This game, and it's "random # generator" are BS!

[sumthinelse]

I upped the attack/defense for modern units. I understand that may put the AI at a disadvantage because it doesn't really adjust very well to changed stats sometimes....

Did anybody read The Babylonian Lottery, a short story by J L Borges? It's about a supposedly random game that gets weirder as it evolves. (At first there are only 2 possible outcomes: you win money or you don't win money. Then the game evolves to have a third outcome: you can lose money, a LOT of money! Then you can get really good outcomes or a sentence of death. But if you get a sentence of death, another player may get a result such that he tries to intervene and save you from your death sentence. It's hard to explain but I think some of you guys would like the story.) One of the points to the story may be that we, the game players, LIKE a little mystery in our games.

 

[chiefpaco]

The random number is un-accountable by enough to make me use it in my favour too. I started using defensive units on the attack & found it surprisingly rewarding. I've used vet pikemen to successfully attack 2hp pikemen defending towns. I've used longbowmen to beat riflemen. Musketmen to beat cavalry. Even use Impi to damage defensive units & retreat. Once I started using it in my favour, I started LOVING it. The defensive units stay behind anyway, so I don't care if they're very damaged. I think it merely a subjective frustration - i.e. the person who gets gipped is angry. Perhaps we rely too much on the ratings...

Now, I am frustrated about Leader generation. We have no control there. 1/16? Bah. Last game it took 38 elite victories (I counted) to get my 1st leader. Then 1 the next turn. Hardly consolation. 2nd game in a row where I had a +25 stretch. I know I don't win a prize every 6 Coke bottles when it's 1 in 6, but this one got to me somehow...

 

[Moff Jerjerrod]

Originally posted by chiefpaco
The random number is un-accountable by enough to make me use it in my favour too. I started using defensive units on the attack & found it surprisingly rewarding.

You see that could never happen in Civ2 or SMAC/X. The combat model was such that units that were on the offensive but with low attack factors would lose most of the time. That is logical. Even in real life you don't use purely defensive units for offense and expect to win.

Why was combat in Civ3 rigged to favour the AI 99.98% of the time? That right there is what is frustrating. Not the fact that person who gets gipped is frustrated and poo poo to him for being a poor sport. I guess I'm supposed to jump up for joy and buy Sid a beer every time I lose a battle that should have gone my way mathematically. I guess I have a poor attitude. After all a true gamer only plays to watch his units die, not to win.

 

[chiefpaco]

Originally posted by Moff Jerjerrod


You see that could never happen in Civ2 or SMAC/X. The combat model was such that units that were on the offensive but with low attack factors would lose most of the time. That is logical. Even in real life you don't use purely defensive units for offense and expect to win.

Why was combat in Civ3 rigged to favour the AI 99.98% of the time? That right there is what is frustrating.

You totally missed my point. It's not rigged to favour the AI. Just unpredictable outcomes. Your 99% probably comes from the fact that you always use superior forces & don't experience the brighter side of things. If you want to argue that it's too random, I can accept that argument. I don't agree with the AI favouritism though.

 

[BlueBaron]

This thread really took off, didn't it?


Well to those who continue to be smug and condenscending, you know what you can do with yourself. Go find yourself a convention of l33t peers and stfu.

For the rest of you who have actually contributed to this discussion . . . well, thank you.

I may try to get hard numbers to support my claim. However, let me remind you . . . saying that the randomness may be a "long term" situation is a cop-out. I'm sure if I played 2000 games, over several different platforms . . . you could probably find a case for randomness. That's no excuse. These games aren't supposed to last for a month, therefore the only randomness that's useful, is in the short term. Furthermore, Alc0p0pz hit on something that happened to me which started my scrutiny of the battle outcomes. I don't care WHAT the circumstances are, one, lone unit should not be able to defeat an army. That's assinine. C'mon, an army of cavalry falls at the hands of a lone rifleman? Doubtful. Where's the law of randomness there? The army has enough HP that in the course of that battle, there should have been plenty of room for even unlucky rolls. BS . . plain and simple.

I'm glad to see some non Sid-drunk people are in this discussion as well. Keep it up.

 

[Tassadar]

Originally posted by chiefpaco
The random number is un-accountable by enough to make me use it in my favour too. I started using defensive units on the attack & found it surprisingly rewarding.

i put a defensive catapult on a hill defended by a pickman in my last game. beleive it or not but my catapult always fire a shot at any barabarian attack. catapult is now a defensive weapon in 1.17

 

[Moff Jerjerrod]

Sorry if I sounded mean Chiefpaco, I guess for me Civ3 has been a rollercoaster ride of emotions.

I've decided to take a break from Civ3 until the new patch comes out. Despite my bitterness I'm still strangely addicted to this game.

 

[etj4Eagle]

I seem to remember a thread a while back from someone who had managed to run statistical tests on the random number generator. And the tests showed that the generator was working properly and generating numbers with the correct distributions. That seemed to prove that the generator was not flawed. Though there is always the possibility that there is some kind of system dependent behavoir in it as well, which may make it flawed on some systems.

 

[damunzy]

Originally posted by etj4Eagle
Though there is always the possibility that there is some kind of system dependent behavoir in it as well, which may make it flawed on some systems.

This leads to my thread (shameless plug). http://forums.civfanatics.com/showthread.php?s=&threadid=20211

I was wondering if it could be dependent on the computer you have but someone in the thread made a valid point as to how thinks I saw as weird could be based on game rules that I just don't look for.

Well, anyways, back to Jedi Knight 2

 

[Sanaz]

At higher levels, does the AI have an advantage? I only play at the lower three levels and don't see much difference - sometimes I get a strange victory, and sometimes a strange loss, but it sure seems random to me. Generally, the stronger unit wins, depending on who's territory the fight is in, the terrain, etc...

 

[Velikovsky]

In my experience, your complaining is unjustified. Perhaps the default game rules lead to too many combat upsets, but that isn't really a random number generator problem.

What I have noticed, however, is that the random number generator seems to be predetermined. If you have a combat with an outcome you don't like, and you reload the game to do it again, the combat will not simply have the same winner or loser, it will have the same exact order of battle wins (hit-point reductions). If you do the same thing with 5 units attacking the city, you get the same exact hit-point outcome. I think I tried up to about 10 units.

There is a way around it, though. All you have to do is use your units in a different order. This leads me to believe that the number generator is recursive, or at least based in some ways on the outcomes of previous battles -- something I would guess without these experiments. Battles definitely seem streaky. Not that I have a problem with that. Momentum in battle seems fairly sensible.

So anyway this is an actual experiment that points out a quirk of the random number generator. I did this a while ago, and never intended to post about it, so I haven't done a long documented analysis of it, but if anyone else out there cares enough, you're welcome to it.

Whatever the case, don't let the random number generator get you so pissed off. There's way too much anger on this board. You guys all need to smoke some herb before you post on here. jeez.

 

[ainwood]

Velikovsky:

I tried to explain (part of) this in the thread the PH (shamlessly ) linked to - I probably failed miserably . The random number generator is simply a mathematical formula that spits out a result based on the number that is input into it. This input number is called a "seed" number.

Every time Civ III runs a random number, it saves the results and uses this as the 'seed' for the next random number. Hence, when you reload, it is still using the same seed, and hence will return the same "random" number.

I understand that this was a conscious design decision to somewhat prevent people reloading just 'cause they get a bad result.

 

[Kill-O-Byte]

Thanks for the posts Killer, Velikovsky and ainwood. That explains a lot. While I tend to have few unnatural results in my battles, my wife has had strings of very, very bad luck that frankly I wouldn’t have believed if I hadn’t seen it myself.

BlueBaron, you’re not the only one this happens to. Hopefully your luck will change soon.

 

[rev063]

Originally posted by Moff Jerjerrod
why should we have to change anything in the editor to make civ3 more enjoyable? CIV3 SHOULD HAVE BEEN ENJOYABLE RIGHT FROM THE START WITHOUT ME HAVING TO EDIT THE GAME TO MY LIKING.

This is clearly a specious argument -- who is the mythical "we" in the above statement? Different people have different tastes, and as the saying goes, you can't please all of the people all of the time. I enjoy Civ3 as it is.

Incidentally (and to get back on-topic), I think part of the reason for the disgruntlement regarding the randomness of the Civ3 RNG (which, as a statistician, I am thoroughly convinced IS truly random) lies in the way the battle animation is displayed. Attacks are shown in pairs (attacker and defender), and for each pair exactly one is a "hit" (reducing the HP of the "loser" by 1). This means that, say, three successive rolls in favour of the defender appears as three "misses" for the attacker and three "hits" for the defender in the animation, making the event seem more unlikely than it really is.

Here's something to chew on: 5 successive defender wins (which would kill an elite attacker without loss of HP) can happen with 1/20 probability (the usual statistical test of "significance") when the attacker's advantage (probability of hit) is as high as 45%. Consider that defenders often have a significant advantage over attackers, and a complete mauling of an elite attacker isn't as unlikely as you may think.

- rev

 

[SirPleb]

Originally posted by Park Ranger
Is this a problem? Not for me. I tend to plan like Killer does - have lots of extra attack units, Just In Case. But this won't translate into an advantage over a riskier strategy that gets lucky, in fact the cautious player has spent more to win than the risk taker. HOWEVER, the cautious style DOES pay off if you know how to turn the lucky breaks (which you never count on), into real advantages.

I think that's a very important point!

Despite my greedy warmongering tendencies ( ) I think of myself as a cautious player. To my mind the key, for the cautious player, is to go for your primary goals only once you have overwhelming odds of winning those goals. When planning a war, decide first on your primary objective. (E.g. getting a resource, a luxury, a wonder, a key geographic point, a toehold on another continent, etc.) Go for it once you have overwhelming odds of success. Perhaps think of it as a casino - do you want to play as a gambler in the casino or as the house? The gambler sometimes wins. The house always wins, by using large enough numbers that individual lucky or unlucky streaks don't matter. Two Knights (or even 3) against one Spearman is not overwhelming odds. Very good odds yes. But not good enough to treat as a sure thing. On the other hand, 20 Knights against 10 Spearmen is overwhelming. Use large numbers to your advantage in your overall strategy.

After gaining your primary objective(s) with overwhelming force it is very likely you'll have a large force left over. (Probably 16 or more of those 20 Knights ) If not, fine, you gained your critical objective. In the more usual case with lots of power left over, go for secondary objectives and keep going until you need to stop. (Either because your force becomes too small or because you want to refocus toward a new primary goal elsewhere.)

If your objective is fuzzy (e.g. "expand territory") the same principle works fine. Start with enough force to be nearly certain of gaining at least two or three towns.

The only goals I can think of which don't work well with this general approach are:
1) Earliest possible conquest. I think that often an early gambit may lead to the earliest possible win. If 10 people are competing for an earliest date on a tiny map, the winner will probably be someone who got lucky in early battles; not one of the cautious players, and not one of the ones who tried the early gambit and lost. Luck will be a significant factor.
2) Getting a leader. The number of elite wins required before you get a leader can vary enormously. I've had it take from 1 to over 40. When your goal is to get a leader (e.g. to build Forbidden Palace) there isn't any reasonable amount of overwhelming force which makes success predictable. You have to throw the dice on this one. (And sometimes keep throwing and throwing. )
3) A tight spot. Most games will have times where things don't go according to plan and you just have to make a move with less than overwhelming odds. But these situations should be ones where your hand is forced, not ones you choose.

 

[Tweedledum]

Originally posted by sumthinelse
I upped the attack/defense for modern units. I understand that may put the AI at a disadvantage because it doesn't really adjust very well to changed stats sometimes....

There seems to be a lot of people here who have tinkered with the values supplied in the game.
This tells me something... That these initial values are simply unrealistic (yeah, I know about the realism argument)

BTW: never mind Spearman/Tank, I had an AI warrior kill a Modern Armour the other day (presumably the blood and bone gummed up the caterpillar tracks somehow...)

 

[sumthinelse]

Originally posted by SirPleb



The only goals I can think of which don't work well with this general approach are.....

2) Getting a leader. The number of elite wins required before you get a leader can vary enormously. I've had it take from 1 to over 40. When your goal is to get a leader (e.g. to build Forbidden Palace) there isn't any reasonable amount of overwhelming force which makes success predictable. You have to throw the dice on this one. (And sometimes keep throwing and throwing. )

Yes. Getting a leader is a 1/12 or 1/16 probability, and don't be too surprised if it takes 30 or 40 attacks with elites. As explored earlier in this thread, one way to cheat the game is to do things in a different order. If you REALLY need a leader, like for the Forbidden Palace, save the game before you attack with an elite unit. If you don't get a leader, restore and bombard or attack with a non-elite unit, then save again before you attack with the elite. The same principle applies to a strategic resource drying up. Save your game at the end of each turn, such that there is one more bombard or attack you could do. (You could even bombard one of your own roads!) If your last coal deposit dries up, restore your game and bombard something before you end your turn. That creates a new "seed" for the RNG.

Or don't worry about bad luck. Play the sh***y hands you are dealt and have fun with it. Spend 2000 years building the Forbidden Palace at one shield per turn and then laugh and have fun when some other CIV finishes it in one turn!

 

[sumthinelse]

Look at it from the designer's point of view. He could make the number really random: the "seed" would be calculated based on the number of microseconds (or other small time unit) elapsed between 2 events in the game, one of which could be when you last clicked the mouse. But since you are free to save the game before you attack, if you didn't like the combat result you could keep restoring and attacking until you not only won the combat (and you would win because the number generated would be different each time), but you won and did not take a single damage point. You can do that kind of thing in games like Panzer General (SSI). So the solution that CIV3 took was that it would base the result on some static condition. A kind of number that would give you the same result no matter how many times you saved and restored. Now players have found ways to exploit that too. For example, bombard, or sacrifice a crummy unit (come to think of it, sacrificing crummy units works in real battles sometimes), when you know the "roll" is going to be bad.

Even if the number is really random and changes every time (as in Panzer General), the human player thinks something is wrong when he loses too much. I guess it's human nature. And when the result never changes, as in CIV3, the human player can't remedy the "injustice" of losing when he attacked with a superior force by restoring and trying again (except by changing the "seed" by making a different attack).

If I take back moves, I can beat a really strong chess program, because I will eventually find its weakness, but it will kick my *** if I don't take moves back. I guess the Firaxis people wanted to make the game more challenging. Don't we b**ch when the game is too easy to beat? You bet we do!!!

Maybe we should have an option to get a real, new random number on each combat. I think that would make a lot of these complaints go away. But then they might disallow saving a game before combat. (Did anybody ever play Spellcasting 301? You went into the casino, and before you placed your bet you tried to save the game, but the casino manager said, "No saving games in the casino -- house rules!")

But if you who are reading this never saved the game before combat and tried it again because you didn't like the result, that option wouldn't help.

I have not seen anything in my games that looks like the AI is favored in combat, but that doesn't mean that it can't happen. Maybe I haven't encountered the right environment for it to happen.

 

[Hurricane]

Originally posted by Tweedledum
BTW: never mind Spearman/Tank, I had an AI warrior kill a Modern Armour the other day (presumably the blood and bone gummed up the caterpillar tracks somehow...)

With your modern armor at 1 HP and an elite warrior attacking (5 HP), the chance for the warrior to win is 24.2%. Not that improbable, actually.

 

[Carbon_Copy]

It's been reasonably proved that the RNG itself works as random as can be expected, but people still claim the game to be really streaky, and I am not so sure that every incident can be blamed completely on just plain bad luck.

Perhaps it is a programming error (or undocumented feature) that the defending units only have one random combat number per turn. If this were the case, even if the RNG were pristine the game would still be streaky because the units with high values would be consistently hard to beat that turn (which is something that I've noticed), while units that drew a low value at the beginning of the turn are pushovers. Does this seem to jive with what people are noticing in their own games?