Battles aren't random, they're scripted

[Sirp]

Hurricane: No I hadn't read that post you referenced. I have now. Thank you for referencing it. That's the kind of test I was talking about. I am now almost certainly convinced, the only reservation I would have is whether or not his simulation of the CivIII RNG is an accurate one.

etj4Eagle: That's another good post. I am now fairly convinced that the civilization III RNG is pretty solid.

I would like to know more about how it chooses the random technology though. Is the heavy bias towards Monotheism and Nationalism intentional? It wouldn't be hard to make it evenly distribute it. Just call your random number function, take it to the modulo of the number of candidate technologies, and choose the technology with the corresponding number that pops out.

As for floating number operations slowing the game down, I seriously doubt it. Surely there couldn't be more than several thousand random numbers generated in the latter turns of a game, and for a modern computer, even 100,000 floating point operations is a piece of cake. I'd imagine it's primarily the AI and pathfinding that slows it down later on.

-Sirp.

 

[sumthinelse]

As usual, excellent research, etj4Eagle.

 

[Hurricane]

Originally posted by Sirp
As for floating number operations slowing the game down, I seriously doubt it. Surely there couldn't be more than several thousand random numbers generated in the latter turns of a game, and for a modern computer, even 100,000 floating point operations is a piece of cake. I'd imagine it's primarily the AI and pathfinding that slows it down later on.

Good catch. hwinkels actually tested it and found out that it has no effect on game speed, so he withdrew that statement in a later post.

 

[etj4Eagle]

Originally posted by Sirp

I would like to know more about how it chooses the random technology though. Is the heavy bias towards Monotheism and Nationalism intentional? It wouldn't be hard to make it evenly distribute it. Just call your random number function, take it to the modulo of the number of candidate technologies, and choose the technology with the corresponding number that pops out.

My guess is that there is some kind of weighting function for the different techs and that unintentional affect that the developers were referring to is the severity of the disparity in probabilities. It is possible that the weighting function was made too complicated and so it in theory it would give a better spread but in terms of normal game values that would be passed to it, it gives a massive bias towards Monotheism and Nationalism.

 

[dguichar]

...read my post.

So i repeat:

Dices do no have memory. They don't remember you rolled a 6 just before. That's called statistical independence...
In computer the number is rememebered: is called seed.

Keep civilized

David

 

[EuropeanPotato]

This is the only game so far(prolly other civ's on this forum too that I haven't look at) that I've seen get picked apart so much. All these ideas are interesting. All kind of wacky threads.

 

[EuropeanPotato]

MAYBE firaxis and all affiliated people with this game's engine and fancy things didn't care. Maybe the just winged it... maybe its just this fancy front. Maybe the think no one would really going into the scripting and take a look how this functions. Maybe some of these probabilities do not have a complicated algorythm. Maybe.

 

[TheNiceOne]

Originally posted by dguichar
So i repeat:

Dices do no have memory. They don't remember you rolled a 6 just before. That's called statistical independence...
In computer the number is rememebered: is called seed.

I think noone answered your post because there wasn't much point in it. No, dice don't have memory and yes, a PRNG does, but the whole point is that a good PRNG behaves exactly as dice: random without memory.


P.S. Thanks to those who found the threads I was unable to find. This thread now contains all the proof we need to hunt down and kill anyone who complains about the civ3 RNG

 

[Grille]

Originally posted by Sirp


I would like to know more about how it chooses the random technology though. Is the heavy bias towards Monotheism and Nationalism intentional? It wouldn't be hard to make it evenly distribute it. Just call your random number function, take it to the modulo of the number of candidate technologies, and choose the technology with the corresponding number that pops out.

-Sirp.

I guess the intention here is keeping a certain balance in the game. Consider a
game with some scientific civs and their free techs aren't biassed towards
Monotheism etc. Generally, research costs then decrease for the (intented)
new expensive techs in a new age when other civs aim for them. Furthermore,
trading could even spread those techs among all or some civs in a very short
time. So a new era is *half-over* at its start.

Another question concerning random numbers:
Looking at the choice of random ai-opponents at game start, are there any
dependencies, especially regarding your civs' traits and/or 'culturally
linked...'-check-box? I've been playing Iroquois often in last games with
7-10 random ai-civs, Aztecs and Americans were always beyond them (and
due to mentioned check-box my neighbors). Ok, could be by chance, but maybe
someone made similar observations or even a whole test on this.

 

[Hygro]

I don't think it's nice to hunt down and kill people

 

[Lt. 'Killer' M.]

Originally posted by Hygro
I don't think it's nice to hunt down and kill people

strike people, replace with flamers

 

[dguichar]

Originally posted by TheNiceOne

I think noone answered your post because there wasn't much point in it. No, dice don't have memory and yes, a PRNG does, but the whole point is that a good PRNG behaves exactly as dice: random without memory.

Somebody stated that was more likely to get 50win-50lose and win-lose-win-lose-...-lose than every other combination. Each one has the same probability to come out, the order does not matter.

Keep civilized

David

 

[Zamphyr]

I would like to know more about how it chooses the random technology though. Is the heavy bias towards Monotheism and Nationalism intentional?

It is intentional, mainly for the reasons Grille stated above. I'm pretty sure this question was asked in one of the Firaxs heavily is chats. I remember the answer being that they intentionally skewed it Monotheism 90% Feudalism 9% Engineering 1%
I cannot remember (and am too lazy to search) which chat it was. It just stuck in my mind because it was this heavilly weighted.

 

[Renata]

I'm almost positive you're wrong about that (the monotheism/nationalism/rocketry thing). I read a thread just very recently where someone quoted a post by the guy from Firaxis saying that the skewing was unintentional, and the result of a mistake in some or other algorithm. He said that it was intended to be an equal chance of getting each starting tech.

Renata

 

[Hygro]

While it makes sense that it would be intentional as to not make the ages already reseached by the time they start, I'm pretty sure it was unintentional.

 

[Park Ranger]

Trying to imagine what it would be like if the scientific civ's free tech was evenly random...no more nationalism slingshot (at least you couldn't count on it), for one. Also if two or more scientific civ's got different techs around the same time they'd probably pull way ahead of the pack.

Some one also pointed out a while back that it would help the human player playing a scientific civ if they were second or third to reach an era, still having a chance at a tech with the max trade value.

I'm so used to it being predictable, it's one of those things I hope they (Firaxis) don't fix.

 

[TheNiceOne]

Park Ranger: My guess is that this will be fixed, since Fraxis' inention is that it should be random. Here's a quote from Mike B. of Firaxis:

Basically, it's a result of the algorithm used to randomly select an available tech. It's unintentionally heavily weighted towards the available tech that appears first in the list of techs (in the editor). It is supposed to select, at random, one tech from the new era that is available to the player (i.e., it has no prerequisites).

As an example, with unmodded rules you have the following available techs in the modern era with these approximate chances of being received:
91% Rocketry
1% Fission
1% Computers
7% Ecology

As you can see, Rocketry is heavily favored (simply because of it's placement within the list). This is unintentional and all 4 techs should have an equal chance of being selected. I will look into improving this for the next PTW patch.

 

[Ribannah]

Originally posted by Midgard Eagle
The "random" battle system isn't random; nor is it about luck or AI cheating: My theory is that the battle system runs on a script.

No script.

You are mislead by the fact that the random number sequence of this turn starts from the last one of last turn, which is kept in the save and autosave files unless you turn 'preserve random seed' off.
Therefore, if you replay a turn exactly as you played it the first time, you wil get exactly the same results.

That said, random number generators that are used in games tend to be slightly flawed towards (usually positive) autocorrelation, because that saves up to half of the calculations and anyone who isn't an expert won't notice this anyway.

RNG's can be flawed with regard to outliers, too, possibly causing Spearmen to defeat Tanks more often than they already should, but again the flaw is so minimal that you won't notice. The shock of seeing this happen at all has a much greater effect on your perception.