stand for Random Number Generator, Random is the key word here.

[LordBlade]

Yeah, RNG can screw you, easily. I mean, I've had games where it tells me I rolled a 1 over TWENTY times, and maybe got 1-2 rolls of 2, 3 or 4. Not a single 5 or 6.

Admittedly, when that happened three games in a row, it makes you think that the RNG isn't working. But in theory it's doing its job. You just can't really tell until you've played 1000+ games.

[Qfasa]

Hello there.

We have data on all rolls that were made in the game for the past couple of years (plus/minus a few months, I don't remember the exact date when we started to gather it). 49.677.144 rolls has been made in that time. The difference between the most often result and the least often result is 0.43%

So, while the results of an individual session or an individual player in a session might fluctuate, the overall result of the RNG shows that the numbers are evenly distributed across all rolls.

If we start messing with the RNG in order to provide an even distribution of numbers for each individual session or each player in that session, there will be no RNG at all – just a complicated system of rolls distribution that emulates RNG.

[Nekthael]

Plus, as you can't know when a session finish, and what will be the last action, you cannot determine the number of roll. And someone could know when he will roll a 1, more or less, so he won't gamble. And so, you go back to the first problem :p

[WingedKagouti]

Originally posted by Nekthael:

Plus, as you can't know when a session finish, and what will be the last action, you cannot determine the number of roll. And someone could know when he will roll a 1, more or less, so he won't gamble. And so, you go back to the first problem :p

You can technically know what you will roll. This is mostly applicable to single human sessions, but is a tradeoff between letting players save scum to either reroll until they get the optimal result or know exactly what they will roll on every dice. The individual dice are still as random as they can realistically be in either scenario and you won't know how they will fall unless you cheat.

"The dice aren't random because I experienced <x> event" is a common complaint made by people who let their perception of a singular event colour their expectations. I've had Blood Bowl dice that basically went 1+1 on two dice rerolled into 1+1 followed by 6+6 and 6+8 (d6+d8) rerolled into 6+8, which in that situation translated into "Your really experienced player just failed to properly hit that weak and fragile opponent and got killed without being saved by your medic. Also that means your turn just ended after your first action and you don't get to move the rest of your team".

Roll enough dice and extreme "patterns" will occur along the line. That doesn't mean the dice aren't random, instead it more or less proves they are. And as has been stated already, a small sample size offers no proof of randomness or a lack thereof.

[Name Lips]

There's also a very common gamblers' fallacy where we expect past results to influence future results. If we keep rolling 1s and 2s, for instance, we expect other numbers to start coming up. "I've rolled enough low numbers, surely the high numbers will show up soon."

In reality, the number is random each time. A die doesn't know or care that it has rolled a lot of low numbers in a row -- the odds of getting another low number will remain the same for the next roll. The die isn't going to think "well I guess it's time to roll a high number now."

You'll see this all the time in families that have several children of one gender. "Surely if we have another one, it's bound to be a boy after all those girls!" 50/50 odds of gender hold true across the whole population on average, but not necessarily in individual families. Odds are that there will be LOTS of clumping on the small scale.

[Kevin]

All very good points, well taken. Certainly past results do not influence future rolls. If you roll 8 1's in a row, the chances are still 1 in 6 that you'll roll another 1 the next time. I also understand about "clumping", the best illustration of it is in the random dots in a square effect. People are shown two squares on a sheet of paper, one with a nice uniform and even spread of dots, and the other with open spaces and clumps, and are asked which is more random. Many people pick the uniformly spread square, but in fact the clumpy one is the random one.

I guess I don't see that rolling a dice 23 times is a small number, and remember this has happened on two of my games now. In fact, it was somewhat comical that when I landed on a misfortune, I'd say "here comes a one!" <click> and sure enough, it was a one.

There are statistical ways to tell how likely it is, with 23 rolls of the dice, to get either the same "badness" of a result I got, or worse. Maybe someone out there is a statistician, and can chime in. I'm not a statistician, but I do program computers, and am interested in Monte Carlo simulations. So, I wrote a little C program to run 10 million trials of rolling the dice 23 times using Visual Studio 2017's RNG to see what I should be expecting. Maybe I'm way off base, or maybe not, but at least I'll learn something.

An issue is how to weight a particular die roll. In the game, it seems that rolling a 1, 2, or 3 is generally bad, while rolling 4, 5, or 6 is generally better. A 1 is worse than a 2, which is worse than a 3, etc. I needed a way to "score" a game of 23 die rolls. So, I somewhat arbitrarily wrote the code so that during a "game", a roll of a 6 would cancel a roll of a 1, a 5 would cancel a 2, etc. I assigned a higher impact to rolling 1s and 6s, than 2s & 5s, etc. because in the game, 1s and 6s do have more impact to gameplay. I first ran it with a weight (multiplier) of x3, x2, x1 for rolling any imbalance of 1&6, 2&5. and 3&4. BTW: The "score" for my OP game using this method was -26 (negative means "bad" rolls, the more negative, the worse. Positives for "good" rolls, 0 would be a balance of 1s & 6s, etc.). I also tried it with a multiplier of x25, x15, and x5 (for which my OP game score would have been -215.) I then tallied up any games that scored the same or lower than my game.

With the first weighting (x3, x2, x1), the result I got was 64762 games out of 10E6 games were the same or worse score (more negative) than mine. That means you'd expect my game or worse to happen only 0.64% of the time, if the RNG was fair. When I used the weighting x25, x15, x5, I got 48203 games out of 10E6 that were the same or worse score than mine, or 0.48% of the time.

Remember, this happened to me twice, in two seperate games.

ok, so I thought about this and wrote the code pretty quickly, so there may be an error in my logic, if so, please let me know. Otherwise, sorry, I just aint convinced.

[LordBlade]

At least the game doesn't abuse the seed to cheat with the rolls.

Not like say Dokapon Kingdom. THAT game has some serious BS AI.
Because there is no real RNG with computers, it's basically just using numbers off a seed. And when you play DP, with the AI on the hardest setting, it flat out cheats. The AI knows exactly what the seed is, and it plans its gameplay around that. So you'll see it move to a certain spot on the board, which seems pointless, then again, then the third time it gets the perfect roll to land on a great spot, because it knows the order of all the roll results, so it plays based on that.

Every player rolls on their turn, so the AI works in how many players have to go before its next move and all. And you'll see that if you do something that messes up with that order, like using an ability that lets you roll 2 dice on your turn, it will throw off the AI's planning and make it change direction on its turn as it reworks the numbers needed to do things. :p

[Lauleyhn]

You should play Talisman, they have preset dice rolls there, in order to create pseudo randomness. Same as some skills in Dota 2 to create 'balanced' gameplay.

[Standard of Ur]

Goddess of fortune is a harpy.