Combat - Am I Missing Something?

[Zapaan]

At first I thought maybe I just wasn't paying enough attention to things like terrain, or perhaps I misunderstood how terrain came into play, or maybe even that random chance could produce a rather ridiculous result. But I'm now convinced that the AI is flat out cheating.

Someone explain to me how an Elite Battleship at full health can EVER lose to an Veteran Ironclad? It's 18 to 4 for chrissakes! More than 4 to 1 odds! Even forgetting the fact that the Battleship is Elite and the Ironclad is Veteran, forgetting that there's no terrain modifiers involved being both ocean squares, how is this even conceivably possible?

But even better than this, is a Veteran Pikeman defeating an Elite Tank. HOW?! I don't care if the Pikeman is on Mountainous terrain, in a fortress and the tank is attacking from plains. It simply should not be possible under any circumstances!

The only possible conclusion that I can come to (especially considering I'm only playing on Chieftan difficulty!) is that the AI is bold-faced cheating like a crack whore on speedballs.

So am I missing something here? And for those of us with only a cursory knowledge of what can affect the outcome of combat, would someone outline the key factors in determining odds for combat?

I'm ][ this close to losing it and unleashing my nuclear arsenal on the world, but considering how shamelessly the computer is cheating, it wouldn't surprise me if they all fell out of the sky in transit and landed on all my cities

- Z

[Gramphos]

It's the randomness that makes it possible for some strange attack results. The AI ain't cheating in combat. The designers wanted to bring back the randomness of Civ1 when making Civ3...

[Zapaan]

I can see random results as being a factor to consider, but how does one explain results like out of 8 combats with the above Battleship vs Ironclad scenario, 5 times I lost?

That's ridiculous! What is the point in even having stats like attack/defend numerics then? At the very least the random factor is WAY too large an effect. I can see having 1 occurance in the above 8 combats, but even then, that's far more than should ever be possible, even forgetting the logic behind something like an Ironclad vs a Battleship confrontation.

Is there ANY way to adjust this idiotic behaviour? I'm a warmonger by trade, and this is completely killing my enjoyment of the game.

An ideal randomness factor would allow the randomness to affect combat on a bell curve of technology. Essentially, the closer in line the two units are tech-wise, the higher the random factor. The further disparate the tech levels are of the two units, the lower the randomness factor approaching zero for 2 or higher tech levels of seperation. It's ludicrous to think that a lowly warrior could ever hope to defeat a Mobile Infantry unit, under any circumstances, but it's happened to me, and not only in a single isolated incident either.

- Z (One cheesed off warmonger)

[Gramphos]

Yuo may increase the hitpoints for the levels, and even increase it for some units on era basies. The more hitpoints the lesser the random effect will be. I'm not an expert on the exact calculation, but you migt look at combat calculators.

[Zapaan]

Hmmm...that sounds promising. How would I go about editing unit health? In the editor? I only see a place to affect bonusl HP by +/- 20. Is there a way to adjust it in general? As in for all games? And I assume that only affects scenarios, rather than the core game mechanics.

Perhaps there are some mods out there that adjust this in some way?

And thanks for the help!

- Z

[Solomwi]

I believe the calculation is fairly simple, at least the way I understand it:

A/(A+DM)=L, where A is attack value, DM is defense value after modifiers and L is the RNG "line". A random number is generated and a hitpoint subtracted based on which side of the line it falls on. It would also work as DM/(A+DM), just backwards.

For instance, if your swordsman is attacking an unfortified spearman on flat ground (10%), it would be: 3/(3+2.2)=.577 (although I don't know the level of precision the game uses). Then random numbers between 0 and 1 are generated. In this case, if it falls below .577, the spearman loses a hitpoint, and vice versa if it falls above .577.

In your tank example, there's a 175% cumulative bonus for the pikeman, making his defense 8.25, so your "line" is 16/24.25, or .660, giving your tank a roughly 2 out of 3 chance of winning each individual hitpoint battle. It should, holding true to the statistics, lose 1 or 2 hp, but win the battle. However, it's not that far-fetched to imagine the pikeman getting three or four lucky rolls, especially if you've ever played roulette and bet the columns.

Also, concerning the number of "lucky rolls", don't forget that when dealing with random events, past outcomes have no bearing on future odds. IOW, most people are tempted to scream "what are the odds of that 1 in 3 chance coming up AGAIN?!?" if, say, your tank has lost a couple of hp in a row. The answer is 1/3, just like it was last time and just like it will be next time, regardless of whether or not it comes up this time.

Sounds like you've just hit a run of really bad luck.

As for the mods, what you mentioned is the right place to look. You may want to add one or two starting hp per age past ancient, based on when the unit becomes available. The "Bonus" part is a bit misleading, bump that up one for say Cavalry and regular Cavs are produced with 4 hp, vets with 5, etc.

[Zapaan]

Originally posted by Solomwi
I believe the calculation is fairly simple, at least the way I understand it:

A/(A+DM)=L, where A is attack value, DM is defense value after modifiers and L is the RNG "line". A random number is generated and a hitpoint subtracted based on which side of the line it falls on. It would also work as DM/(A+DM), just backwards.

For instance, if your swordsman is attacking an unfortified spearman on flat ground (10%), it would be: 3/(3+2.2)=.577 (although I don't know the level of precision the game uses). Then random numbers between 0 and 1 are generated. In this case, if it falls below .577, the spearman loses a hitpoint, and vice versa if it falls above .577.

That has got to be the stupidest combat odds calculation I've ever seen in any game, especially considering how low the health element is. It's no wonder the game can produce results of the kind I've been seeing. Thanks for explaining that to me, combat makes MUCH more sense now.

In your tank example, there's a 175% cumulative bonus for the pikeman, making his defense 8.25, so your "line" is 16/24.25, or .660, giving your tank a roughly 2 out of 3 chance of winning each individual hitpoint battle. It should, holding true to the statistics, lose 1 or 2 hp, but win the battle. However, it's not that far-fetched to imagine the pikeman getting three or four lucky rolls, especially if you've ever played roulette and bet the columns.

You see, there's the problem. The odds calculation here is WAY off. Granted, the Pikeman gets several bonuses here that jack up his defense dramatically, but it's still 2:1 odds (roughly). If the game is going to use such low health to determine outcomes, the results must bear out significantly more linearly. Now that I understand how combat is actually resolved, I can see why the random element comes about far more often than it should. Adjusting health to max for all units should reduce this significantly.

Also, concerning the number of "lucky rolls", don't forget that when dealing with random events, past outcomes have no bearing on future odds. IOW, most people are tempted to scream "what are the odds of that 1 in 3 chance coming up AGAIN?!?" if, say, your tank has lost a couple of hp in a row. The answer is 1/3, just like it was last time and just like it will be next time, regardless of whether or not it comes up this time.

I'm unfortunately far too familiar with statistical odds What I was referring to in my posts was more of a logic perspective of a Pikeman somehow whipping out a Spear from God to take on a tank kind of thing There are certain confrontations that IMO, should always bear out the same, no matter what. Damage should be applied, to be certain, but a Pikeman facing a tank, should ALWAYS lose, with a very very small chance of winning. The same holds true for something like a Battleship vs an Ironclad. It's ludicrous that an Ironclad should ever be able to go toe to toe against a Battleship and win.

Sounds like you've just hit a run of really bad luck.

Yes and no I suppose. Now that I can see how combat is resolved, it's clear that even a slight dip towards the 'bad' rolls can have significant results. Though, none of this compares to going 12 down hands straight of Blackjack

As for the mods, what you mentioned is the right place to look. You may want to add one or two starting hp per age past ancient, based on when the unit becomes available. The "Bonus" part is a bit misleading, bump that up one for say Cavalry and regular Cavs are produced with 4 hp, vets with 5, etc.

I'm fiddling around with the bix file now. Hopefully that will give more expected results. Hopefully

- Z

[Vlado]

Do you have preserve random seed on? Because if you do you'll get the same result everytime you check the combat after a reload unless you do something inbetween. And all this other stuff about random chance etc...

[Solomwi]

That has got to be the stupidest combat odds calculation I've ever seen in any game, especially considering how low the health element is.

I've grown to kind of like it, due to one piece of advice I found here on 'Poly. Instead of sitting there thinking "wait a minute, no way can a Pikeman beat a Tank!", I view it as "How in the @$!^ am I supposed to win a war with such asinine tank commanders? Do they not know to maintain those things?!?" Actually, due to that and considering that tanks and other machines at times beat themselves (malfunctions, operator stupidity, etc.). The second doesn't always hold up considering the scale each unit represents, but it helps while playing.

Though, none of this compares to going 12 down hands straight of Blackjack

Twelve?!? That's ALL?!? Rookie. Craps is my game, and I've seen just as strange or stranger there, but I have had those unbelievable blackjack runs, too (6 straight 16's not too long ago, on the way to a horrible horrible run).

[Dr. A. Cula]

This is a very old thing: I remember one game of Civ 1 where I attacked an enemy city with a knight. Turned out it was defended by a mech. inf., but I still took the city.

[Zapaan]

Obviously I'm doing something wrong here. I've edited the civ3x.bix file to adjust all unit health as well as the level health, but I'm not seeing any difference ingame. I assume I'm missing a critical step somewhere? Or am I mistaken in thinking that the change will have any kind of effect in an existing game?

- Z

[Gramphos]

No edits will effect an ongoing game (since civ3 1.17f IIRC)

[Zapaan]

Doh! Ah well, guess I'll have to fire up a new one

Btw, I'm having trouble with your C3MT utility. For some reason I can't save at all anymore, as whenever I try, it crashes. I've tried uninstalling and reinstalling, but that did jack squat. Any ideas what might be causing this?

- Z

[Gramphos]

I would need to see the files that causes this probelm, and a list of things doen to it to make the problem.

[Stuie]

Zapaan - I would like to make one additional suggestion beyond modifying the game: soften your targets before an outright attack.

The way combat is calculated in this game, the use of combined arms is greatly rewarded. You should make sure you use some artillery (catapult, cannon, etc.) to knock a couple hit points off your intended victim, then attack them. This will greatly increase your odds of victory.

Likewise, if you can keep some arty stacked with your defending units, they get a free shot at any attackers before the main combat is resolved, again increasing your odds of success.