Combat system - statistical stuff from >20 games

[Lt. 'Killer' M.]

Originally posted by demetrious
Back to the initial post -

If the battles were true 50/50 outcomes - the odds of 389/526 is 0.02% or 1/5000 that this is just bad luck and despite even odds the AI won that many times - i.e. you really must assume the randomization is infact broken.

The problem is that if the battles that were "approximately 50/50" were in actuality 60/40 (because of the assumptions) the odds that this is a reasonable result jumps to 12.5%. That is the same as flipping three straight heads on a coin. Unusual but not unheard of.

So the sample size is big but the upfront assumptions in the battles and which ones you selected as "approximately 1:1" I think raise some uncertainy.

Again: I made sure it was at least 50:50 for me . Just calculated some of the commoner battles - they`re closer to 60:40 for ME! (exact numbers depend on how rounding is handled).
So if it is bad luck, it`s somewhere up there in the 1/5000 area! But I`m the guy who just walk down the street and was hit by a brick falling from a house. no constrction site, just an ordinary house. Bad luck!

Two weeks later, the same happened again in a different town!

bad luck? Yes! That`s just me!


btw: there`s a guy who got struck by lightning 8 different times! BAAAAAAAADDDDDDD LUCK!
(but none of the hurt him seriously)

 

[Zachriel]

Originally posted by Killer


Again: I made sure it was at least 50:50 for me .

Could be incredibly bad luck. But assuming each stroke of bad luck is independent, and not due to you somehow dislodging the bricks, then cheer up! Your luck may change.

On the issue of Civ3 combat, I'm not sure that your complex scheme for testing is accurate. It would be better to go mano a mano, warrior to warrior, 1a-1d, on open terrain with no terrain advantage, I think.

 

[Zachriel]

Perhaps the "faulty randomizer" is due to a psychological effects.

When flipping a coin (or any randomizer), a streak of arbitrary length will always occur sooner or later. In other words, 100 heads in a row, or 100 tails in a row will eventually always happen -- if you flip the coin enough.

When we win combats with our tanks, we expect that. If we do it without a scratch, we may note it, but immediately our attention turns to the next target of our advance.

If we lose several combats in a row, especially if it stops our advance, perhaps even leaving us defenseless to a counterattack, we remember. We really remember.

Over time, these incidents continue to recur, reenforcing our perception. This is similar to the gambler with the sore foot, standing on one leg, winning the big bet. For the rest of eternity he will stand on one leg thinking somehow or other this will make his horse faster. He just "knows" it works for him.

 

[Lt. 'Killer' M.]

Zachriel: I know about strings of luck! I just knocked out 3 Hoplites (veteran) with 3 Horsemen (2 regular, 1 veteran)! Die, Alexander!

On the other hand yesterday 2 Elite Pikemen of mine, on hill behind walls fell like leaves when 2 conscript warriors and 1 horseman (all regular) went for them....

Still, even though especially 50:50s are prone for rows, it still sucks because the AI gets an enormous advantage since it produces so many units so fast.... I always wonder how they can pay for these armies????

But actually I think the problem may be too complex to show in 1A<-->1D combats. May be somthing how the rounding of numbers is handled.....

When i find time I`ll do some tests (but those I can think of that would help to show if there`s a bug in the combat calculation require an editor) and then we`ll see.....

Esentially I had a gut feeling and a rough estimate confirmed it. It`s not really more then that - it`s still a gut feleling, but I think the test shows that the combat thing need looking into!

 

[etj4Eagle]

Originally posted by Killer


But actually I think the problem may be too complex to show in 1A<-->1D combats. May be somthing how the rounding of numbers is handled.....

Question, why are we assuming that the numbers are rounded? Since this is a computer there is no reason for it not to keep a decimal place or twenty. (Other than the variable type, but you can still use integers and keep a decimal place by just multiplying everything by ten). I see no reason for a warrior on plains on defense not to have a defense value of 1.1 (instead of rounding that to 1 or 2). I noticed that the combat calculater assumes no rounding as well.

 

[Rhandom]

If you put a save game file up here for everyone to test, won't it have its random seed already generated and so everyone is going to see the same results?

 

[Marzipan]

If the observed results really do not match the expected distribution then I bet that they are not handling rounding/truncation properly when factoring all the different considerations in. I noticed that there are some numbers on the city screen that have what are either rounding or truncation errors. Wish they would fix it!

 

[damunzy]

Still no test battlefield from Dan? Must because of the weekend. I understand.
As to battles being fair, I have seen both sides of a coin. I have been extremely unlucky and I have just recently took out the germans (3200 BC) with my one spearman versus Germany's spearman and warrior (all regular with my spearman getting verteran status after beating enemy spearman, but still only had 2 health versus the warrior, as he lost 2 to the spearman). I am not saying that people are having weird results, hell, they could have something to do with the way different processors handle numbers (remember the first Pentiums? They had a math problem. Maybe Firaxis missed something when programming the game.). I have noticed extremely different results when attacking from different squares, and maybe it has to do with a hidden river, or maybe there are certain angles of attack that are sweet spots.

I am now tryingto figure out how Germany went from 1 warrior to 1 warrior and 1 spearman after I declared war on him and attacked the same turn. Now that seems fishy to me.....and I have the pictures to show the weirdness. They will be posted to www.geocities.com/palehorse76/weirdciv3.html when I get a chance.

 

[gonzo_for_civ]

Visit the civIII combat calculator to find out the chances of winning a certain battle. I find that usually i will win a battle with my vet. vs. enemy regular and only lose 2 hp from my vet. This is opposed to losing with a regular agains their regular(my unit taking 3 damage as opposed to only 2 as a veteran). This is strange considering that there is no offensive or defensive bonus given to veterans or elites other than hit point addition.

 

[Lt. 'Killer' M.]

Hi folks!
Used the weekend to play LOTS of games. At first it was the same as before - I lost more then my share on all levels, and also winning units often (but not always) took unbelievable damage. Like an Elite Knight (15HP) going for a fortified Conscript Warrior (6HP) in the open on grassland and surviving with 2HP.....

Then I had to install several other programmes - and things changed. I don`t think it had anything to do with the other programmes but it was another strange coincidence.

I played the Persians - my Immortals swept across the continent. 40 battles - no loss. OK, OK the other guys were using Horsemen at best but still I`d come to expect to lose about 1 unit / city taken.

And then: uh-oh, this guy has Knights! And Pikemen!..... I built some Horsemen (best I could do) - and swept him away with NO losses!

this really shouldn`t be! I guess the AI "felt" as i normally did!

Different game, same story! My Babylonian Bowmen killed everything in sight. I lost a few, then the rest were elite and then I won a domination victory using extremely few Knights, backed up by the Bowmen. 9945 points, Deity level - not bad, hu? Actually the combat results in that game checked out at roughly 55:45 for me - I believe in using superior power but what bothers me is that the 5 % extra were fights that really counted. Like a single Elite horsman fighting down a veteran Impi with only 1 of 15 HP lost, then take the Horsman counterattack and tell Rome where to stuff it (again only 2 HO lost), then go for the next Impi and kill it with another 3 HP lost - and get a leader along with the Zulu`s last town! shouldn`t be!

Strange results may happen - but they also should be randomly distributed, and these were al for me, as for many games before, they`d all been against me.

What I`m saying: the random number generator sucks !!!



Another thing: it seems that the Babylonians get some advantage! I never see them killed off by other civs, I hardly see them loosing battles, I hardly see them loosing cities. And I always have a hell of a bad time killing there Spearmen with my Cavalery! Usually it takes 2-3 Cav / Spearman!

Don`t flame me for this, please, this is not some late night hallucination; I can use that to my adcantage by picking the Babs for me; it`s not AI cheating but rather a bug (or did i miss something in the manual about the Babs being extra good fighters???).


Dan: what do you say?? i hope you got my e-mail.

 

[damunzy]

Killer, I have noticed in most of my games that my Samurai do awesome, not even needing catapults, while just a few turns before my swordsmen using enough catapults to damage the enemy to 1 hp were taking enough losses for me to notice. Now these battles weren't versus the same cities or even the same civs and they weren't conducted in a scientific matter. I just know that my samurai were kicking some booty versus the 2 defenders when my swordsmen where having problems. Now my samurai were still kicking tail when they ran into pikemen(3 defense) but they retreated a great deal also and in a few rare cases, died.

In case you have never played the Japanese, the stats for a Samurai are 4/4/2 and require only iron. Unfortunately you cannot upgrade them and you cannot upgrade into them (So your horsemen are stuck being horsemen until you get calvary as you don't get knights. I am changing my .bic to take this into account so the horsemen will be able to change into samurai.).

 

[cephyn]

Originally posted by PaleHorse76


In case you have never played the Japanese, the stats for a Samurai are 4/4/2 and require only iron. Unfortunately you cannot upgrade them and you cannot upgrade into them (So your horsemen are stuck being horsemen until you get calvary as you don't get knights. I am changing my .bic to take this into account so the horsemen will be able to change into samurai.).

Samurai are cool. You cannot upgrade TO them (just as with any other UU) but you can upgrade FROM them (just as with any other UU) -- They replace knights, so just as knights, they upgrade to Cavalry.

However, there is a disadvantage to upgrading -- they become worse defensive units. 4/4/2 to 6/3/2.

Ouch.

 

[Marzipan]

I'm wondering if I'm seeing this in the game I have going now. The German navy is unbelievable! They have won all but one or two sea battles out of maybe 20, when it should have been a little closer. In their last victory, a battleship with one hit point left attacked and sunk one of my undamaged veteran subs. Hummmm. . . and then there was the transport that sunk my destroyer.

 

[Ironikinit]

Killer, I hope you're keeping in mind that the defender (from what's been said on Civ related boards, anyway) always gets a bonus. It sounds like you're attacking a lot, maybe that has something to do with it.

 

[Lt. 'Killer' M.]

kittenofchaos: as I said, a very rough test os I gave them a bigger bonus then they actually get. but I couldn`t find what bonus rivers give and if it`s cumulative... so there`s a certain degree of error in it but it shouldn`t really come out the way it did.

And those later games show the opposite: I keep winning (A and D) against insurmountible odds..... Good for me, but not fun since I don`t have to be careful about overextending my troops - they`ll soak up counterattacks and just go on, exactly as I always said the AI did.....

 

[Akka]

Originally posted by Killer
kittenofchaos: as I said, a very rough test os I gave them a bigger bonus then they actually get. but I couldn`t find what bonus rivers give and if it`s cumulative...

The bonus is : +25% defence for the unit attacked, if the attacking unit have to cross the river to reach it.

 

[Lt. 'Killer' M.]

thanx, Akka

that and the non-displayed walls in cities that have shrunk below 7 may explain a lot!

I just played a long game, huge map, one me and another civ. he had no iron, no horses and i swept him away with knights. worked as it should, but:

why did his longbowmen (essentially Knights with 1 movement and weaker defence) always take out my pikemen (15 times, sometimes even w/o damage), but often a spearman in the same status (level, damage and bonusses) damaged my knights to retreat???? Strang that the small sample showed not even one probable exception, but the large one many rarer ones!!!!!

 

[DarkwingGT]

Killer, I had assumed a while back that any distance weapons get to fire first in the round. At least that's what my battle results support. Is this not true? All my battles lead me to believe that ranged weapons (not bombard, but ranged weapons) seem to get to do their damage first.

BTW, by ALT-TAB friendliness I meant that I can keep IM open while playing. Kills the dead time between turns.

 

[Lt. 'Killer' M.]

Originally posted by DarkwingGT
Killer, I had assumed a while back that any distance weapons get to fire first in the round. At least that's what my battle results support. Is this not true? All my battles lead me to believe that ranged weapons (not bombard, but ranged weapons) seem to get to do their damage first.

might be the case - I don`t know.

as for ALT-TAB: when do you think I work?????

 

[Lord Thorium]

Whether the AI gets an unfair advantage or not, one thing is clear to me. The game should not rely on randomness so much. Doesn't it just make you want to delete the game when you decide to attack a weaker civ only to see your tanks slaughtered by their cavalry because you are "unlucky" this time around?