Civ III: Conquests Patch Notice

[T-hawk]

Kami_Mercenary:

If your army is comprised of mostly weaker units than your enemy, did it occur to you that you mighht *deserve* to lose?

NO. Reducing the combat gap between units of different power levels was a deliberate design decision made way back when. It's not good for game balance for whoever gets cavalry first to obliterate anyone with muskets, or for tanks to rip apart any civ that happened not to get rubber. The original system somewhat reduced the impact of this.

I also add that this entire change has absolutely nothing to do with streakiness. I ran some tests a while back and any result is equally likely to come after any other result. All this does is *change* the streaks to be more likely in the favor of the higher-valued unit.

BTW, the programmers don't necessarily understand the game better than us. Soren Johnson has posted that he's mostly a Monarch and Emperor player and he gets beaten at Deity more often than not, while several of us have little trouble beating Deity in almost all situations.

 

[Speaker]

I would also like to give my thanks for such a quick patch, and add my voice to those who prefer the "old" combat model.

It seems to me the simplest fix would be to give a choice to the player which model they would like to use. Those of us who like the "old" system could continue to use it while those who are interested in beta testing could try out the new model. Can't see too many people being unhappy with this, assuming you Firaxians are dead-set on including the new model.

 

[Charis]

Mike B. -- thanks for your clarification

Think of each round of combat as the attacker rolling dice to see if they hit the defender. If the number rolled is greater than or equal to the defense value, the attacker hits; otherwise, the defender hits. The change we made was to the way the attacker rolls the dice. The attacker now rolls multiple times and the result is the average of all the rolls. ... A single spearman can still beat a single tank but it's highly unlikely that a single spearman will be able to stand up to a stack of tanks (which is as it should be).

Quite interesting, it's an asymmetric change for attack vs defense. That's actually quite a bit different from the 3-of-5 approach (and graph shown earlier). Here's an updated graph, if I understood it correctly that attacker gets multiple rolls with a virtual die of his strength and defender gets one with his defense.

There are several interesting things to note:

1. For low attacker chances (spear attacking a tank), the new chance goes asymptotically to 1/2 of ths old chance. (ie, 10% -> 5%)
2. For high attacker chances (tank rolling over a spear), the chance of the weak defender surviving go as the old percent *squared*. (ie 90%win/10% loss now becomes 99% win, which is I believe "as intended")
3. A 50% chance before remains a 50% chance now
4. The results in the low range are nearly identical whether you average 2, 4, or an infiinte number of attacker dice.
5. However on the high side, the more values you average the sharper the cutoff. (In the case of a very large number of rolls averaged, the attacker would ALWAYS win whenever the att:def ratio was greater than 2:1, i.e. a 67% chance in the old days)

But... since everything is by hit point, and the hp system is still there to 'even things out'
- On the low side, a 1/2 chance to win on a given hp is probably about a 1 in 4 to 1 in 8 chance to win the battle! (That's rather large, perhaps as intended)
- On the high side, what was previously a 67% chance to win on a given shot now becomes an 82% chance to win one shot, or for vet vs vet, a 83% overall chance to defeat the unit goes up to 98%. If your odds are greater than 2:1, the chance of defeating the unit go *WAY* up.

So yes, averaging rolls is much different from re-rolls, and is asymmetric as I think you desire, but should the win odds go up *that* much for more than about 1.5 or 2:1? Beware the Berzerk!

(BTW Mike, Tavis, I hope this is all taken as constructive comments with an aim to help and point out the impact of "math" changes, and not as screams of anguish from folks who haven't even played the beta yet As a math hound / developer myself, such discussion is like waving fresh meat in front of a starving man )

Charis

PS in EDIT -- an interesting approach would be to take what vet-on-vet win percentages should be over the wide range of warrior vs MechInf up to MA vs spear, and plot that as a function of att:def ratio. From that one can back-calculate what the single-hp shot percentage must be, and then come up with an easy-to-implement/code function to get this. I think the problems most folks are raising here is that 2:1 of cav over pikes should see no more than a small increase in real chance-to-win, and both the proposed new combat system and the misunderstood version lead to a cav vs pike or MDI vs spear blowout that will result in a race-to-tech. It's also likely that "the masses" are way over on the right side of the curve upset that their tank loses once in a blue moon vs the pike they do in fact face, while the deity players see themselves on the opposite side of the fence, knowing that now there will be no hope for them, still researching Feudalism while Germany gets Mil Tradition!

 

[waylander]

Originally posted by Mike B. FIRAXIS
Again, though, this is a BETA patch (but several of us have been playing with it for a few months now) so things could change before the final release. I hope this info helps clear some things up .

If they have been playing with this change for months it's obvious that the change is in no way as drastic as some board members are trying so hard to make us believe. You people need to sit back and wait for the patch and give it a try before you predict the end of CIV as we know it. Even after completely misinterpreting the initial message some people can't seem to stop themselves from crying doom and gloom when they have less than the full picture.

For my part I want to thank everyone at Firaxis and Breakaway Games for the excellent job they have done on C3C. Even with the corruption bug I have been playing it solid for weeks and having a wonderful time with the game. Thanks for a great addition to the CIV lineup.

 

[eliliang]

Originally posted by T-hawk

I also add that this entire change has absolutely nothing to do with streakiness. I ran some tests a while back and any result is equally likely to come after any other result.

This doesn't mean that the PRNG is not streaky since your tests were presumably conducted on the behavior of the game, after the PRNG result was transformed by the game mechanics/calculation. A better test would be to examine the raw output of the PRNG and test that using something like George Marsaglia's "Diehard" tests (George Marsaglia's Diehard Tests ) or something similar.

 

[alexman]

Charis, I don't think the defender gets a roll at all.

 

[eliliang]

Originally posted by waylander

Even after completely misinterpreting the initial message some people can't seem to stop themselves from crying doom and gloom when they have less than the full picture.

There wasn't any misinterpretation. We were given incorrect information.

Now we are interpreting on the basis of what we believe is correct information. If this is still not correct information, then of course our interpretations could also still be pointless. If the information is now correct, then I believe that my interpretations are correct as well.

 

[Arathorn]

OK...some numbers, under the following assumptions!!!! It may or may not be correct.

Old method:
- To determine who "wins" a round of combat, first a real number (R) between 0 and 1 is generated (something like a random 32-bit integer over 2^32). There are a number of semi-equivalent ways of doing this.
- That number (R) is compared to the real number value of attacker offense divided by (attacker offense + defender defense), where the latter number (Wpct) is obviously affected by terrain, fortifications, etc.
- If R<Wpct, the attacker wins the round and the defender loses a hp. If R>Wpct, the defender wins the round and the attacker loses a hp. (R=Wpct has very little chance of occurring, so I ignore it)

This matches the current combat calculators results, if not necessarily the method.

BIGGEST ASSUMPTIONS HERE:
- R is now calculated as average of R_1, R_2, R_3, R_4, where each R_i is an independently-generated random "real" between 0 and 1
- The rest is the same.

Under these assumptions, which may or may not be accurate... All units are assumed to be vets, for right now.
- A longbow/immortal/MDI (4) attacking a spear in the open (defense 2.2) will win 98% of the time. (was 79%)
- A longbow (4) attacking a fortified spear in the open (defense 3.3) will win 77% of the time. (was 60%)
- A warrior (1) attacking a warrior in the open (1.1) will win 36% of the time. (was 45%)
- A cavalry (6) attacking a musketman fortified in a city on flatlands (6.3) will win 35% of the time. (was 47%)
- A sword (3) attacking a fortified spear on flatlands (3.3) will win 39% of the time. (was 45%)
- A horse (2) attacking a fortified spear on flatlands (3.3) without retreat will win 4% of the time. (was 25%)

It's a huge, drastic change, no matter how you look at it. I just don't see how it will be balanced.

Arathorn

 

[eliliang]

Originally posted by Charis
Quite interesting, it's an asymmetric change for attack vs defense. That's actually quite a bit different from the 3-of-5 approach (and graph shown earlier). Here's an updated graph, if I understood it correctly that attacker gets multiple rolls with a virtual die of his strength and defender gets one with his defense.

I think that we have to wait for a more exact description of what is being done before we can run the numbers. My interpretation based on the high-level info Mike gave seems to be different than yours.

 

[Charis]

think that we have to wait for a more exact description of what is being done before we can run the numbers. My interpretation based on the high-level info Mike gave seems to be different than yours.

Eliliang - this is quite true. I'ld be delighted if we find out more, in this direct fashion. If not, you know that some experimentalist among us will spend a few hours testing to find out the data, and a theorist (cough) will then fit the data.

Alexman - That amplifies the above comment about a need for more info. If there were no defender roll at all, then an attacker strength less than defender strength would give 0% chance to win, so I doubt that's the case. Hrm, unless it's somethink like what Arathorn said. My assumptions were similar except with the defender roll you mentioned. Darn, now I have another set of simulations to run...

Charis

 

[Arathorn]

It's not about winning/losing, it's about game balance and strategic options. I feel very comfortable that T-hawk, Speaker, Charis, me, and a number of others could still comfortably beat deity with the changes...but there would be one path to do it. Get a tech lead, beat up on others. YAWN!

Now, I can play from behind and go for a GL grab, or I can push culture aggressively early and catch up on techs with better land later, or I can do all kinds of wild and crazy things (no fast troops, no slow troops, etc. etc.). You can play with better troops, similar troops, worse troops and still have a game. With changes, you can play with better troops or lose. I'd find another game first.

And I still note that nothing is done about streakiness. The results will be just as streaky -- they're just distributed sharply more in favor of the guy with the stronger troop -- even if it's only narrowly stronger (1 vs. 1.1 which is about as even as you can get goes from 55/45 to 64/36!).

Arathorn

 

[Grey Fox]

Originally posted by Arathorn
... a fortified spear on flatlands (3.3)

A spear fortified on flatlands is NOT 3.3.
It's 2.35. 10% for the terrain, 25% for the fortification. <- EDIT: This is wrong I know (it's 2.7)

 

[T-hawk]

I'm mostly worried about the cases where the attacker's strength is somewhat, but not prohibitively, less than the defender's strength. The common situations of this are knights (4) against muskets fortified in a size 7-12 city (7), cavalry (6) against rifles fortified in the same (10.5), and so on. If Charis' assumptions and calculations are correct, we'll need twice as many knights or cavs as we used to in these situations. This completely and totally changes the combat balance of the game.

 

[padlock]

Originally posted by Mike B. FIRAXIS

Think of each round of combat as the attacker rolling dice to see if they hit the defender. If the number rolled is greater than or equal to the defense value, the attacker hits; otherwise, the defender hits.

I've been trying to reconcile this statement with what we know (thought we knew) about the expected combat odds in Civ 3, and I think I understand how it might work.

Assume R is generated by the random number generator, and is a float between 0 and 1.

Then using the following rule:

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

would give us the expected results.

It seems like the change that has been implemented is that R is now averaged over 4 'rolls' and thus will tend to .5

Intuitively, I feel that this change does have a fairly major, and detremental, affect on combat when there is a moderate discrepancy between A and D, but I am far from a statistician, so I'll leave it to others to figure out how much.

 

[Grey Fox]

Originally posted by Charis
Alexman - That amplifies the above comment about a need for more info. If there were no defender roll at all, then an attacker strength less than defender strength would give 0% chance to win, so I doubt that's the case. Hrm, unless it's somethink like what Arathorn said. My assumptions were similar except with the defender roll you mentioned. Darn, now I have another set of simulations to run...

The defender don't have a roll, it has a value, for example; if the RNG is between 1-1024, if its a battle between Attack: (3) and defense: (2) (to simplify), that's 40% chance for the defender. Which means that the defenders number is 410. So the RNG is a number between 1-1024, if it's higher then 410 it's a hit.

 

[varyag]

Nobody noticed my comment. That is depressing. Well, I am not a very proud type - I repost

The current battle calculation is flawed , but the proposed in beta 1.10 of Conquest remedy is even worse . Remember the goal - to mitigate the probability of impractical outcomes without unbalancing the game. Here is my solution (just 2 cents actually)




Roll the dice (the way it is done now). After you get an outcome, toss an unbiased coin - 50/50 chances for both combatants
If the toss confirms the roll, then the outcome for 1HP is decided.
If not, then do TWO more rolls and tosses. Now we have SIX outcomes. Assign weight of a roll to 2 and weight a toss to 1. Whoever gets better score wins the battle for 1HP


[dance] [dance]

 

[eliliang]

Originally posted by Charis

Eliliang - this is quite true. I'ld be delighted if we find out more, in this direct fashion. If not, you know that some experimentalist among us will spend a few hours testing to find out the data, and a theorist (cough) will then fit the data.

Actually, it's the experimentalist that fits the data. The theorist tries to find a closed-form solution.

But that wasn't my point. In this particular case, I had a different interpretation of what Mike said. In other words, I don't believe the defenders get a roll at all. That's why I wanted to wait to see the exact description.

 

[eliliang]

Originally posted by varyag
Nobody noticed my comment. That is depressing. Well, I am not a very proud type - I repost

The current battle calculation is flawed , but the proposed in beta 1.10 of Conquest remedy is even worse . Remember the goal - to mitigate the probability of impractical outcomes without unbalancing the game. Here is my solution (just 2 cents actually)




Roll the dice (the way it is done now). After you get an outcome, toss an unbiased coin - 50/50 chances for both combatants
If the toss confirms the roll, then the outcome for 1HP is decided.
If not, then do TWO more rolls and tosses. Now we have SIX outcomes. Assign weight of a roll to 2 and weight a toss to 1. Whoever gets better score wins the battle for 1HP


[dance] [dance]

Sorry. This changes the probabilities also, just to a differing degree and in a different way than the scheme Mike described that would be in the patch.

 

[Arathorn]

Padlock -- that's the same as my interpretation, except I divided everything by 1024 to get a float between 0 and 1. Except same concept but a slightly different way of phrasing it.

Grey Fox -- WHOOPS! Where did I get 50% fortification bonus into my head???? I think it's Charis' fault

4 vs. 2.35 -- 97% new (77% old)
3 vs. 2.35 -- 82% new (63% old)
2 vs. 2.35 -- 28% new (41% old) Note: no retreats calculated!

These are true but also mis-leading...spear should defend at 2.7, no? 2*(1+.1+.25) = 2*1.35 = 2.7 (10% for flatlands and 25% for fortification)

4 vs. 2.7 -- 92% (was 70%)
3 vs. 2.7 -- 64% (was 56%)
2 vs. 2.7 -- 13% (was 34%)

Arathorn

 

[varyag]

Originally posted by eliliang


Sorry. This changes the probabilities also, just to a differing degree and in a different way than the scheme Mike described that would be in the patch.

That is the entire point as I said to improve the system without unbalancing the game. Do the math and distribution on what I said, and you shall see