What RNG does Civ use?

[Brain]

Some bad assumptions are being made. For a start a pRNG is not a random number generator. It uses an algorithm to generate a string of numbers that appear to be random, but have an underlying predictability.

Okay, but some PRNGs pass statistical significance tests and are used for research that is published in serious scientific papers. You need to arange the numbers in like 600+ dimensions to hope to see any systematic pattern. Not the one Oystein presented though. That one has a very short period. But for a computer game it doesn't really matter anyway unless the programmers did some stupid manipulation of the numbers like I described in an earlier post.

Brain. Nice diagrams. How did you produce them?

"A" was produced using a random shuffle algorithm. "B" was produced the same way, but I manually moved the squares around to satisfy the condition "no sequence of more than 3 of the same kind either horizontally or vertically". I read about a psychological experiment like this in a magazine, but I don't have the original article handy.

When people complain about the RNG producing long strings of wins or defeats they do not neccessarily mean that it is not producing a good approximation of randomness, just that it looks unfair, and they would prefer the pseudo-randomness of Brain's diagram B, to that of A.

Randomness in games is meant to be unpredictable not fair. "A" satisfies this (as far as a computer game is concerned), but "B" does not.

 

[FinnMcCool]

Brennan, I seriously love your signature.

Brain, you end your last post with a crucial assumption. It can't be said, accurately, that randomness is used in a game in order to be unpredictable but not necessarily fair. It's a serious point, so bear with me.

For examples, card games shuffle the cards in order to provide unpredictability. Conveniently, it is also fair because it's always the same cards. This gives rise to that dreaded "exploit" since some players are keen enough to remember all the cards played so far, and they can better predict what card is likely to come next. It's definitely an exploit, and IMHO it's one that shouldn't be avoided. On the other hand, early wargames using a dice system often ran into players who complained about "unfair" strings of bad luck. One solution was to use chits, and every time a RNG was needed, you'd draw a chit, then put it in the discard box. You had two chitpools, both equal, one for each player. Neither player could claim unfair strings of bad luck. This is also exploitable, and some people don't like that; but some people think the exploit is less a problem than the fairness it provides...

Randomness in a game is supposed to provide both unpredictability and fairness--if at all possible. Civ's RNG bites so, because it's questionable whether it's either fair or unpredictable...

 

[Brain]

Randomness in a game is supposed to provide both unpredictability and fairness--if at all possible.

I disagree with you entirely. If you want completely fair then play chess. A game with any kind of random factor is by definition unfair.

 

[FinnMcCool]

Wow Brain, thanks for going so far out on a limb. Your clarity is refreshing.

You are, however, incorrect. "A game with any kind of random factor is by definition unfair" is a completely erroneous statement, but I appreciate where you're coming from. I don't want to get bogged down in semantics, and if someone whips out a dictionary I swear I'm gonna load my .38, but "fairness" isn't by definition removed as soon as randomness is introduced. It's fair as long as all players suffer the same randomness, under the same circumstances. The strategy comes in by manipulating the situation to minimize the cruel twists of fate. Maximize your combat strengths, utilize your mobility, maximize your options, crap, read The Art of War.

The annoying part is, it seems the RNG is more variable than anything you can manipulate easily. Why send in Zerks, or MIs, when the RNG will make an Archer just as likely to win. (I know that's just sour grapes, I apolagize... but the idea remains: how wide a variance is acceptable before the RNG becomes counter-strategic?)

 

[Theoden]

Maximize your combat strengths, utilize your mobility, maximize your options, crap, read The Art of War.

Part of the Art of War is to prepare for unlikely results. In real life generals have to take bad strings of luck into account so why not in civ.
Also, with your method you would have a combat which would normally have 50/50 in outcome to be like 40/60 if you are having a good string of luck. That is not reasonable IMHO.

 

[FinnMcCool]

Sure. Prepare for unlikely results, but it isn't game-friendly when you can expect unlikely results fairly often. Unlikely events should be rare.

Not sure Theoden and I are talking about the same thing tho. I'm not suggesting anything like changing an odds; 50:50 should still be 50:50.

 

[Rohirrim]

It would be nice if they would program out the possibility of a Spearman beating a Tank. The odds of that kind of occurance should be astronomical.

 

[FinnMcCool]

My favorite rant whenever this subject comes up is USE MORE CURVED PROBABILITIES. I won't rant, but just say that if there is a curve at all to the probabilities, I don't see evidence of it. With a steep enough curve, an event like you describe would be truly astronomically rare; meanwhile events like a 1:1 would be extremely variable, which they should be. And events like a 2:1 should be exactly twice as likely as the 1:1, but it doesn't seem like they are. At all.

(edit) an afterthought: well, a 2:1 combat odds shouldn't win twice as much exactly, I suppose. But it should be a lot more likely than a 1:1. We all know how huge a force you need to take out those 4 Spearmen the AI has stashed in every city...

 

[Theoden]

Not sure Theoden and I are talking about the same thing tho. I'm not suggesting anything like changing an odds; 50:50 should still be 50:50.

Sorry if I didn't make my point clear enough. I was referring to your example with cards. If you draw three aces to begin with then the chance of getting another will be 1/49 rather than 1/13. My impression was that you wanted to implement this kind of system to civ so for example a 50/50 combat would always result in 50 wins for either side after 100 battles.
edit: crosspost

 

[Rohirrim]

I'm no mathemetician but it seems to me that while the variabilities in combat (who goes against who) are large, they are finite within the game. I don't see why you couldn't create a pre-loaded RNG that is dependent on the strengths and weaknesses inherent in particular matchups and that also includes in its equations values for terrain, fortifications, etc. In other words, have a pre-weighted RNG formulated for particular matchups, in particular situations. If a Spearman goes up against a Tank, his odds of winning are (say) 10,000 times worse than if he's going up against another Spearman. Your odds of a Spearman victory become astronomical.

 

[FinnMcCool]

Rohirrim, your post is critical, thanks. For your example, if 10,001 people played so far that a Spearman was facing a Tank, one of those players would by all rights defeat a Tank with a Spearman. So now we have two players in 10 000 who've experienced a Spearman killing a tank. Chances are, both players will post in this forum, and probably expressing vastly different opinions.

The surviving Spearman player will likely complain that the RNG is very fair; the defeated Tank player will likely complain that the RNG is unfair.

It's my opinion that it should take more than merely 10 000 events to allow a Spear to kill a Tank. Maybe 1 in one hundred million would be fair.

 

[FinnMcCool]

Sorry if I didn't make my point clear enough. I was referring to your example with cards. If you draw three aces to begin with then the chance of getting another will be 1/49 rather than 1/13. My impression was that you wanted to implement this kind of system to civ so for example a 50/50 combat would always result in 50 wins for either side after 100 battles.
edit: crosspost

I'm expanding this argument way past the first case, but thanks for the numbers in the cards example. If I've been dealt 3 aces, I should have a very high degree of confidence of winning; it's my opinion that the RNG in Civ3 is so drasticly variable that even if my odds are 3:1 against my enemy, I'm not confident of winning.

Sorry to bring up sour grapes, but this idea came to me from seeing 800 years of Berzerkers losing to Archers, which might not be 3:1 (Zerk 5 att, 2 def, amph) but it's close enough to motivate me...

Ahem...

 

[Brain]

Sigh!

There's always going to be people with such experiences no matter what system is used. Personally, I doubt any game programmer will bother implementing a gaussian distribution or whatever other system when a straight uniform distribution is so much simpler.

Anyway, I'm can contradict almost all your statements but I'm tired of this argument. You need to clear up some serious misconceptions.

 

[el toro loco]

if all this is random and not cheating like i expected. then one person out there has to be the most unlucky civ player. i think that might be me : )

 

[Theoden]

it's my opinion that the RNG in Civ3 is so drasticly variable that even if my odds are 3:1 against my enemy, I'm not confident of winning.

You shouldn't be confident of winning. If you could automatically expect to win in certain cases the game would be boring IMHO. I believe a completely random RNG like in Brain's example 1 is the solution.

Sorry to bring up sour grapes, but this idea came to me from seeing 800 years of Berzerkers losing to Archers, which might not be 3:1 (Zerk 5 att, 2 def, amph) but it's close enough to motivate me...

Ahem...

If your berserk was attacking an archer without def bonus the chance of winiing would be very high. But that still doesn't make it 100%. Just remember that you also have wins when your chance to win is 5%.

 

[FinnMcCool]

Sigh!

There's always going to be people with such experiences no matter what system is used. Personally, I doubt any game programmer will bother implementing a gaussian distribution or whatever other system when a straight uniform distribution is so much simpler.

Anyway, I'm can contradict almost all your statements but I'm tired of this argument. You need to clear up some serious misconceptions.

No "gaussian distribution" (or even a whatever) is necessary. I'm always amazed that it doesn't occur to the experts.

A primitive example in the primitive BASIC:

VAR=RND(0)
generates a number betwee 0 and 1 (not inclusive)
VAR=INT(RND(0)*6)+1
generates an integer between 1 and 6 (inclusive) like dice do

or,

INPUT "how many dice", COUNT, z=0
FOR c=1 to COUNT
z=z+INT(RND(0)*6)+1
NEXT c

totals all dice

now, to make a probability curve all you need to do is
VARIABLE1=RND(0)*1024
VARIABLE2=RND(0)*1024
CURVEYVARIABLE=(VARIABLE1+VARIABLE2)/2

This is rude to be so specific, but if you assume the curve has to happen in the RNG, you're wrong. It can just happen in your own use of the generated number.

Now all you have to do is decide where you want your mean to be, and that, fellow Civfanatics, is the hard part. It's where game designers earn their salary, IMHO.