Civ III: Conquests Patch Notice

[Dominae]

Sounds like an awesome idea to me!

[DrSpike]

It's important to realise that firepower, more hitpoints, and averaging techniques are statistically all similar ways of attaining the same goal (edited much later: i.e. altering the probabilities for a single battle). It doesn't really matter what label you put on it, except to appease the 'reality' whores.

[Dominae]

SpencerH's suggestion is a much toned-down version of Firepower. It addresses the Tank versus Spearmen "problem", while keeping true to the original design of Civ3.


Dominae

[RobC]

That's not a bad idea, making the averaging only come into play when units are from different eras. The dominance of cavalry will be certainly blunted once Riflemen become commonplace (it's just a couple of techs away, and the AI will usually beeline for Nationalism). It would also slightly reduce the useful lifespan of units like the immortal, gallic warrior, and dromon, but they are pretty powerful anyway.

[padlock]

Originally posted by DrSpike
It's important to realise that firepower, more hitpoints, and averaging techniques are statistically all similar ways of attaining the same goal. It doesn't really matter what label you put on it, except to appease the 'reality' whores.

I disagree. Let me give you an example:

Unit with an attack 2 attacks a unit with a (modified) defense of 10

If both units had only 1 hit point, the attacking unit would have a 1/6 chance of winning (2/10+2). Another way of looking at this is to say that it would take an average of 6 units to win the attack.

Now if each unit had 6 hitpoints, the chance of the attacking unit to win the battle completely goes down dramatically. However, the attacking unit will on average do 1 point of damage, and therefore it will still take an average of 6 units to win the fight (assuming no healing is done in between the attacks).

The same holds true no matter how many hit points you add. In fact, the higher the number, the more likely you're going to need exatcly 6 units (as opposed 3, or 4, or 7, or 8, etc.).

As you can see, adding more hit points doesn't affect the relative power between the units, it simply reduces the uncertainty.

This is in sharp contrast to the 'averaging' system which was proposed for the beta, which acts simply as a bonus multiplier to the stronger unit. While I haven't done the math, I can tell you that under that system, averaging 4 dice would mean that instead of needing 6 Units to win, you might need something like 100.

Quite a different affect!

[DrSpike]

Well I have done the math, many times whilst analysing the combat model for Civ2, then later when Civ3 came out. There is not a perfect matchup (hence why I said similar) between the techniques, but they all essentially have the same goal.

You even said it yourself, you just don't realise it.

"As you can see, adding more hit points doesn't affect the relative power between the units, it simply reduces the uncertainty"

Correct. This is what all 3 methods do though.

"This is in sharp contrast to the 'averaging' system which was proposed for the beta, which acts simply as a bonus multiplier to the stronger unit"

Incorrect. The averaging system proposed (well in fact there were 2 subtly different ones analysed) works to change the overall odds of the stronger unit winning by reducing the variance in the distribution(s) from which the key numbers come.

I'm not going to say any more on this..........I knew I shouldn't start the debate. I teach this stuff all week, the weekend is a break.

[Arrian]

Spencer's idea sounds interesting.

If implemented, however, I'd suggest a free upgrade of Immortals to Med Inf, to prevent disadvantaging the Persians. As it stands now, IIRC, Immortals would upgrade directly to guerillas.

Any other units that have the same stats as a unit in the next era, in the same upgrade chain? I can't think of any...

-Arrian

[alexman]

Hoplites and Numidian Mercs.

[gunkulator]

Originally posted by SpencerH
Let me contribute this to the discussion though. Is there a way to use the combat averaging idea to bring back the firepower concept from civ2.

ancient unit v ancient unit no averaging
ancient unit v middle age unit average of 2
ancient unit v industrial age unit average of 3
ancient unit v modern age unit average of 4

etc

How would such changes effect the game?

Interesting idea, however it would have some unforeseen results with a few UUs. Immortals would actually have a greater chance of defeating Pikemen despite pikes being a more modern unit. Similarily, Sipahi would crush Riflemen in the open.

[SpencerH]

I agree that changing HP and A/D values has similar effects. Thats why they have been so heavily used in many mods to try to balance the combat inconsistencies. Unfortunately, there are some pretty severe graphics limitations to the number of HP you can add. That problem became especially true for armies with elite units with 7-8 HP each. My own mod used completely re-done A/D values to the same aim but if you wish to play other humans you all have to use the same mod (which can be problematic).

This is the first time that I'm aware of that Firaxis has tried to openly correct the more glaring strings of combat uncertainties while trying to maintain an element of 'the imponderable' that occurs in combat. I believe that they have been limited in this regard by the original design concepts (ie no firepower and no unit superiority flags) and that this is an idea that was believed to be a workable solution within the original game framework. Averaging combat results may be the only practical way to address this question. I think we should put forward ideas that may be actually implementable.

[DrSpike]

I actually think your proposal is reasonable, as would be expected in replicating an effect with a long pedigree from Civ2 and SMAC. I just don't like people not realising the equivalence.

It seems probable you do though, and I confess I had not thought of graphical limitations as the binding constraints here!

Firepower was dropped and there was a statement some years ago about it making it too hard for people to calculate the odds of winning. This is why they dropped it and simplified the way hitpoints worked. I shudder at the thought of having to explain the variable averaging model to posters with a limited grasp of statistics............it's even worse!

However, as someone who kicked up hell when firepower was removed I am warmed to the core to see people trying to replicate its effects once more. Really the truest form of flattery.

[Catt]

Surprised that this hasn;t made its way over to 'Poly yet. A poster called TheNiceOne was responding to a post from someone who was disappointed with the decision to pull the combat change and stated:

Originally posted by TheNiceOne at CFC
But the problem is that since the patch is so necessary (to fix corruption and GPT bugs), this wouldn't be an opportunity to try out the feature, but rather us being forced to use this "feature" or live with corruption/GPT bugs for maybe half a year.

With that for the context, Jesse (as "Tavis") posted the following:

Originally posted by Jesse at CFC
Other than the half year part, you nailed that one TheNiceOne.

The plan for the addition was always to make it a configurable option. However, it's always better to hardcode for initial testing. I liked the change a lot, but as it has been correctly stated in the forums - Civ3 was designed with 1 roll, better to fix what was actually 'really' broken (corruption) then dig into playing with the other gameplay algorithms.

My primary concern was for multiplayer - the game is already time consuming, if you suddenly add on needing 5 warriors to kill a spear that wasn't fortified standing on desert then games would be REALLY long... most of the tests I ran initially were with higher A/D level units or with 31 AI and things were really smooth. After sending out the list and getting initial feedback, I got these test results:

(test performed attacking 0 terrain bonus tile defended by an unfortified unit (if fortified F )

10 Warrior vs 10 Spearman

1 Warrior Win, +3 other HP's, 3 PROMO's

10 Warrior vs 10 FSpearman

4 HP's removed 2 PROMO

10 Archers vs 10 Spearman

6 Archer Wins -2HPRatio, 1 PROMO

10 Archers vs 10 FSprearman

4 Archers Wins +3HP

Once towns, forts, hills, rivers, and mountains were added it became apparent that the Ancient Era would require significant rebalancing (especially if you ended up without iron!!)

My final tests were 20HP vs 20HP with a 100 percent offensive bonus and a 100 percent defensive bonus. You already know the result.

Keep the feedback coming! The patch is now 1.11 - I'll re-rerun the tests to make sure Mike isn't trying to be funny =)

Jesse

Interesting that the plan for the addition was always to make it a configurable option. Also interesting that the conclusion on retesting some aspects was that it "became apparent that the Ancient Era would require significant rebalancing."

As to the

Keep the feedback coming!

As I posted at CFC, unless I am wildly mistaken, I think it is safe to assume that the responsiveness and engagement Jesse and Firaxis have displayed at the fan sites will ensure plenty of feedback (hopefully constructive and reasoned) from the fans.

Again, thanks for the engagement from Firaxis

Catt

[Theseus]

[padlock]

Originally posted by DrSpike
There is not a perfect matchup (hence why I said similar) between the techniques, but they all essentially have the same goal.

Well, we will have to agree to disagree. As my example above pointed out, using one method would leave relative unit strengths the same (meaning it would always take, on average, X times as many units to defeat a unit X times stronger), whereas the other would change them drastically.

In any case, I too am ready for a weekend off from all this analysing

[DrSpike]

Look, all 3 methods can leave relative strengths the same, whilst improving the odds for the more advanced unit through addressing the variance (or replicating this effect) of the relevant distributions. That is why I made my post above. It is indisputable, so I'm not agreeing to disagree.

[padlock]

I don't want to get dragged into a long debate about this (I don't have the energy) but you are mistaken. The effects are very different. Averaging the random rolls does indeed affect the relative strenghts as opposed to the distributions. This is because the averaging is occuring in between the smallest quantumn unit (in this case the unit hit point).

In fact, it can be clearly seen that averaging, is very similar to increasing the difference between A/D values which is also quite similar to instituting the concept of fire power.

So in summary

Averaging rolls == FirePower == modifying A and D values (all 3 of these change relative strength)

Averaging rolls != increasing hitpoints (keeps strength the same and reduces variance)

That is indisputable.

[Arrian]

Thanks for that repost Catt! Great stuff, particularly the part about their intent to make it a configurable option.



-Arrian

[Sarxis]

So when will the patch be out? Friday (the norm)?

[Catt]

Originally posted by Tiberius

I have the feeling that most of people don't want the change because it'd change their style of playing, their good old strategies and tactics that worked so far. It doesn't really matter whether this is good or not, it is just: "Oh My God, what will happen now with my horse rush?"

Originally posted by Tiberius

Why would the new rules detract from gameplay is beyond me. I have no means to prove it whatsoever, but I am confident that if the new, now withdrawn model, had been applied from the very beginning of civ3, the crowd now asking for its withdrawal would have asked to keep it.

Sorry to take this thread back a bit off-topic, and in what may be an unpleasant way, but this is a serious pet peeve of mine and a serious (though often unintended) challenge to vigorous discussions in the forums.

Argue for your view! Debate the reasons underpinning opposing views! Poke and prod and challenge the views of others! Engage in menaingful discussion!

But for goodness' sake don't leap to impugning the motives of those who disagree with you. Disagreeing, or finding the opposing view "beyond you," does not invite you to (1) cast aside the stated reasons underpinning the opposing opinions and instead rely on the nefarious "feeling" you have that the real reason must be something else, particularly something less noble; or (2) argue that the "crowd" that disagrees with you is lemming-like and wants no change simnply out of a dislike of change.

Putting aside that such arguments are not directed at the issue and therefore are tremendousaly weak substantively, they are extraordinarily rude and they discourage the free exchange of ideas.

[/rant] I will stop now.

Catt

[DrSpike]

Originally posted by padlock
I don't want to get dragged into a long debate about this (I don't have the energy) but you are mistaken. The effects are very different. Averaging the random rolls does indeed affect the relative strenghts as opposed to the distributions. This is because the averaging is occuring in between the smallest quantumn unit (in this case the unit hit point).

In fact, it can be clearly seen that averaging, is very similar to increasing the difference between A/D values which is also quite similar to instituting the concept of fire power.

So in summary

Averaging rolls == FirePower == modifying A and D values (all 3 of these change relative strength)

Averaging rolls != increasing hitpoints (keeps strength the same and reduces variance)

That is indisputable.

Settler: The whole point of what I initially said was that all 3 methods (hitpoints, firepower, the averaging technique discussed recently) change the relative odds of winning whilst KEEPING THE A/D VALUES THE SAME.

Of course the desired effect for any given 2 units could also be brought about by changing the A/D values themselves. This distinction of yours about whether the methods affect the relative strengths is just stupid. All the methods increase the odds for the better unit. What is more for any of the methods used a good statistician could pin down the way to get the equivalent result for any given pair of units using any of the other methods. This does not imply identical results over the entire range of values possible.......far from it in fact. But all the methods are pursuing the same aim. Period.

[padlock]

Originally posted by DrSpike


Settler: The whole point of what I initially said was that all 3 methods (hitpoints, firepower, the averaging technique discussed recently) change the relative odds of winning whilst KEEPING THE A/D VALUES THE SAME.

Of course the desired effect for any given 2 units could also be brought about by changing the A/D values themselves. This distinction of yours about whether the methods affect the relative strengths is just stupid. All the methods increase the odds for the better unit. What is more for any of the methods used a good statistician could pin down the way to get the equivalent result for any given pair of units using any of the other methods. This does not imply identical results over the entire range of values possible.......far from it in fact. But all the methods are pursuing the same aim. Period.

First to your "Settler" comment, am I supposed to deduce that the fact that you have more posts means you must be right?

Anyway, I'm going to try this one last time and then I quit.

Take the example I mentionted above (unit A is attacking unit B which is 6 times stronger)

Now if I multiply the hit points of both units by X, it will still take on average 6 units of A to defeat 1 unit of B (again if no healing occurs between attacks). As X approaches infinity, it will now take exactly 6 units of A to defeat 1 unit of B. The result will be completely deterministic, and what's more, the relative strenghts between A and B have been maintained (6 : 1)


Now lets say that instead of multiplying hit points, I average the roll X times for each round of combat. AS X approaches infinity, the roll will approach .5, and the attacker will always win. That means that I could attack with 1000000 units, and never even take off a single hit point from B. Sure the results are once again deterministic, but completely different from what they would have been before.


Do you see now just how far off from equivalent these methods are?

Please, before you hit reply spend some time and think about this. You seem like an intelligent person, and I'm sure you'll be able to see this.

As for me, I'm done with this discussion.

[Master Zen]

@Catt: Thank you for re-posting that

I for one would have liked to have this as an option, just to see how good/bad the change would be like, plus, if it ended up "bad" it could have been a good challange to try and beat the game with the odds even more against you.


@SpencerH:

That is an EXCELLENT idea. This is actually a more refined way of making A/D values even higher as the eras progress. Even so, Civ3 made units more powerful from a numerical point of view. Remember in Civ2 a Tank had only 10 HP and it could swarm against anything else. Now it has 16 (and MA 24) and they still sometimes lose (I've had MA's lose to Musketmen).

I think the "realism" would apply to units within the same era getting lucky against each other like what happens with the current system. But I still don't buy a spearman even having the KNOWLEDGE of how to creep in and disable a tank!


@padlock

Originally posted by padlock

First to your "Settler" comment, am I supposed to deduce that the fact that you have more posts means you must be right?

don't worry and keep on discussing, don't listen to spammers who seem to have their ego as high as their post counts

[DrSpike]

Originally posted by padlock


First to your "Settler" comment, am I supposed to deduce that the fact that you have more posts means you must be right?

No, the fact that I am right does. The fact that I teach statistics at university level frequently means I am right. The fundamental laws of the universe mean I am right.

[Edit: apologies to all for any immodesty......I see from MZ's post people who don't know me might get the wrong impression from these posts. Please try to understand, at times like these when the statistics behind Civ come into the limelight everyone has an opinion, often a very strong one, on things that they may not truly understand (even for those with some knowledge in the area, as it is evident Padlock has). As such it can be incredibly frustrating, so I never get involved. After dodging for several days I made one small post on stats...........that was a mistake. ]

Originally posted by padlock

Take the example I mentionted above

No one in this thread has claimed that the 3 methods are identical for all possible attack and defend values. In fact I went out of my way to say that several times. A fundamental equivalence can be derived for any one set of values.

For instance using your example above in your first post I could derive the exact extent of averaging necessary to give the exact same results in your example as increasing from 1 to 6 hitpoints.........it would only be slightly higher than 1 given the extent of the difference between the A/D values, but it could be done. This in no way proves that averaging that exact number of times would be exactly equivalent to a different change in the level of hitpoints, or indeed that for any averaging model there will be more than one change to the hitpoints that is equivalent

BUT THERE IS ONE! IT IS UNDENIABLE. IF YOU MAKE ME I'LL DO IT TO PROVE IT TO YOU.

You are not entirely wrong in your initial post (though some of your later analyses are spurious), but you were criticising something that no one said. When averaging 4 times the extent of hitpoint change necessary to be equivalent in your 2A and 10D would be enormous. Hence showing that changing the hitpoints from 1 to 6 is not equivalent to averaging 4 times is not at all difficult - even someone with no statistical skill can do it.

For the last time, all 3 methods can alter the probability of the stronger unit winning whilst retaining the same A/D values. What is more you can derive the change in one model that is equivalent to the change in another. That was all I said, and it is correct, I would stake my entire career on it.

[Myrddin]

padlock

Are you sure you understand the proposed averaging?

Surely the effect of the averaging in your example is a give a win percentage for the attacker for each round of 0.166

[padlock]

Originally posted by Myrddin
padlock

Are you sure you understand the proposed averaging?

Surely the effect of the averaging in your example is a give a win percentage for the attacker for each round of 0.166

As intuitive as that seems, that isn't the case.

Assuming combat results are calculated as follows:

Attack successfull if: R * (A + D) >= D

where R is a random number between 0 and 1.

Then averaging R a certain number of times will make it tend toweards .5 (Averaging it an infinite number of times will make it exactly .5)

This would lead to the stronger unit always winning.

Even averaging R over a relatively small number like 4 will have a drastic change on the odds, with even relatively small differences between A and D (say 25% or so). In the case above a successfull attack would go from your quoted value of .166 to something well over an order of magnitude lower.

[padlock]

Just to follow up on my last post.

Someone on the civ fanatics board had an excellent analogy.

Let say you are playing a dice game where you have to roll a six to win. If you're only using 1 die, then you've got a 1 in 6 chance to win. If instead you are rolling 4 dice and taking the average, then all 4 have to be a 6, so you would now have a 1 in (6 * 6 * 6 * 6) chance of winning (ie. a 1 in 1296 chance).

[DrSpike]

Yes, such games are the way to proceed.......I commend you on that. Let's say you have multiple sets of 4 dice, and roll each of the sets. Each of the sets has a 1 in 1296 chance, but you have multiple sets now, so could you still have an overall chance of 1/6 of winning?



Ooh look I offset the averaging by increasing the repititions. And I could do it for any finite amount of averaging too.

Hehe it's not a perfect analogy for the previous argument, but it should help nonetheless.

[vmxa1]

Thanks for the post Catt. It is very gratifying to see that Firaxis is paying attention and is flexible.

I also commend you pointing out that just because one does not agree with someone, that does not me they are evil and incapable of thought.

[Vince278]

Originally posted by vmxa1
Thanks for the post Catt. It is very gratifying to see that Firaxis is paying attention and is flexible.

I also commend you pointing out that just because one does not agree with someone, that does not me they are evil and incapable of thought.

Ditto.

[Tiberius]

Originally posted by SpencerH
ancient unit v ancient unit no averaging
ancient unit v middle age unit average of 2
ancient unit v industrial age unit average of 3
ancient unit v modern age unit average of 4

That is a good idea, SpencerH