Ok, what is the deal with combat resolution

[Clutch-3]

Ulthien:

I'm not going to tell everyone on this list my resume, but I will say this:

I never had any problem with people's "feelings" about the game. In fact, I've specifically said a few times that I'd rather not even discuss feelings. If people feel that the combat sucks, they are entitled to those feelings. When people start throwing around math that's seems invalid to me, and using it to prove a point, then I feel I should say something if I disagree. And instead of attacking others personally, as you have, I have offered to help analyze these results with other players and try to figure out some answers (see post w/ my name in it in strategy forum). If you have so much experience and want to contribute to this discussion, I suggest you do the same so we can compare results.

Of course if you'd rather just complain about how horrible the game is and then flame anyone who tries to systematically attack the problem, then keep at it. And don't try to say that the "math" I use is not applicable. Not only is it applicable, it is the only real way to prove/disprove claims about bad random number generators or statistically unlikely combat results.

cheers,
clutch

 

[Loopy]

Originally posted by Akka
2) the random generator generate very low/high numbers in streaks.

Actually a truly random number generator will output "streaks" a lot more often than people expect. In the simplest example of flipping a coin many, many times you'd actually expect to see a streak of twenty plus heads in a row. Ask a human to write down a "random" order of coin flips and they'll never write down long streaks like that. Doing a Chi-Square on number of streaks is one way to test for truely random data.

I've said it before, but it's worth mentioning again and again. It's the oddities that stick out when we remember combat results. I also don't think most people do the math to factor in the effects of attacking across a river, when someone's fortified, or even the actual base A/D/H numbers. They just see ancient era Immortal and assume it shouldn't be able to take out their more modern rifleman, nevermind that the Immortal has a very nice attack rating.

The way the computer "cheats" in combat is that it has a firm grasp of what all the multipliers do to defense values. The AI programmer also realized how effective swarms can be. There's been a Firaxis chat transcript that said that testers felt that the AI had an unfair combat advantage initially and it wasn't until the programmers gave the player a combat bonus that the testers felt the playing field was level.

 

[freakboy]

Even if there are times in the game where you get a stupid streak of bad luck, if you're playing so that the loss of a few units will make such a big difference, you're in trouble.

I've been using a broad attack scheme - bombers, ships, artillery, mobile units and infantry. Even if my tanks get smoked by a bunch of warriors, it doesn't really matter because I'm bombing their country into dust, my navy is blockading their ports, and my artillery is nailing anyone that comes out of their cities.

My only comments would be things like:

A battleship should never lose to a frigate or ironclad. A batteship (even circa 1915) could fire round so much farther than an ironclad that the ironclad would do it's best to just run away.

I think that this is one element missing from naval warfare. A lot of naval engagements ended with one unit evading the other and getting away. The battleship would always win, but it has to be able to see the ironclad (there might be fog or a storm.)

Modern armor has lost engagements with calvary in modern times. Armored units are at their best on flat open terrain, and fare poorly in mountains, jungle or hedgerows (afganistan, vietnam, normandy). If tanks go into the jungle, a bunch of guys with machete's can hack them to bits when they get out of their tank to take a leak.

Freakboy
AKA
Toolboi
AKA
Freakboi

 

[Malys Faisent]

Ok I will bring up a weird point, which doesn't involve any strangeness from the Tank vs. Spearman aspect of the combat generator.

I always explore with warriors, and often pop a hut and out come 2~3 barbarian conscript warriors. Of the numerous times this has occured I see one of two results:

My unit takes 2 (total) health levels of damage and is elite at the end of the attacks by the 3 barbarians

or

My unit takes 1 (total) health level of damage and is veteran at the end of the battles.

I've never, ever had any other result from popping a barbarian hut that released conscript warriors.

My 1/1/1 3 hit point warriors that fight barbarians are the meanest sons-o-*****es ever to walk the earth, they decimate entire clans (outnumbered 3 to 1) repeatedly regardless of difficulty level or barbarian activity. And you know what, it is entirely unrealistic!

If anyone else has seen different, let me know...but unlike Civ 2 I am entirely unafraid of popping a hut now...at least in Civ 2 you had to worry that you might get your explorer killed or your nearest town captured.

 

[Iron Chef]

Malys Faisent: I too often have that same result--a regular warrior surrounded by three conscript barbarian warriors attacked by each and eventually ending up a vet (once or twice an elite). But I have had the situation where my guy was beat by a single conscript, so yes, it does happen. With what frequency I cannot say.

 

[jim1013]

well i had a very lucky combat in my latest regent level game. now i had barbarian levels on a fairly low setting... second i think.

Anyway, i had an uprising or something. Which was 3 stacks of around 16 or 17 barbarian horsemen!
ARRGHH. thats a lot.

Whats more one stack was near a fairly new city, which was defended by ONE regular spearman. (the city was size 2 or 3). I thought i was sure to loose the city, however my one regular spearman defeated 17 conscript horsemen. It was elite at the end of it.

Talk about good luck !! If only GL could be produced by battles with barbarians.



I've also noticed incidences where the game seems to decide that a particular unit IS going to win almost no matter what.. although perhaps this is all within probabilities

 

[FlubberDubs]

Originally posted by Loopy


Actually a truly random number generator will output "streaks" a lot more often than people expect. In the simplest example of flipping a coin many, many times you'd actually expect to see a streak of twenty plus heads in a row. Ask a human to write down a "random" order of coin flips and they'll never write down long streaks like that. Doing a Chi-Square on number of streaks is one way to test for truely random data.

Ok, I couldn't help myself, 20+ head flips in a row??? 2^20 > 1,000,000 so I don't think we would expect to see that very often.

 

[Loopy]

Originally posted by FlubberDubs
Ok, I couldn't help myself, 20+ head flips in a row??? 2^20 > 1,000,000 so I don't think we would expect to see that very often.

I didn't say that you'd see it very often, just that you'd see it. Besides, just so we're on the same page here, the "official" definition of "many, many" is 50 million give or take one million.

So I might have gone a little overboard with the 20+. Ten plus probably would have made a better example. Regardless, short of seeing the actual code, you'd need a large sample size to determine whether or not something like combat resolution was actually random. The anecdotal evidence you see here just doesn't cut it.

Oh yeah, there is a bonus when defending against barbarians at most levels. So those victories shouldn't really count when talking about combat resolution.