Combat system - statistical stuff from >20 games

[Lt. 'Killer' M.]

There`s been much whining here about the combat results and so I decided to play lots of 16 player tiy world games (lots of combat) and just mark down for every combat: the units, their status, terrai, other factors such as walls, HP before, after, result. I modified my units by tripling hitpoints so goofy results are a lot less likely than with the original game.

I used a simple system to calculate attack and defence that favors the AI. So I should get results that show the AI winn less often then expected. To give an example:

A veteran Swordman (A=9, Base HP 9) attacks an unfortified regular Spearman (D=6, Base HP=6) in plains across a river. For terrain defence bonus I now assume 3 defence points for the plains which is A LOT more then plains give. So the swordsman should win >50% since he`veteran and the Spearman isn`t. For the river I assume the Spearman is veteran, too. Now it should be 50:50.

Admittedly a very simple and rough calculation, but if the AI now fares better then 50:50 regularly I know that the number generator is buggy - you follow me this far?


Here`s what I found:

A) Somehow there`s a graphics glitch and town walls often aren`t displayed when the town`s on a hill. explains loads of "wonder defender" battles.

B) AI won 389 out of 528 battles that should be roughly 1:1 (see above). That`s WAY TOO HIGH!!!!

C) there`s a tendency for strange battles when a Wonder is at stake or whem the AI is close to being wiped out. Then they get "wonder defender" - or are cut down like grass. Sadly it comes out 1 in 2 for the human on average where it should be 1:1 so the whining was with reason.

D) 1:10 odds happened 3 times out of around 95 battles- that`s OK. but the damage done by the loser was 23 times incredibly high (like a trireme regular taking 5 HP from an elite battleship. lol)

E) if AI is battling AI ( I observed close to 200 battles) strange things happen, too. But less often by far.

 

[Lt. 'Killer' M.]

One thing I now did: I upped all attack values by 1/2/3 for A=3/A=6/A=9 units and so on. this will make the AI units stronger, too but I rarely am defensive with equal or inferior troops anyway.

Firaxis should change the combat system so that different units have different attack capabilities against different opponents. i.e. a Battleship can sink a Destroyer with a single hit, can only rarely (20% of the time? 10%?) attack diesel subs (when they are on the surface), cannot attack a nuclear sub, can "bombard" overflying airplanes (but Destroyers and Aegis cruisers would be way better at this). A Pikeman would be an excellent defender against mounted units (but not modern ones), but suck against Archers. And so on...


Whaatyouthink???

 

[LaRo]

There is a combat calculator:

http://www.civfanatics.com/civ3combatcalc.html

 

[homegrown]

This question of attack bonuses against specific types of units is certainly mentionable. A lot of players here know that pikes kill knights pretty well.

IMHO the answer is that Civ 3 is still a civilization 'building' game, not a war game. Both playing styles are accommodated but the marketing generally reads "Build an Empire" rather than "Destroy everyone else on the planet". :crazyeyes

 

[DarkwingGT]

Killer, I'm glad you got those results only in that it means I'm not crazy and AI really does get a combat advantage.

One thing to note, with out the tripled hitpoints, the results would be amplified because the human unit has less chances.

Weren't specific bonuses in Civ2 and did they not take them out? At least the pikemen versus mounted was.

Regardless of if it says "Build an Empire", I dare you to play a game of Civ3 on the same continent as other civs without being attacked. It's impossible as at least once during the game, an AI civ will attack. And if the human military can't defend against the AI, well geez, it's hard to "Build an Empire" when you have no cities. Military is a large part of this game and if it's flawed then it cuts directly into the civ building aspect.

 

[Spawnbob]

read my post in
http://forums.civfanatics.com/showthread.php?s=&threadid=14314

it is possible to have an empire on the same continent as the AI who are constantly in wars with each other but for you to remain totally peacful for the whole game. It's hard but i think it 's about inadvertantly pis***ing the AI off, and not just going back on treaties etc i think a lot of it is to do with, do they see you as threat,or weak, have you got something they want , would crushing the human be useful to them.
I think also a lot of the reason for a lot of human AI wars, is the way people play inc me, in the rush to expand that very expansion can possibly be seen by them as a threat which it is in a way, even though your intentions might be peaceful.

Also in that game it wasnt i did everything th AI ask of me, i just remained neutral but firm aganst there demands and i think that helped keep the peace for me

 

[OneInTen]

I dare you to play a game of Civ3 on the same continent as other civs without being attacked.

8 civs, one continent, standard sized map. I only had a single city the entire game (was playing for 20000 point culture win with one city, to see if it would work). Didn't even come close to being attacked, and won easily ...

 

[SkyShagger]

Yeah, I went for a culture victory, normal map, max number of civs, only had 5 cities, and i won easily, but I did come close to going to war and talked them out of it. I only had Babylonian 2/2/1 UU's in time of calvary. *Shakes head* Those Germans, always looking for trouble.

 

[Beam]

Killer,

Thanx for recording the stats and it is not as regular as one would expect.

Yet, what difficulty you playing on?

 

[Zachriel]

Originally posted by DarkwingGT
Killer, I'm glad you got those results only in that it means I'm not crazy and AI really does get a combat advantage.

I've played thousands of combats and the randomizer works the way I would expect, though you may argue with the odds.

Six thousands years of history with no war would be nice, but hardly a simulation of human civilization. North America has had just one major war in 200 years, which is considered pretty good by historical standards. Europe has had two huge wars just in this century.

 

[Dan Magaha FIRAXIS]

I'm firmly convinced that if we posted the source code and it showed conclusively that the AI doesn't get a combat bonus, there would still be talk of conspiracy and AI cheating.

Killer, how often did you save and reload the same game when performing combat calculations?


Dan

 

[Lt. 'Killer' M.]

Dan Magaha: Never!!!

It may be that the way the random numbers are applied to the combat results produce a strange effect that is enhenced by my style of play though....

Is there the possibility of draws that have to be rounded off to one side? That could be enough if in those cases it`always the AI that gets the win, even if it only happens 1% of the time since especially 50:50 combats produce "rows" for one side. So it might happen there more often.....

Yesterday I had a 29 HP Army in a size 7 City, Defence=9 attacked across a river by a regular (6HP) war elephant (Attack=12). He took it out, only taking two hits, then got a promotion for it..... Essentially an (admittedly demaged) army of three Knights was wasted by 1 Knight that had roughly a fith of the HP, dedspite defence bonuses, even though the Army should have won 12 out of 21 fights --> that would be around uuuuhm


let me put it this way: the army could go over 34 rounds, then be down to 1HP and the elephant would be dead. (28+6)... The army would only have to win 6 out of these.. only 6!!!!! That`s pretty low....

 

[s3d]

Originally posted by Dan Magaha FIRAXIS
I'm firmly convinced that if we posted the source code and it showed conclusively that the AI doesn't get a combat bonus, there would still be talk of conspiracy and AI cheating.

Killer, how often did you save and reload the same game when performing combat calculations?


Dan

So I dare you to post it. I'm sure three are some problems with code. Not that you intentionally did it, just some bugs. For example I have clear impression that there is a bug in factoring hitpoints into combat.

 

[Lt. 'Killer' M.]

Originally posted by Dan Magaha FIRAXIS
I'm firmly convinced that if we posted the source code and it showed conclusively that the AI doesn't get a combat bonus, there would still be talk of conspiracy and AI cheating.

Killer, how often did you save and reload the same game when performing combat calculations?


Dan

Dan: Just for the record: I never said "cheating AI" meaning Firaxis planned the game unfair. It might be a simple mistake after all. I remember random number generators from the Amiga - hardly ever did one work. Later on the PCs had trouble cause what gave "random" on one system often gave predictable numbers on others.....

 

[Zachriel]

Originally posted by Dan Magaha FIRAXIS
I'm firmly convinced that if we posted the source code and it showed conclusively that the AI doesn't get a combat bonus, there would still be talk of conspiracy and AI cheating.
Dan

The combat system works as advertised. Of course, every once in a while a tank commander still takes a wrong turn and ends up in an elephant pit, but the odds are within the parameters specified by the combat system.

Planning a proper attack can minimize any bad results with the randomizer. Here is the solution to the "Killer Phalanx" problem:
http://www.crowncity.net/civ3/killerphalanx.htm

 

[Shockwave]

*still remember the time two of his marines were killed by ONE Japanese archer each* (that's one archer killed a marine, and another archer killed the other marine)

Both marines were made with barracks and fully healed..... Damn bad luck

 

[DarkwingGT]

Sweet, Dan, I look forward to the source code :grin:
(Note: Obviously Firaxis would never do anything like post the source)

Seriously, I never said the AI cheats, I said it has the advantage in combat. Even if it's only that the random number generator always sides with them, well then that's still an advantage. People often complain about luck, but luck is an advantage like any other. I think Arcanum has it right, a random generator that somehows keeps pretty close to the mean. Yes, every once in a while there's a fluke result, but that keeps it to that, a fluke result.

As for the use of superior force, well, I shouldn't have to outnumber my enemies 15 to 1 using superior units just to defeat one guy. That's a little ridiculous.

And for the people who play an entire game without a war, you must be God or something because no matter what I do, no matter how strong I am, civs always attack me. I play on Warlord. Not Monarch, not Regent, Not Emperor or Diety. Warlord and yet the civs always gang up on me. At least if they had some tension and in fighting, but usually it's very little if at all.

EDIT: One of my fonder memories was having an Elite Warrior attack my Veteran Rifleman who fortified in a city of size 12. The warrior was cut down to 1 HP and then proceeded to make a miraculous comeback and get my Rifleman down to 1 HP. I did fortunately win that fight, but of course that meant he couldn't defend against the Knight who attacked next.

 

[LaRo]

DarkwingGT, maybe you should play low profile when you are not very powerful enough to win. So if someone want 100 gp. give it, it will allow you to stay in peace. I know it is not fun to stay low profile war is not the only way to win. And if maybe they will ask for your techs. Sell before they do, it will bring you lot of cash to increase your science to 100%.
One thing, if you don't have great library, make sur to sell tech to the one that have it otherwise he will get it for free.

 

[Zachriel]

Originally posted by DarkwingGT
As for the use of superior force, well, I shouldn't have to outnumber my enemies 15 to 1 using superior units just to defeat one guy.
. . .
EDIT: One of my fonder memories was having an Elite Warrior attack my Veteran Rifleman who fortified in a city of size 12. The warrior was cut down to 1 HP and then proceeded to make a miraculous comeback and get my Rifleman down to 1 HP. I did fortunately win that fight, but of course that meant he couldn't defend against the Knight who attacked next.

On the first issue, most experts believe a 10-1 advantage against a fortified position is advisable. Even then, you can lose if things break bad. Perhaps your men were fat with Christmas dinner and not prepared for a surprise attack.

One defender in a war zone might be considered too little, especially if things break bad. If you always fortify with one unit, then expect to lose a city once in a while.

People tend to remember the exceptions, not the hundreds of battles that had the expected result. For instance, we remember John Paul Jones because he did the "impossible."

 

[DarkwingGT]

My guys should've been prepared, the war had being raging for 100 years already.

10 to 1, right, sure. So that indicates that regardless of technological superiority, I still need a 10 to 1 ratio. That's crazy.

I remember the battles of exception, for example in one game as the Persians, I attacked an unfortified spearman in the open with an Immortal, and get this, I won! I was flabbergasted. Their spearman had been repelling all of my Immortals and suddenly I won one. I'll remember that for a long time.

And one defender in a war zone, well I garrison each of my cities with 4 of the top defensive units, but against odds of 30 Warriors who can bring down my Rifleman, well it's a little daunting. The only thing I haven't seen is Mech. Infantry being overwhelmed. Then again, the computer never has anything better than Cavalry by the time I get Mech. Infantry. Maybe you're right, I need 8 garrisoning units of the top class. Forgive me, what was I thinking.