This game is so arbitrary, makes me quit.

[Bartleby]

I don't think you can count a stack as big as that.

 

[MAS]

Or even easier, mod the units so they all have 1hp and just count how many are left.

 

[MAS]

I don't think you can count a stack as big as that.

Then make it 10 stacks of 100 units, or 20 stacks of 50 units.
Put it up so that units from one stack can only attack the corrosrponding defender stack.

So make 20 2 tile island and on each island you put one stack for the offensive units and one stack for the defensive units.

Or better yet: make that 40 2 tile island, So you can put both tests in one scenario. 20 island where the human defends, and 20 islands where the AI is the defender.


(Edit:Or maybe someone can program a small app to check a save file and do the counting? )

 

[Chamnix]

But the claim has been that the RNG only favors the AI at some points (for example, when the player is winning too easily) - what happens in a set-up scenario will not disprove that claim.

I'm willing to accept meisen's data as proof of something as long as it is reproducible.

 

[MAS]

But the claim has been that the RNG only favors the AI at some points (for example, when the player is winning too easily) - what happens in a set-up scenario will not disprove that claim.

Give the Human 2000 extra units and 50 extra metropolis cities (on a separate island, so it doesn't interfere with the test battles)

I'm willing to accept meisen's data as proof of something as long as it is reproducible.

I have an additional demand: make it easy to understand.

With my test, you don't even have to know odds, just compare AI results with human result. If the AI cheats, it can be seen at one glance.

1st attack: 6hp p vs 4hp ai odds=97p/3ai win=p with -3 with -hp odds of 13/97 or worst 1/4 of the range for the p unit.

VS

Human loses 100 units AI loses 80 units: so the AI has an advantage

 

[Jokeslayer]

But the claim has been that the RNG only favors the AI at some points (for example, when the player is winning too easily) - what happens in a set-up scenario will not disprove that claim.

I tested this myself. 100 AI warriors vs. 100 human warriors with no defense bonuses and all set to 1hp for all experience levels. Gave the human a continent full of cities, all techs and a bunch of mechs (about 50, I think). Then let the AI attack the human warriors. Ran it several times. The human side won 280 out of 572 trials. No evidence there of AI cheating.

 

[timerover51]

I tested this myself. 100 AI warriors vs. 100 human warriors with no defense bonuses and all set to 1hp for all experience levels. Gave the human a continent full of cities, all techs and a bunch of mechs (about 50, I think). Then let the AI attack the human warriors. Ran it several times. The human side won 280 out of 572 trials. No evidence there of AI cheating.

Jokeslayer, where I have seen the problem is not when the AI and the human player are at the same tech level, or close to the same. Where I have seen the problem is when I have a fairly good tech level edge on the AI. Like riflemen against legionaires or spearmen, or galleys against ironclads or cruisers, both with boosted stats, and the Ironclads with an 8 hitpoint boost. When one spearman on a hill can chew up three cavalry and send them all retreating, while surviving an incredible amount of time with a single hit point, I start having problems with the system. Or when a war elephant takes down a 12/18 mech infantry unit without loosing any hitpoints. I actually do better attacking Roman legionaires with Crusaders than I do with Infantry units set to 8/10 with 6 hitpoints for being veterans.

And MAS, I am on Macs as well, and the weird results were so irritating that I bought a surplus Windows box from my son's high school just so that I could do editing on Civilization 3. In the test scenario that I am running now, I am building Space Ship Parts and nuke subs with tactical missiles while the AI Aztecs are hitting me with Medieval Infantry, galleys, caravels, and trebuchets. I have boosted the stats on my naval units and some of the land units. What I am seeing is not in the realm of reasonable probability. I design miniature and board war games, and have a dice collection from 4-sided to 30-sided to use to generate various probabilities or achieve a somewhat historical outcome, playing with combinations until it feels right, with the low tech guys having some chance to win, but the high tech players getting the benefit of investing in high tech. What I am seeing is as if the AI is rolling a 12 to 16-sided die while I am rolling a 6-sided die. By comparison, in the WW2 in the Pacific game that I am playing, where I have worked some on the unit values, I am not seeing weird results, but the combat results are in line with what I would expect and in line with historical results.

Where the AI bias shows up is when you have either a high tech verses low tech situation or when you have a disparity in hit points. Your matchups at even tech levels are not going to trigger whatever subrountine is operating at the point. I do know that if I designed a miniature or board war game where ancient galleys could chew up ironclads and WW2 type cruisers, I would be laughed out of the convention where I was playtesting the game. Zulus taking out a British infantry battalion equipped with single-shot Peabody-Martinis can fly, Zulu Impis taking out a WW2 tank battalion will not fly.

 

[Chamnix]

Save please .

 

[Pyrrhos]

A few of things wrong with asking for saves all the time, Chamnix: First, there has to be a save, obviously. Second, the game has to be set to "preserve random seed" for you to see the same thing - and you're going to cry "foul" if you don't even if you knew that it was not set to "preserve random seed". Third, if saves with the "preserve random seed" were provided and you saw the same thing, you would instantly dismiss it because "all the times the oucome favours you have not been included and this is just a single occurrence when it didn't out of tens of thousands of games."

Now, I may be wrong here, but I tend to trust people to tell the truth and not make things up. This is probably because I am an honest person myself.

 

[Jokeslayer]

~snipped for length~

So your argument is only that more advanced units are disadvantaged by the AI? I misunderstood, thinking that people were saying it was the civ that was penalised, not the units specifically.

Well, why don't you test it? Run and record a proper number of trials of your claim, rather than just stating what you believe to be true. I am now creating a test with Infantry vs. Warriors and will post when I'm done. And if somehow this doesn't meet your requirements, I think you should post them explicitly, or do it yourself.

@Chamnix: is that "save please" aimed at me?

EDIT: My results are in. Out of 600 combats, the infantry lost 57. Since they should lose one out of every 11, that's an expected loss of roughly 54. If that's the kind of bias the RNG is throwing out, I'm not too worried. Obviously this doesn't speak for if the RNG allows the disadvantaged units (I'm not entirely clear if you think the "low tech bonus" only applies to AI units, or human ones as well) to take extra HP from the better ones than they should. I might test that later.

 

[Chamnix]

A few of things wrong with asking for saves all the time, Chamnix: First, there has to be a save, obviously.

Meisen played his battles 3 times – he must have had a save to do that, and he could have posted the save right when he posted the results. So, maybe he played the save 3 times, was satisfied with the results and so deleted the save immediately afterwards. That’s OK. This bias is something that happens fairly regularly, no? How hard is it for one person to get to a point where this bias occurs and then go back to the latest autosave? Why is it that not one person has one save where this bias happens if it happens regularly?

Second, the game has to be set to "preserve random seed" for you to see the same thing - and you're going to cry "foul" if you don't even if you knew that it was not set to "preserve random seed".

Actually preserve random seed should be off. I don’t want to see what the person saw – that would simply prove that the person found one statistically improbable but not impossible event, and you are right that that would not satisfy me. What I want to see is a battle where the RNG is biased in favor of the AI run over and over and over and over and over and over and over and over and over and over again. I don’t care that someone can find and post a strange result – I know strange results happen. But if the RNG is biased, then I will not get the exact result the person got, but I will get a biased outcome over the course of many trials – that is the only way to prove or disprove this theory.

Third, if saves with the "preserve random seed" were provided and you saw the same thing, you would instantly dismiss it because "all the times the oucome favours you have not been included and this is just a single occurrence when it didn't out of tens of thousands of games."

Give me some credit here Pyrrhos. If you give me the ability to test it as many times as I like, then I would never claim that it wasn’t tested enough times. Although I am very skeptical, believe it or not, I actually want to know whether this theory has any merit. If the RNG is biased in favor of the AI based on some triggering event, then I want to know what that event is for my own potential advantage.

Now, I may be wrong here, but I tend to trust people to tell the truth and not make things up. This is probably because I am an honest person myself.

You can trust people to tell the truth and still be aware that people are sometimes wrong.

@Jokeslayer – no, that save please was not directed to you. It is directed toward anyone who has an example of a point in a game where the RNG is biased.

 

[Marsden]

I think a big part of the problem is that the game is too random for realistic results. Most games do use different systems, whether it be different tables or sided dice where civ 3 employs a ping pong ball type lottery machine which could potentially give 18 '1' s in a row giving a warrior a chance to kill 3 modern armors. I lost a modern armor to a swiss pikeman once. At the time it struck me as more odd that a problem. The one place where I do take some exception to the AI's "abilities" is when their cavalry attack my tanks and I loose quite a number of them. Also, the naval combat results are very erratic. I've had far too many enemy galleys just fight off what ever I attacked them with. I think this has more to do with the arbitrarilly assigned A.D.M values than the system itself, but clearly a galley should not be able to sink a metal vessel. The most a galley should be able to do is hope for a hurricane to come and sink the enemy ship while they row away.

 

[Pyrrhos]

@ Chamnix -

 

[PaperBeetle]

The one place where I do take some exception to the AI's "abilities" is when their cavalry attack my tanks and I loose quite a number of them.

Nothing unexpected in that is there? ATT 6 vs DEF 8 + terrain etc, so the cavs can be expected to win about a third of combats. If I am conducting a trench war against a runaway AI in the industrial era, I am actually pretty happy when they get Motors, because they will start spamming me with tanks instead of infantries, and I can kill tanks more easily.

 

[Marsden]

Nothing unexpected in that is there? ATT 6 vs DEF 8 + terrain etc, so the cavs can be expected to win about a third of combats. If I am conducting a trench war against a runaway AI in the industrial era, I am actually pretty happy when they get Motors, because they will start spamming me with tanks instead of infantries, and I can kill tanks more easily.

I understand that, but I think it's not the rng I'm upset but the somewhat arbirary rankings that allow cavalry to kill that many tanks. Does it seem reasonable that in 1944 when the Axis was being set upon by two fronts that all Germany needed to do to stop the onslaught of Shermans and T-34s is get a bunch of horse cavalry? That's what I mean. But it's a game, obviously, and I think safe to say, as I have before, we've put more thought into this than the designers that were paid to think about it. The arbitrary nature of X amount of longbowmen (which we know a higher level AI can easily accrue on an island) killing infantry or even armored units can just frustrate a person.
If your privateer can't take out a galley because our cannon are no match for their arrows, you... might just be suffering bad rng.

Finally, I'd like to add, I've named the rng Chrom. Just thought I'd throw that extra silliness out there. If you can't have fun, why bother?

 

[classical_hero]

This is where bombardment is very useful in civ3. Bombard the crap out of the units you want to attack and then the odds are stacked in your favour. Even in Civ4 you still get some stupid results. I know that i lost a unit that had 99.5% chance of winning and yet it lost. Needless to say, that was very frustrating to lose that battle. Stupid RNG.

 

[FLump]

Well here's my take on the the whole thing

Obviously, if a bunch of dudes with spears charge a battalion of tanks, they would lose stupidly. But what if a bunch of technologically challenged spear-folk instead lure the tanks onto a narrow path and trigger an avalanche which sends the tanks into a gorge? Total -age. That is how I justify it in my mind. War isn't about fighting fairly. If you are sneaky enough you can win.

And as far as the Cavalry beating up tanks - didn't you watch Indaina Jones 3? Indy + horse pwns tanks.

 

[TheOverseer714]

Maybe it is time to put this one to bed. The RnG sucks. Face it, live it, deal with it. No amount of tweaking can fix it. If the game sucks, why play it and kvetch about the negatives? If you like the game, but hate some things about it, what can you do? At some point, you can't tweak it enough to fix it without fundamentally changing the game beyond recognition. The uncertainty of combat results is probably one reason why this game endures where others that have more certainty are just "flavor of the month" games. Everyone in this thread who still plays Age of Empires 1 regularly, raise your hand. Everyone who still plays Rise of Nations regularly, raise your hand. Those games had pretty much certainties in their combat results, and not many still play them. So you might as well just enjoy the game as is, or change it to a different game, but this one still has those programmed dice rolls that are going to make you Live with it. Besides, admit it, you like a game that isn't perfectly predictably, or you wouldn't be on this forum.

 

[Pyrrhos]

Maybe it is time to put this one to bed.

For you and others who have been here a long time, perhaps so. But new people will "discover" this and be intigued/upset/infuriated by it to the point where they feel the need to vent their ire publicly.

But there is another side to it - hopefully Firaxis will realise that Civ III is, by far, the best version to date and release a Civ V. Now, hopefully, it will not simple be a graphics tweak, but a completely revised Civ III, which addresses the points that "attract most negative attention". Now consider this:

If there are NO complaints, there is no reason nor a market for a Civ V...

 

[timerover51]

Maybe it is time to put this one to bed.

I would put it to bed if it were not so irritating. I have been playing board and miniature war games since 1970, and started tweaking rules shortly thereafter. I have been doing game design since the mid 1980s, plus doing some game design work for the government. I can understand about unpredictable results and use them, ranging from chance cards for weather to vagaries in the reliability of bomb fuzing. The low tech verses high tech unpredictability is what irritates me so much. I bought a Windows box for the sole purpose of editing Civ3, which as a Mac owner borders on heresy. Since I do not have an editor for PTW or Conquests, I did not have a choice. I will keep tweaking things until I figure that I have done the best that I can within the parameters of the game, and try not to ask my brother to crack open the source code and do my own reprogramming of it.

I keep wondering if I should break out Trevor Dupuy's Qualitative Lethality Index equations to determine exactly what values to use for the units, but I am not sure that I am interested in that much work, plus the equations start breaking down when it comes to armored vehicles, and really break down when it comes to aircraft, and he never had the foggiest idea as to what to do with warships. However, clearly I am not going to be able to totally remove the problem, but hopefully, I can get it to a point where I can live with it. I like the game, otherwise, I would not be putting this much time and effort into it.