AI cheating in combat?

[erhsan]

Playing in Monarch mode...

My veteran archers vs. regular warriors is 4:1 ratio on equivalent terrain. No defense bonuses.


It takes 5 units like 3 warriors and 2 archers to take a city defended by 1 spearman. No city wall...

What's going on? This bias is just too much.

[jimb0v2]

there is no bias at all. You just had bad luck. Also, there is a seeding system in the way the save game works so that if you reload the game and just try again you will get the exact same results.

[PhillipII]

Ive noticed some very odd things about combat too.
firstly i have had determined that combat results appear to be calculated before the actual combat begins.
Save a game right before your units A and B attack enemy units X and Y, when X and Y are in two seperate squares.
make A attack X and B attack Y.
you will get result 1, for example A gets a perfect win while B is defeated having done 2 damage to Y.
now reload the game as many times as you want and refight the combats.
You will find that you will always get the same result unless you switch targets. very odd.

Ive also noticed that Babylonian Bowmen seem to be ALOT more powerful than thier profile suggests, being able to survive Immortal (A4)charges and take out Pikemen fortified on forest squares regularly.

[PhillipII]

dammit Jimbo v2! I hate it when people beat me to the post AND do it more succinctly than i did!

[jimb0v2]

Originally posted by PhillipII
Ive noticed some very odd things about combat too.
firstly i have had determined that combat results appear to be calculated before the actual combat begins.
Save a game right before your units A and B attack enemy units X and Y, when X and Y are in two seperate squares.
make A attack X and B attack Y.
you will get result 1, for example A gets a perfect win while B is defeated having done 2 damage to Y.
now reload the game as many times as you want and refight the combats.
You will find that you will always get the same result unless you switch targets. very odd.

Ive also noticed that Babylonian Bowmen seem to be ALOT more powerful than thier profile suggests, being able to survive Immortal (A4)charges and take out Pikemen fortified on forest squares regularly.

You actually raise an interesting point with your illustration.
If you alter your example slightly and X is a veteran unit and Y is an elite unit. The theory holds sort of. if you attacked with X and it died then attacked with Y and it had 2 hp left after combat then reloaded it and reveresd who attacked first (does'nt have to be the same defending unit, but must be same hp/defense) the Y unit might not die (but it will DEFINITELY have 4 hp or be daed) and the X unity will have 2 hps left.

This has a lot of wierd implications. I'm sure the goal of the seeding process was to discourage people from reloading until they win. This seeding process does discourage is, but doesn't fix it. Since you can either a) wait a turn and try again b) make sure your first attack is with a wimpy unit that would have died anyway or a stronger unit that might survive the carnage you know will ensue.

Other threads, with posters much more knowledgeble than I have suggested the problem with hp vs power/defense. It seems hps massively outweigh the power of a unit. Which is why you see late game horsemen still damaging and destroying infantry and such. And also accounts for why bowmen seem to do very well vs 3/4 defensive units. The odds just arent' that bad. I'm sure someone could come up with a system that simulates civ 3 combat in dice, a 2 power unit vs a 3 defense unit is a much different battel than a 1 power unit vs a 2 defense power unit. Overall a lot more hp is taken of the units in the 2 vs 3 than the 1 vs 2 battle. As for the defense bonuses and such, they only effect the defense number I believe.

Another observation that could be COMPLETELY WRONG related to that is that I believe how much damage you do to an enemey unit is independent of how much damage it does to you. If you had 20 elite horsemen vs 5 mechancial infatry common sense tells us all the infantry should survive, but this is just simply not the case. the defense value does nothing more than put a clock on the horsemen. Ie if the horsemen gets to swing 5 times because the infantry probably is not going to miss with such a high defense. I guess what I'm getting at is I'm not sure whether the 2 is really pitted VS the high defense of the infantry, or the 2 is just a general chance to hit vs ANY unit. It seems like the latter to me, in which case, an army of elite ancient horsemen is actually incredibly powerful. Which seems to be supported by the majority of actual players illustrations.

Sorry for all the rambling.

[PhillipII]

if you switch the targets of your units then you will get a totally different result, which also will be identical over multiple replays.

You are right though about the HP being a major factor which, unlike in Civ II, really makes it essential to build barracks right from the start even if you are only going to fight defensively.

Im am certain that Defence has to do something. i was once in a bad situation with Samurai attacking Spearmen and killing quite a few of them. Then i upgraded them all to Musketeers and the Samurai were no longer able to put a dent in them.
So, as i said, i dont know how, but defense really does work.

Actually i just remebered something i read in the manual. Defence is not the units ability to avoid taking damage but the chance that an attacker will take damage. So the higher the defence the faster attackers will die, thus doing less damage.
So if your horseman A2 attacks Mech inf. D? then the chance of the horseman receiving damage si much greater than the chance of the Mech Inf receiving damage.

[jimb0v2]

if what you say is true, it really does mean an army of horseys are better than a small number of more strategic units. (perhaps equivaltent in shields)

This is becuse a 2 or 3 power unit damages a good ammount of the time, it's just that they only get 5 hits because that infantry is going to hit the horse every time.

In my opinion this is a design flaw of sorts.

[nonamefits]

The % chance to hit

It is realy quite simpel to calulate the % chance you have to hit. And it explaines quite good why a spearman can beat a tank.

warrior versus spearman
1 vs 2 ,spearman is inside a 7 city. (givin 50% bonus)
this gives a 1 vs 3 chance.

the warriors have 25% chance each time to take a life from a spearman.

a 20 attack rating tank against a non fortified spearman:
20 vs 2 or 10 vs 1. This gives the spearman a whooping 9%! chance to wound the tank.

edit : the % chance is calulated as follows: your units rating : u
enemies rating : e
%chance = (u / (e + u) ) * 100

If on the other hand you average out the number of units required to take another unit u get the following :

3 warriors to take 1 fortified spearman.
10 spearmen to take 1 tank.

The effect of this system is that its the more the total number of life on your side thats gona win you the battle rather than the total number of attack/defense point. Another thing one can see from this is that if the % change in your attack/defense rating is big (say 1 to 2) will give a big diffrence while a 20 to 24 will almost not matter at all.

So the best way to build your army is to build the the biggest units u can, that still takes no more than 1 turn to complete. (oh, and get barracks!!)

[sophist]

Originally posted by jimb0v2
there is no bias at all. You just had bad luck. Also, there is a seeding system in the way the save game works so that if you reload the game and just try again you will get the exact same results.

Regarding the seeding. If you reload from the previous turn, the result is exactly the same. However, if you reload from the turn before that, the result will most likely be different. It appears as though they randomly calculate the seed for a turn in the previous turn; if you go back two turns, the seed for the "present" will be different. This is an interesting tradeoff, as in general, players wouldn't have the patience to go back two turns to get a desired result (and retrying every time it didn't work). However, I could see someone in a tournament doing this. So what it does is interfere with your ordinary at-home play, but makes it possible to, well, not cheat, but play with a stacked deck when participating in such a contest just as making it completely random would. Of course, calculating the seed further back is also a bad idea because then you would just avoid the goodie huts that are actually baddie huts, etc., as you could predict everything and that would alter your decisions.

[nonamefits]

Regarding the seeding.

As far as i see it a fairly standard random calulations thats beeing done here. They seed the "random()" funcition with a certain number and just saves the status of the random generator with each save. This means that if take 3 actions in a diffrent order will give a diffrent result. So you wouldnt need to go back 1 turn to get a diffrent result, but rater do somthing else that invoces the random number generator.

[barefootbadass]

The seed stuff is not a mystery, it is loaded when you load up civ III, and the next 'random' number always is related to the value of the seed and whatever modifications have occured to it each time a number is generated. If you reload to do a combat, the state of the random function will be the same so you get the same results. If you go back a turn, things will probably be done in a different order which will change the results. Also note, that the combat is indistinguishable from a random sequence if you play legit and don't reload. The seed is probably set to the time when you load civ3 and even if you leave it running for hours of play and if the sequence repeats(which probably will never happen if it is a good algorithm), the odds are essentially 0 that you would even have 1 identical combat.

Anyway, yes tons of horsemen can take down any defender if you have enough of them, but in all cases, units with higher attack will have on average less casualties. Only 1 unit hits each round, either the defender or the attacker based on their defensive or attack ratings, respectively, so its not true that an elite horseman attacks he gets exactly 5 shots. He gets at least 4 shots(he will run at 1 hp if he can). Its a cheesy strategy(although not cheating IMO) that can work on the ai but in MP people will deal with it or counter with like strategies. You will likely be mutilated by similar sized hoards of better units.

[barefootbadass]

Re: AI cheating in combat?

Originally posted by erhsan

It takes 5 units like 3 warriors and 2 archers to take a city defended by 1 spearman. No city wall...

What's going on? This bias is just too much.

Umm, is the city size 7+? that's a 50% bonus there. also the spearman is fortified +25% I think, and the terrain has +10%(all terrain has at least this bonus). Are you attacking across a river? That has a pretty hefty defense bonus. Anyway, also are you sure there is only 1 spearman? You will only see 1 when you right click on the city, but if you damage it in one battle the next one will come to the top and that can be easy to miss.

[vmxa1]

If you play enough games you can write a book on the "strange" combat results. Some may even be reasonable, if you look closely, but it does get to be annoying to see regular caravels beat your ironclad vets. I was losing elite galleys to reg galleys when they attack, unless I was able to fortify them before they stuck. I had an elite swordmen attack my army of elite calvs who were in the city with walls, barracks and forted and they dam near got killed. You can live with it, but I still feel it is too common.

[fanatic civver]

An issue that nobody has raised yet - remember that back in civ 1/2, barracks gave a +50% addition to a unit's attack and defense values. Civ 3 does NOT do this. This could be why results seem more random in civ 3 compared to previous civs. It is also why tanks get a huge increase in their attack value.

Units are a lot easier to build in civ 3 because there are less city improvements to build vs civ 2, so the fact that you can build more units sort of balances against the increased randomness.

With a large army of tanks I'm having no problems killing infantry fortified inside 6-12 sized cities with next to no losses because of the retreat rules. When everyone gets infantry then you're forced to use tanks, otherwise you're going to have to use the artillery + cavalry combination to take cities.

[teknokrat]

Sometimes I get the feeling the combat system is broken. What I have been seeing far too often is this. I attack a unit (say a fortified rifleman) with lots of riflemen (say ten) and they will all be killed even if they are vetran or elite WITH THE RIFLEMAN NOT LOSING A SINGLE HITPOINT! (I have even seen a rifleman go from conscript to elite this way). I wait a turn and then attack and the battle is a lot more "correct" that is, I lose a few units but manage to wipe out the rifleman with both sides taking damage.

The problem I see is not so much spearmen taking out armour ( I can live with that) but what appear to be "runs" when a single unit always "wins". Personally I think the combat engine has some kind of bug in the way it calculates probabilites.

later