Oh come on!!!

[Flak]

Originally posted by Setsuna
Ah yes, Barbarians - the one part of combat that isn't messed up.

According to the <strike>Manual</strike> Strategy Guide you get a 400% bonus against Barbarians on Chieftain.

Oh well, at least SOMETHING works.

Yeah I'm not sure that has an effect though. I thought it did. I'd been able to send regular horsemen all over the place and win battles, get up to elite, and win even more battles. Never had a Great Leader though. I'm starting to think that's a myth.

Anyway, I'd just gotten up to Knights and was having some trouble with barbs out near my precious saltpeter resource. I figured this would be a good chance to see how a Knight fights. I was thinking that just the sight of an elite Knight would scare the barbarians into yellow. My first fight with a regular barbarian warrior and my Knight gets his ass handed to him. He wins, but is on his last red notch and is forced to retreat the field as the other barbarian horsemen and warriors give chase.

Sure this was probably bad luck. But it sure seems to me that such extradinary events should be atleast as rare as say... getting a Great Leader. Probably much rarer than that even.

 

[Valant2]

Poochface:
Xerces started out referencing the origional post, but his later posts take specific quotes from Terser's post.

Xeres:
Ah Xerses in the later posts you was speaking about Tersers outcome eh? Well lets run through his two of his more specific examples probability wise eh?

Battle #1
1 computer controlled standard greek warrior attacks 1 human controlled vetran horseman fortified on a mountain . The end result is his unit is killed and the computer only lost 1 hp (usit was standard and now is yellow...hence the 2hp).

So he lost 4 turns and won 1 turn in the battle.

Computer attack power: 1
Human defense power: 2.25 (fortified +25%, on mountain +100%)

Lets run the math shall we?
P(x) = (5 over 1 factorial) * (2.25/3.25)^1 * (1/3.25)^4
= 5 * (.692)^1 * (.308)^4 = 3.11%

Its not as improbable as chimera's example but the computer was quite lucky in the battle to get an outcome that would only happen in approximately 3 in 100 battles just like this one?

Battle#2
1 computer controlled standard greek warrior attacks 1 human controlled elite spearman which is fortified in a city that has walls. The end result was the spearman having red (1hp left) and a dead computer warrior.

So he lost 4 turns and won 3 in the battle.

Computer attack power: 1
Human defense power: 3.5 (fortified +25%, walls +50%)

Lets run the math shall we?
P(x) = (7 over 4 factorial) * (3.5/4.5)^3 * (1/4.5)^4
= 35 * (.778)^3 * (.222)^4 = 4.00%

Once again its not as improbable as chimera's example....but once again the computer is quite lucky on the outcome.


Long story short my opinion is that the computer, statistically speaking, is winning outcomes to battles in its favor in an improbable frequency. Hence one of the following must be true..

1) There is something else included on the per turn outcome calculations that Firaxis has not divulged.
2) There is something wrong with their equations that determines the outcome per turns of the battle.
3) The random number generator is not being random enough.
4) The AI is beding the rules....yes I know Firaxis says no...but still would like em to check.

 

[Xerxes314]

Behold! It is... the Civ-ulator.

http://www.columbia.edu/~sdc2002/civulator.html

Try it. You'll like it. And stop saying mean things about me. It hurts my feelings.

Xerxes

 

[Valant2]

Xeres:
Since when did I say mean things?

Checked out the calculator....hmm it gives only a 25% bonus to fortified units. I thought that was 50%? Anyone here have the 411 on this?

Would not take much more to add the probability of the final battle outcome probability on his site.

 

[Xerxes314]

That is the final outcome probability. Sorry if that's unclear.

Xerxes

 

[bdrago]

Hrm. If I click on the "Civ-ulate" button multiple times without changing the inputs, the percentages and HP remaining keep changing by a significant amount. Shouldn't the values stay the same?

 

[Valant2]

Xeres:
I take it your the one who made it? If so might I inqure the equations ya use to get the final battle results?

Also any info on fortified being 25 or 50%?


From SMAC Longetivity Vaccine Wonder...
HEY...Get off my land you peacekeeping son of a .....

 

[Xerxes314]

The numbers change slightly because the process is random, albeit averaged over a large number of samples (1000). I find that there's a variance of about a percent, which I don't consider to be significant. I could push it down by taking more samples, which would make it slower, or by finding an analytic solution, which would require thinking (noooo!).

The +25% bonus for fortifying units is in the civilopedia and also in the default ruleset as a modifiable parameter. All my numbers are from the default ruleset, so if those don't match the actual rules in the game, the calculator won't work. I think they do, though.

Xerxes

 

[Valant2]

Xeres:
Ok your taking a sample of a population approach. Ok sounds good. Is the calcuations being run on the server of the client? If the client is running them I would crank up the sample some more.

Okay the 25% changes my previous calculations...but it would be slight. Not enough to bother with recalculating them.

1) In your calculations are you allowing for redudant outcomes probability wise? For example if you flip a coin 8 times....the probability of getting 4 heads and 4 tails is not (.5)^8. You need to multiply this by the number of combinations that result in 4 heads and 4 tails...

i.e.
HHHHTTTT
HTHTHTHT
HHTHTTHT
THHTTHTH
and so on

Hence the use of the factorial of (8 over 4) = 8!/(4! * 4!) = 70 which is then multiplied to the (.5)^8 to get the correct probability of 4H and 4T in a 8 coin toss....which is 27.34%


1) And how are you compensating for a variable number of trials based on the attacker and defenders hp?

Think your trying to compensate for this by runing a sample. In your samples are you stopping after one of the 2 units hits zero hp?

From SMAC Longetivity Vaccine Wonder...
HEY...Get off my land you peacekeeping son of a .....

 

[Valant2]

Xeres:
Hmm noticing some other things with the calculator as well.
1) Where did you get info on that the number of city improvements increases the defensive bonus of a unit defending in a city?
2) Where did you get info on that the population of a city increases defensive value of a unit in a city for each point of population? From what I read in the instruction manual...units station in a city get the following bonuses:
Pop of 6 and under with no walls - no bonus (maybe it counts the terrain that the city is built on for this level)
Pop of under 6 with walls - 50%
Pop of city 7-12, reguardless of walls - 50%
Pop over 12, reguardless of walls - 100%
3) Where did you get info that the terrain the city is built on is added to a units defense?

 

[Xerxes314]

The calculator works by emulating a battle 1000 times. There's no coin-flipping sort of combinatorial analysis going on at all. If you're interested in the nitty-gritty, you can just view the source; it's all there in the javascript.

The information about terrain bonuses is in the civilopedia. The information about population and improvement bonuses is inferred from the ruleset. To check it, open Civ3edit and open a mod file. Then go to edit rules and check the section on general rules. It contains a number of defensive modifiers, such as fortifying, being in a fort, city walls, and something called "citizens" and another thing called "buildings" which both have values of 4. I presume this means that a citizen or an improvement is worth +4% defense.

Xerxes

 

[kensai]

Greek spearman vs. my veteran horsemen unit(fortified in mountains)=My loss, with spearman unit's health barely edging into yellow.

well, a spearman attacks with one and a horseman defends with one, so even with the combat bonuses to your horseman you can expect the spearman to win randomly once in a while

Greek warrior vs. My elite spearman unit fortified in city with walls=My win. However, spearman unit has a tiny sliver of red left on health bar. Same enemy turn a horsemen unit destroys said spearman with no apparent loss of health.

I dunno, but I think a warrior (attack of 1) and horseman (attack of 2) should be able to beat one fortified spearman.

At least two battles with horsemen vs. my veteran horsemen (and yes, one in which the Greeks did attack across a river. In both instances, my units were dispatched with virtually no health loss to the opposing units.

umm, a the Greek horseman has an attack of 2 and your defends at 1... why are you complaining that you lost these battles?

When you disect it, it doesn't seem like the computer was cheating at all.

 

[Valant2]

Xerses:

Okay found the values of population and city improvements....but I must still ask how you know that they are always plugged into the equation to determine a battle outcome for units defending in a city? For example how do you know that a unit defending in a city gets a terrain bonus + town type bonus + # pop bonus + # improvement bonus + river bonus (if applicable) + fortified bonus (if applicable) + great wall bonus (if applicable) in all cases?

Maybe the # of city improvements bonus and population bonus are only applied to a city of a pop of 6 and under with no walls.

Long story short...
For example you are given a chart of vales of X, Y, and Z values and told in condition A the result is X+Y+Z....how can you assume that condition B will result in X+Y+Z? It could be only X+Z or maybe just Z.

From SMAC Longetivity Vaccine Wonder...
HEY...Get off my land you peacekeeping son of a .....

 

[kensai]

Battle #1
1 computer controlled standard greek warrior attacks 1 human controlled vetran horseman fortified on a mountain . The end result is his unit is killed and the computer only lost 1 hp (usit was standard and now is yellow...hence the 2hp).

So he lost 4 turns and won 1 turn in the battle.

Computer attack power: 1
Human defense power: 2.5 (fortified +50%, on mountain +100%)

Lets run the math shall we?
P(x) = (5 over 1 factorial) * (2.5/3.5)^1 * (1/3.5)^4
= 5 * (.714)^1 * (.286)^4 = 2.38%

Since it's warrior vs. horseman, the horseman will always run away UNLESS the warrior is brought down to one health and the horseman is brought down to one, in which case it will stick around to fight. So this battle outcome must have been made up.

 

[Valant2]

Kensai:
Would the horsemen still flee even if they are fortified?

Also I was just calculating the probability someone elses claimed battle outcome. I was not the one who said it happend.

From SMAC Longetivity Vaccine Wonder...
HEY...Get off my land you peacekeeping son of a .....

 

[Cunobelin Of Hippo]

In my forty minutes of play (had to go to the neighbours 'cause my computers being a little b1tch) I noticed no combat cheating. All seemed fair, especially when my Jag Warriors started rocking American @$$ and allowing me to extort them for all they had. MWAHAHAHAHAHA!

Combat seems fine, but I haven't really played enough to offer a meaningful opinion.

 

[plasm4]

i consistently get by destroying whenever I try anything that looks anything like swarming.. when I attack I use attack units (like longbowmen), defense units like pikemen.. now if I see an enemy I attack with my longbow, when I win, I back him up with the pikeman. If my bowman gets attacked now the pikeman will woop on the attacker and keep the bowman safe. Also, I always move my hurt units back on my side and fortify so it can heal. I've noticed the computer uses the same strategy, retreating hurt units, and I try my best to prevent them from escaping..

this way slowly my army will get much bigger than the opponents, and more of your units will live long enough to upgrade them. I lot of my riflemen who are about to become infantry hopefully if i can get some rubber were once spearmen.. Using this strat you also build up more elite units which is nice.

One more strategy to make you're conquests a bit easier...
I was England allied with France and we were tearing Russia apart.. I was pissed cuz france didn't join the war til I did most of the work. I'll be damned if I'm gonna let france annex those russian cities that I've had my eye on all this time. so this is what I did, I emptied 2 pretty good cities in the center of my empire of any units and stationed 1 cavalry right outside of each. I called russia on the phone and asked for peace, offering them those 2 cities in the heart of my empire, in return I asked for 3 of their cities. Russia agreed. Next turn I put some defensives in my 3 new russian cities, and I retook the cities I gave to russia. All this without a fight.

Only problem now is that the russian cities are too far to be very productive, and my 2 cities have russian people in em so they ain't so productive anymore.. but since they're right next to my capitol they're almost back to speed about 15 turns later

I've experienced a TINY but of the unbalanced fights, but I've noticed it works bothways.. i remember my bowman defending against a knight and winning.. It hasn't really been a problem cuz I get my injured off the front lines as soon as possible, only thing to worry about though is that the comp will try to pick off your injured by horseback so you gotta escort em

 

[Xerxes314]

I changed my mind about the population/building defense value. My new theory is that it's a defense STRENGTH (not bonus) that's applied vs artillery attacks. Which is increased to 8 against artillery by a wall, increased to 8 against naval assault by a coastal fortress, and increased to 8 against air assault by SAM facilities. Also, coastal fortresses and SAMs make attacks vs ships/aircraft with an attack strength of 8.

Hope this helps.

Xerxes

 

[chimera99]

Whew! I really started something. From what I've seen, horsemen are very weak. I assumed that in the open, they should have significant advantages. So much for assuming. I now use them to finish off very weak units such as the enemy units that move a square away after almost being killed.

The fight I described was the opening battle in the war between myself and the Aztecs. They won that battle but lost the war and are now only found in the history books. After fighting probably 100+ individual battles in that war, I have a better feeling for the changes in battles. Civ II was quite predictable (very little varience). Civ III is more like the Panzer General series. With units that are anywhere close in strength, the varience is very large. I still wonder about the reports of swordsmen taking out tanks though. Maybe the programmers need to put in some floors/ceilings in the combat.

Given that, I have to say this is a fun game. It is VERY different from Civ II.

 

[Valant2]

Xeres:
Never crossed my mind those 4's could of been used for bombardment calculations....good call on that one.

Have you found any specific info in the instruction book, stradegy guide, comments from Firaxis on wherver we can confim or deny that defensive bonuses of the terrain where the city is build upon is included into the defense formula of a unit defending in a city?