 |
|  |
 |
|
Golden Bear
|
 |
Oregon
Feb 2002 time: 21:37
|
|
For a lead in... I have tried to remain objective about AI cheating in CivIII. I have tended to believe that it was a figment of imagination built upon poorly documented instances of occasional long shot odds. I mean, other than the way that AI civs always seem to miraculously settle next to resources that won't appear for Eons.
I'm also indifferent to the way the game plays as all the AI civs vs. me. I'm OK with that.
However, I just saw the belief breaker.
Pangea, full land, huge. I have spotless rep. Really. Never broke a treaty. Everybody except the Zulu (I'm executing them) is polite or better with luxury incentives to boot.
Bingo, the Babs break double treaty and attack from ROP rape - sorry, but the "how to beat the AI on builder" does not really work, BTW.
OK, I'm cool with that. I am a couple of techs ahead of everybody. Despite a land grab following AI wars I have AT LEAST 1 infantry in every city while no civ has an offensive unit better than cavalry. Cav, of course is very good. But how good?
How about, 11 cities attacked with single Cav vs. single Inf, for, what would you guess as results? Nope, you are wrong. 11 cities taken. Each single (1 attacker) cav captured or destroyed a city protected by inf. I'm thinking, what, 6 attack vs. 10 (doubled for fortified?) is less than 1:3 chance of victory. I should have lost 3 cities, maybe 4.
Remember now, that cav will not attack a pile of arty with a single inf guarding it. Suddenly, boom, every attack goes good.
Calling the number dudes out there. Can you give me the odds of this one? Aw, I know that it is impossible. Shoot, I just lost 3 tanks and a cav trying to finish off a 1 square musketman in a city. I did not whine (then) but maybe a little moan now.
I really don't mind a little AI cheating. But cmon. Oh well, I guess that I'll have to go make them go cry for their mommies. I'm going to win anyway, but this much desperation by the code is embarrasing.
AFAIC this puts the "does the AI cheat in combat" question out of reach.
Golden Bear (with his eyes wider open now!)
|
|
|  |
 |
|
Golden Bear
|
 |
Oregon
Feb 2002 time: 21:37
|
|
And I did not intend to use only a single defender - it's a period of rapid resettlement due to WWs and it takes time to build up enough units to double everywhere.
I was surprised that I would get any attack since typically the AI does not seem to want to violate a ROP if they are polite with you.
[start edit]
VXMA was kind enought to remind me that there is a 25% bonus for fortification without walls. Thus, infantry defending should be 10+2.5 vs. cavalry attacking at 6. This should mean that the cav has about a 30% chance at winnning. 0.3 to the 11th is .0000018 or .00018%. If I did it a million times it should occur twice. I wish that I had rerun the turn! I just moved forward and cleaned the Babs out of every place they had invaded.
[end of edit]
A problem with ROP rape for the AI is that they don't plan well and wind up with their units spread all over the map in penny pockets far from their home. And I was trying so hard to be peaceful. Sigh.
Golden Bear
Last edited by Golden Bear on 13-08-2005 at 22:29
|
|
|  |
 |
|
Mace
|
 |
Trade Empire of the Carthaginians
Nov 2003 time: 23:37
|
|
quote: Originally posted by Golden Bear
Bingo, the Babs break double treaty and attack from ROP rape - sorry, but the "how to beat the AI on builder" does not really work, BTW. |
I agree, at least on Monarch and above. You can't be
*solely* builder on the high levels, or you'll get your keyster handed to you 99 times out of 100.
quote: Originally posted by Golden Bear
How about, 11 cities attacked with single Cav vs. single Inf, for, what would you guess as results? Nope, you are wrong. 11 cities taken. |
Now, you know, if you or I tried to attack 11 enemy towns each with 1 cav, we would laughed off the continent! I would get maybe...maybe 1 hit on the AI infantry health bars...
|
|
|  |
 |
|
lebensraum
|
|
heyya gb,
i suspect the problem is in the 'randomising' algorithm that is used. basically, the rng doesn't (generate random numbers that is). it does a reasonable job of spitting out numbers that look random, but it is not genuinely random, and sometimes it gets itself stuck in one of these ruts.
much as i would like to blame the ai, firaxis, sid, soren, george bush and the sicko who introduced me to civ in the first place, it's really not their fault. the only way out of it will be when we get a true, hardware-based rng. simple enough, just measuring something like thermal noise across a resisitor would do it, but until someone takes the trouble to build it into the hardware, there's nothing to do but grin and umm, you know.
|
|
|  |
 |
|
UnOrthOdOx
|
 |
Hello....
Jun 2002 time: 22:37
|
|
Ok...
Forgive me here.
In the top post, it was 11 infantry killed:
quote: How about, 11 cities attacked with single Cav vs. single Inf, for, what would you guess as results? Nope, you are wrong. 11 cities taken. Each single (1 attacker) cav captured or destroyed a city protected by inf. I'm thinking, what, 6 attack vs. 10 (doubled for fortified?) is less than 1:3 chance of victory. I should have lost 3 cities, maybe 4.
|
My first reaction was that this was indeed a little out of line, and a tremendous stroke of bad luck. (then, I've seen 32 cavalry lose to a single spear before) My instant question was to ask what experience the respective forces were. 5 HP cavalry on 3 HP infantry wouldn't be TOO especially out of the realm of possibility for example.
A little later, it becomes an entirely plausible attack on 11 RIFLEMEN:
quote: Argh! Once again, here is the situation.
Yours truly is sitting in a position that he has ROPs with everybody and is feeding them luxury bribes to keep them happy. Everybody is Polite or Gracious and has been for a long time.
That is, except for the Zulu, who I was taking out of the Civ pool so that everybody would fit on one screen for F4. And the Zulu had no friends.
At the same time, there were multiple wars going on and leaving settlement space available so that I was putting settlers out rapidly too.
Thus, I had double Riflemen in my core and most of my cities but I had not had enough time to double defender ALL the new ones. I was putting out new settlements almost as fast as defenders at that point. Naturally these were not important cities that I lost.
That is why there were 11 cities with single Riflemen defenders.
|
Well, I'ld expect similar, or close to it, results from such an attack there.
So, which was it?
Last edited by UnOrthOdOx on 16-08-2005 at 02:59
|
|
|  |
 |
|
inca911
|
|
St. Paul, MN
Jan 2000 time: 05:37
|
|
I generally avoid ROPs. I do not want the interior cities of my empire to require defenders, and having enemy forces wandering about is no good. By avoiding ROPs, I can push my defense out to the borders and use a couple mobile forces per section of border as a rapid response in case of problems.
I do use ROPs once I am on my own culturally-covered continent. This allows me to sneak defenders and settlers into the culture cracks created during other civs' wars.
|
|
|  |
 |
|
Drakan
|
 |
Málaga, Spain
Jun 2004 time: 06:37
|
|
quote: Originally posted by Golden Bear
I have spotless rep. Really. Never broke a treaty.
Bingo, the Babs break double treaty and attack from ROP rape - sorry, but the "how to beat the AI on builder" does not really work, BTW.
|
I think Golden Bear you are confusing reputation with attitude.
Reputation allows the gpt trade with the AI
Attitude is what normally holds them off from attacking you.
In the first paragraph of my article How to win Deity Builder-style, step-by-step I warn people to find out what each stands for in C3C before reading any further.
When I posted my strategy the whole point of having a spotless reputation is so you can trade techs with the AI for gpt and cash in on it. They will not allow gpt deals with a flawed reputation and hence the strategy wouldn't work out.
This doesn't mean they can't attack you. Because attitude is a whole different matter.
You may have a spotless reputation and still be attacked with a ROP rape even If they are "polite".
You have to take preemptive measures to avoid being attacked by improving their attitude towards you. To achieve this goal you must:
- avoid razing many AI cities/starving AI citizens
- establish embassies
- establish ROP asap (specially this one)
- gift 100 gold (not more) every ten turns to those AIs which may pose more of a threat (i.e. are nearby)
- gift luxuries if needed
- trade alot
Before posting the article I had won more than ten Deity games using it and never got attacked by the AI (ROP rape) at the most AI aggressive setting. Although in a Sid game I did get attacked using it.
You have to take care both of your reputation (enables the gpt trade) and of their attitude (stops them from attacking you).
Last edited by Drakan on 24-08-2005 at 17:13
|
|
|  |
 |
|
Golden Bear
|
 |
Oregon
Feb 2002 time: 21:37
|
|
Hi Drakan. I may have appeared to confuse rep and attitude because I expressed myself incorrectly.
Also, I overstated the "...does not work" part. It does work and work very well. I got caught because I was deliberately pushing the technique to see how much I could get away with.
Now to respond to your list:
- avoid razing, etc.: I never do this in any game. Also, I never starve cities. I pay for immediate Temples as soon as the city is calm. I don't think this contributed.
- establish embassies: I did this immediately upon getting the cash buildup while using GL for tech advance. Embassies with everybody.
- gift 100 gold: got me here, I never gift gold. I gift luxuries (see below) and sometimes tech. My "impression" is that AI civs might interpret gold as tribute and that I am dealing from weakness. Probably reading too much into the game.
- gift luxuries if needed: When I can, I try to gift a luxury to each civ that I want to stay friendly with. I figure that it keeps them polite or gracious plus it "handcuffs" them from goint to war.
- trade a lot: yep. I did/do that. Keep the cash buildup going. Keep the tech advance close to me so that I have things to trade for.
In the game under discussion, the Babylonians, prior to the attack, were friendly with 10+ turns of ROP to go and about 8 turns of a luxury gift to go. That's why I said that they double broke treaties.
I think that your formula, which does work well, might need an addition to the rep and attitude requirements. Don't show weakness. When you become #1 in score, etc. the AI wants an excuse to attack you. I gave it the excuse by jamming settlers into every little space between warring countries - did so much of it that I could not keep up with adequately defending them. The AI saw an opportunity for some cheap wins - unfortunately they were valueless wins because the towns were start-ups - and took it, despite rep and attitude.
Thus, Drakan, I totally agree with you but my observation of the technique is altered slightly because I think that I took it, deliberately, to an extreme.
Golden Bear
|
|
|  |
 |
|
lebensraum
|
|
quote: Originally posted by Drakan
- avoid razing many AI cities/starving AI citizens |
que??
starving ai citizens affects their attitude? first time i've heard about that. do you know how significant this factor is?
now how am i supposed to get any research done during wars?!! unfair! unfair!
|
|
|  |
 |
|
Dis
|
 |
Las Vegas, Nevada
Feb 2000 time: 21:37
|
|
quote: Originally posted by Golden Bear
For a lead in... I have tried to remain objective about AI cheating in CivIII. I have tended to believe that it was a figment of imagination built upon poorly documented instances of occasional long shot odds. I mean, other than the way that AI civs always seem to miraculously settle next to resources that won't appear for Eons.
I'm also indifferent to the way the game plays as all the AI civs vs. me. I'm OK with that.
However, I just saw the belief breaker.
Pangea, full land, huge. I have spotless rep. Really. Never broke a treaty. Everybody except the Zulu (I'm executing them) is polite or better with luxury incentives to boot.
Bingo, the Babs break double treaty and attack from ROP rape - sorry, but the "how to beat the AI on builder" does not really work, BTW.
OK, I'm cool with that. I am a couple of techs ahead of everybody. Despite a land grab following AI wars I have AT LEAST 1 infantry in every city while no civ has an offensive unit better than cavalry. Cav, of course is very good. But how good?
How about, 11 cities attacked with single Cav vs. single Inf, for, what would you guess as results? Nope, you are wrong. 11 cities taken. Each single (1 attacker) cav captured or destroyed a city protected by inf. I'm thinking, what, 6 attack vs. 10 (doubled for fortified?) is less than 1:3 chance of victory. I should have lost 3 cities, maybe 4.
Remember now, that cav will not attack a pile of arty with a single inf guarding it. Suddenly, boom, every attack goes good.
Calling the number dudes out there. Can you give me the odds of this one? Aw, I know that it is impossible. Shoot, I just lost 3 tanks and a cav trying to finish off a 1 square musketman in a city. I did not whine (then) but maybe a little moan now.
I really don't mind a little AI cheating. But cmon. Oh well, I guess that I'll have to go make them go cry for their mommies. I'm going to win anyway, but this much desperation by the code is embarrasing.
AFAIC this puts the "does the AI cheat in combat" question out of reach.
Golden Bear (with his eyes wider open now!) |
could you tell me if your military is was strong, average or weak compared to theirs at the time of attack?
I know no one takes my theory seriously. But I seriously do think you get better battle results when you have overwhelming force. If I start out with the same size military, I get terrible combat rolls. But if I approach a city with sufficient force, I nearly always win. this is why I never even attempt to take a city with a couple of units even if there is only 1 defender.
I know it sounds hokey. But I swear the more units I have, the better they do in combat.
|
|
|  |
 |
|
Dis
|
 |
Las Vegas, Nevada
Feb 2000 time: 21:37
|
|
quote: Originally posted by lebensraum
heyya gb,
i suspect the problem is in the 'randomising' algorithm that is used. basically, the rng doesn't (generate random numbers that is). it does a reasonable job of spitting out numbers that look random, but it is not genuinely random, and sometimes it gets itself stuck in one of these ruts.
much as i would like to blame the ai, firaxis, sid, soren, george bush and the sicko who introduced me to civ in the first place, it's really not their fault. the only way out of it will be when we get a true, hardware-based rng. simple enough, just measuring something like thermal noise across a resisitor would do it, but until someone takes the trouble to build it into the hardware, there's nothing to do but grin and umm, you know. |
how hard can it be to randomize something? I know it's not this hard. I've worked as a slot technician in Las Vegas. Slot machines can generate winning percentages of about 98%. Sure they vary (this is a given when large jackpots happen). But the machines can randomly select cards and still achieve a payout it's designed for (usually in the mid 90 percentiles). Are you saying Firaxis can't write code as good as IGT (international gaming technologies)
|
|
|  |
 |
|  |
All times are GMT. The time now is 05:37. Apolyton Time is 00:37. |
top of page
|
|
|
Forum Rules:
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts
|
HTML code is ON
vB code is ON
Smilies are ON
[IMG] code is ON
|
|
|
|
|
|