As promised, a Civ III combat test map

[Dan Magaha FIRAXIS]

Basically what I've done is create a normal sized map, Regent level, Babylonians versus Persians. You are the Babs, the AI is the Persians. Terrain is all desert, no cities, rivers, or other terrain features on the map to influence combat results.

There are two rows X several pairs of columns of 10 units each. Each pair of columns is a matchup between specific units. You'll start in the lower-right hand corner of the map (sorry about that) and work to the left.

First column pair starts out one col of Bab warriors vs one col of Persian warriors. Then it moves to a col of Bab warriors vs a col of Persian spearmen. Then a col of Persian warriors vs a col of Bab spearmen. Then two columns of spearmen, one each Bab and Persians. You get the idea.

I've include the world famous "Spearman vs Tank" matchup, as well as Cavalry vs Tank, Spearman vs Knight, and some others.

All units are normal (no conscript, elite or vets). No stacks, armies, or other configurations are used.

NOTE that you MUST use v1.16f of Civ III to use this save correctly and you MUST be using the default civ3mod.bic. If you're using modified rules who knows what kind of combat results you will see.

Also, due to the way the game stores the rand seed in the savegame, I have no idea how different the results will be from one person to the next. I recommend trying different combats in different orders each time to try and mix it up a bit. Also you might want to try ending your turn and defending as opposed to attacking first.

Dan

 

[cephyn]

Ok, i did it just as the game gave it to me, attacked with everyone. Here's my results (anyone wanna volunteer to compile all these?? )

Warrior v. Warrior (A1 v. D1)
10W to 10L

Warrior v. Spearman (A1 v. D2)
3W to 17L

Spearman v. Spearman (A1 v. D2)
3W to 17L

Knight v. Knight (A4 v. D3)
11W to 9L

Knight v. Spearman (A4 v. D2)
12W to 7L

Spearman v. Knight (A1 v. D3)
5W to 15L

Tank v. Spearman (A16 v. D2)
19W to 1L

Spearman v. Tank (A1 v. D8)
0W to 20L

Tank v. Tank (A16 v. D8)
15W to 5L

Cavalry v. Tank (A6 v. D8)
8W to 12L

Tank v. Cavalry (A16 v. D3)
20W to 0L

Cavalry v. Cavalry (A6 v. D3)
13W to 7L

That was me attacking in all cases. I think I show some screwy results up there with the warrior v. spearman and spearman v. spearman. Don't have my calculator handy, but it looks screwy, maybe someone has that in their head for easy reference.

Hope to see more results, I'll do a defensive test later tonite hopefully.

I'd just like to reiterate that my weird results, in game, come up when attacking a capital of a rival civ. Capitals seem to repel attacks far better than other cities of comparable size -- is there a bonus i don't know about?

I also noticed what i consider to be a BUG. Even though animations are CLEARLY off, whenever a cavalry unit is attacked, it still performs its initial animation and sound...raising up a gun and the horse grunts.

 

[Thunderfall]

I think I show some screwy results up there with the warrior v. spearman and spearman v. spearman.

3 wins and 17 losses look pretty resonable to me. Both of them have attack power of 1 and attacked the spearman with Defense of 2. My results are not much different from yours:
--------------------

Warrior v. Warrior (A1 v. D1)
10 Wins to 10 Loess

Warrior v. Spearman (A1 v. D2)
4W to 16L

Spearman v. Spearman (A1 v. D2)
3W to 17L

Knight v. Knight (A4 v. D3)
10W to 10L

Knight v. Spearman (A4 v. D2)
13W to 7L

Spearman v. Knight (A1 v. D3)
4W to 16L

Tank v. Spearman (A16 v. D2)
20W to 0L

Spearman v. Tank (A1 v. D8)
0W to 20L

Tank v. Tank (A16 v. D8)
14W to 6L

Cavalry v. Tank (A6 v. D8)
10W to 10L

Tank v. Cavalry (A16 v. D3)
20W to 0L

Cavalry v. Cavalry (A6 v. D3)
15W to 5L

 

[cephyn]

I didn't edit my previous post for this on purpose. Wanted to keep these separate.

OK here's some defensive stats. Of course not as comprehensive as before, since the CPU doesn't want to cooperate by attacking everything 1 on 1. I smell a conspiracy.

anyway -- format is a WIN is the CPU successfully defeating the PLAYER. Attacking unit is listed first.

warrior v. warrior
5W to 9L

Knight v. Knight
11W to 7L

Knight v. Spearman
10W to 9L

Tank v. Spearman
19W to 0L

Tank v. Tank
15W to 7L

Tank v. Cavalry
20W to 0L

Cavalry v. Cavalry
18W to 4L

Cavalry v. Tank
3W to 7L

 

[Thunderfall]

cephyn, did you fortify the units?

 

[Astaroth]

My results were exactly the same as Cephyn's, except that
Cavalry vs Calvary was 14x6, not 15x5. I counted the escapes
from fast units as losses. Looks like the "random" numbers are
stored in the save games. Maybe Firaxis didnt want us to reload
when we lose?

 

[Dan Magaha FIRAXIS]

Unfortunately there's no way at present for me to do a map like this without using the savegame format which does save the random number seed, so I can't be sure what kind of variation in results we're going to get. This is why I created a lot of units and recommended that everyone try doing the combat in different sequences, so the results should be different.

And yes, the game was designed so that save + reload wouldn't be a very effective or efficient way to cheat, but right now there's no way around this for the purposes of this experiment, unless Gramphos' tool can hack the save game to randomize the seed.


Dan

 

[cassembler]

Perhaps I'm being optimistic, but how exactly did you place the units on the map? Is this an updated editor or creative .bic hacking?

 

[cephyn]

Originally posted by Thunderfall
cephyn, did you fortify the units?

good question.

The answer is no, i didn't -- i was going for straight numbers, no bonuses (subconsciously, hehe)

All I did was declare war on the Persians, then click on the next turn.

 

[Troyens]

OK, what happens in THESE situations?

(I understand that a veteran missing a hit point is, in fact, a regular. But that's not the point).


All are in grassland, nobody is fortified.


1. An elite knight attacks a regular tank (minus one hit point).

2. A veteran cavalry (minus one hit point) attacks a veteran hoplite.

3. A veteran longbowmen attacks a full strength veteran cavalry with no retreat route. (I've seen what happens - the rifle-firing cavalry is destroyed).

4. A veteran musketeer attacks a veteran spearman.

5. An elite samurai attacks a conscript tank.


There are other examples that sometimes give anomalous and non-historical results. Bottom Line: units from one Age should receive about a 25% combat bonus against units from an earlier Age. That would solve a lot of problems.

I've seen strange results: a privateer with an attack strength raised to '3' attacks a galley - the galley sinks the privateer w/o losing any hit poiints. I've also seen a galley attack a privateer; since the galley has rowers the privateers should be able to just sail away. But nope. Sunk.

And galleys firing cannon broadsides (!) in the animation is especially ludicrous.

But for those who want accuracy, I suggest we also change the following:

1. Only infantry, artillery, and leaders get airlift capability - not tanks, cavalry and even elephants,

2. Bombers MUST be able to sink warships.

3. War elephants should be 4.1, and Longbowmen 4.3 (and English specific).

4. Cruise missiles should have a range equal to any bomber.

There are many other examples. Point: the occasional bizarre combat result is not a major problem - there are other problems.

 

[Dan Magaha FIRAXIS]

Originally posted by cassembler
Perhaps I'm being optimistic, but how exactly did you place the units on the map? Is this an updated editor or creative .bic hacking?

I used a debug build of the game which has extremely limited functionality for spawning units (it's a hack put in by the programmers during development to assist in testing parts of the game).

It was brutally tedious, too, let me tell you :crazyeyes:

Dan

 

[cephyn]

Originally posted by Troyens


2. A veteran cavalry (minus one hit point) attacks a veteran hoplite.

3. A veteran longbowmen attacks a full strength veteran cavalry with no retreat route. (I've seen what happens - the rifle-firing cavalry is destroyed).


There are other examples that sometimes give anomalous and non-historical results. Bottom Line: units from one Age should receive about a 25% combat bonus against units from an earlier Age. That would solve a lot of problems.


There are many other examples. Point: the occasional bizarre combat result is not a major problem - there are other problems.

RE: 2. I've seen this happen a lot, only Pikemen not Hoplites. Often, the Pikeman/Hoplite will win -- but that makes sense. Cavalry attack at 6. Hoplite/Pikemen defend at 3. Add in their Fortify bonus (cuz they usually have it) and any other bonuses (city walls, on a hill/mountain, across a river, high city population, etc) and its very quickly going to be an even match or even favoring the hoplite/pikeman. Especially since in your case here, the cavalry is hurt.

RE: 3. Well duh. Corner a cavalry unit, and its horses become useless, even a burden. (if you want to have a 'historical real life' reason for it losing. In game mechanics, the Cavalry has a defense of 3. The longbowman, attack of 4. Since a cornered cavalry isn't usually fortified, i can see how it would lose to a longbowman. You'd have to hope he's on a hill/mountain, in which case, the cavalry has the advantage.

RE: your different age bonus, I have to agree with this. I like that idea and im pretty sure ive posed it in some other thread, which brings me to your last point...

this thread is ABOUT odd combat results and trying to track them down. Don't try and change the subject, just open a new thread. 8)

 

[Cantankerous]

Troyens,

I agree with a lot of what you've said, but there's no way that longbowmen should have a defense of 3.

In historical battles of the middle ages, if the enemy ever broke through the lines and got to the longbowmen, they were screwed. They basically had no defense at close range. So I think a defense of one is the only alternative.

Cantankerous

 

[Hurricane]

Hey, keep to the subject! There are hundreds of threads about combat strengths and historical accuracy.

What this thread about is to test whether the random number generator generates strange numbers or not. IMHO, the test results show nothing unspectacular, except for the lone killer spearman in cephyn´s first test. I can live with this.

As Dan and other point out, the main problem with this test is that it uses only a limited number - and most important - identical random numbers each time it is run. This way you can´t say the good random spread found here is a precise picture of the actual random numbers used throughout one whole game of Civ3.

However, this fact can be used for good, as well. First, everybody (I´m thinking Dillo and other here) who suspect that their computer in some strange way messes the PRNG up can use the savegame. Since everybody should get identical results, this is easy to prove once and for all.

Second, the savegame can possibly be used to finally decide if defenders always get a 10% bonus to their strength or not, by changing the order of the attacks.

 

[damunzy]

Moderator Action: Troyens, keep to the top of the thread and quit trying to push what you think needs to be changed. This is a moderator warning so don't ignore it.
Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889 -PaleHorse76-

Moderator Action: Dear Troyens, I will type this slowly to help you: Threadjacking bad, on topic good, posters not do bad threadjacking thing, or maybe bad thing will happen to poster.
Lefty Scaevola
Please read the forum rules: http://forums.civfanatics.com/showthread.php?t=422889

Thank you Dan for posting this map. I have not yet tried it out since I have taken a vow of no games and no CFC until I find a job (what am I doing here? ....well, baby steps!) but I am glad that you guys are supporting the community. Thanks.

Glad you got use out of Trillian, Dan. Since version .7x has come out it has been the best IM IMHO.
Talking about sigs, thanks for the link to the Real Ultimate Power site. Ninjas rule!

 

[Dan Magaha FIRAXIS]

PaleHorse,

We'll see what kind of variance we get out of this map, and if the consensus is that it's useful perhaps we can do a naval unit combat test as well.

P.S., thanks to your sig I'm a bonafied Trillian evangelist. Great program!


Dan

 

[Loopy]

Say Dan, can you give some insight into how long many combat rounds are "saved" by saving the seed in a game? If it's for say 10 rounds, then wouldn't having a modern armor taking out an elite army of warriors or two as the first "battle", make all subsequent battles random? If armies are hard to do in your debug version, you could always do it as three squares of warriors on railroads for the armor to roll over. Or if you really wanted to, you could send me a copy of the debug version to play with. Thanks for the work regardless.

And for someone actually compiling all the results, if everyone posted their results at the end of their post as a nice 12 by 2 listing (with zeros) it'd be pretty easy to get everyone's in. For example Thunderfall's would be:

WW LL
10 10
04 16
03 17
10 10
13 07
04 16
20 00
00 20
14 06
10 10
20 00
15 05

 

[cephyn]

Once again, thanks Dan. You are continuing my faith that you Firaxis guys are a-otay folk.

This whole saving the random number seed in the game is a bit of a problem. 8/ Riddle me this, though, I thought the save/reload reset of the RNG could be accomplished by exiting the game entirely, and then going back in -- therefore, to get a new seed, shouldn't all we have to do is load up the sav game, sav it AGAIN, bounce out to windows and then back in, and load up the re-saved game? If this doesn't work, why am I wrong/confused/babbling?

As for naval battles, I'd like to see that too. I lost a battleship to a frigate last nite. 8( But maybe I'm just here to balance the universe with my very unlikely events.

Wanted to mention that this little combat tester is giving me insight into how the AI thinks -- for example, it will almost never attack a unit it feels it is most likely to lose to. Maybe this is why we feel we lose to the AI a lot -- it simply picks battles better?

Loopy, your format is cool, but uh, what is your proposal for standardizing which line corresponds to which battle set?

And yeah, though not related to palehorse, I too am a trillian user.

 

[bloodysmurf]

Dan, thanks for posting the test map!

I just wanted to take a moments to state what may be obvious to many, but perhaps not all of you: you are going to need a much larger sample of battle results to come up with a statistically significant argument that the number generator is anything but random. Who knows, maybe the thing really is skewed... but I think the majority of people complaining about bizarre combat results are reading patterns into otherwise random results (losing a tank to a spearman is going to be far more memorable than the ten times you got the spearmen stuck in your tank treads). I happen to agree with the arguments that these between-era battles should be modified, but I suspect that given the current rule structure, the outcomes are on average exactly what you'd expect.

In summary: Just because you flip a coin and it turns up heads nine times in a row doesn't change the fact that your next flip has a 50/50 chance of landing heads or tails... unless, of course, the coin is weighted

-ollie-

 

[Loopy]

Actually, you could draw statistically significant results from this sample size depending upon what question you're looking to answer. If you're trying to find out if combat is biased in Civ3, you have sample of 240 battles from which to draw your conclusion. This is a much better design than only 240 spearmen vs tanks. When you combine results from multiple users, the power of the test goes up. Of better design would be to randomly select participants for 12 battles from the entire population, but Dan's a programmer, not a statistician.