Programmed Irritation

[Buttercup]

Well now, quite a lot for me to respond to here so please don't be offended if I miss out something that someone was hoping would spark my interest more than it does, lol. I shall try to go one-by-one.

The AI does seem to be programed to react more to player moves, than to have been programed with much strategy.

I like this quote immensely and is one that relates to pretty much every game I play. An excellent example of this is the difference between attempting a 20k Cultural single city win and a game where one pumps out military Units. Both times I've tried the 20k victory on Regent (one Large Pangea, one Large/Huge Archepeligo) no-one sent any troops in my direction and both games were what one would describe as 'quiet' games. However, whenever I start to mass produce Units, either for pre-emptive defence in a Space Race or as preparation for conquering, I find a civ sending over a fully determined stack, as illustrated in the Opening Post of the thread.

Again, just one more example among many. This may not have been programmed as Irritation but it certainly suggests a programme which is more occupied with what the human player is doing than what state it's own affairs are in which, by default, is essentially going to create an overall agenda of Programmed Irritation.

It's all reactive.

Yes, indeed!

I propose that is for the same reason you'll sometimes find my units fortified next to barb camps: not because I expect Cyc's archer to show up any turn and I want to tease him, but because I found two barb units in the camp, got wounded while killing the first barb, and fortified to heal up before attacking the second.

Yes, I have done this before as well. The issue of taking out Barb camps is one I'm not overly familiar with so I'm just all ears on this subject, but I know I like everything both you and Cyc have said in this thread and you both strike me as the kinda players I like so please don't fall out over Barbarian Camps. I'm happy to accept Cyc has noticed Barb Camp Irritation and I'm happy to accept there might be other reasons for this occurrence other than purely to irritate.

the AI will not only place units on chokepoints, no, they will, if they are able, build a fortress/barricade too.

Ah, now, this is an example on 'logical tactics'. I have experienced the AI build Fortresses and Barricades at choke points in many games. I would not call this Irritation because it's precisely what the AI *should* be doing if it were a human player going for a win. The Irritation I am referring to is pointless path blocking that has no strategic value other than to 'slow down' the 'inevitable', what might be described in old board games as 'ha-ha, you miss a turn!'. Also, and the irony here is quite amusing for me, you actually inadvertently highlighted another Irritation in the programming - That being that the AI might build these defences, but it wont bother defending them with anything more than a token force. It's using a very powerful tactical tool to simply 'irritate' progression rather than put a halt to it.

or it could be "We don't know exactly what we want either, but after some tweaking this looks ok."

I like this quote, it suggests a very human angle to the programming process. Balancing is one of the hardest factors to get right in any game of any type. Resources are, without doubt, the key balancer in Civ 3 and an area they clearly had major problems with upon release (hence the introduction of non-resource Units post-vanilla) and might even be the key in explaining some of the key 'problems' with ludicrous battle results. The net result of 'trying to keep the game playable for people without resources'. One can't just 'guarantee' everyone gets resources, that destroys the point of them, but, at the same time, one can't just make them all ridiculously hard to attain as the 'majority' of 'normal' players will just rage quit in anger.

I think they should have made resources Civ specific, this would have made much more sense and added far more Role-play-ability to the concept they were trying to achieve. As has been noted on these very forums, what exactly is the point of playing Scandinavia on an Archipelago and then not giving them Iron on their isolated island. They might as well have started as any other Civ and all the effort put into roleplaying different civs is just totally wasted as one is likely ending up using Vikings to attack Riflemen or Musketeers, which is totally out-of-shape of them being the biggest menace in the late ancient/pre-middle period in history.

entrail reading

Mmmmnomnomnom

Are we really having this discussion?

Apparently you have me on ignore so I can say anything I want about you. HA-ha. Big nose.

One of the prime aspects of the AI programming is to identify a rival tribe which is weak, and attack it.

Not at all. Though I enjoyed all of your post, this line stands out as suggesting a reply is vital. The most common complaint I read on the forums in this respect is that the AI attacks you *when you are ahead* and NOT when you are behind. I remember reading one post where someone played all their games deliberately 'just behind' the leading AIs simply to avoid this kind of Programmed Irritation, regardless of their civ's actual output capacity. I myself have had games where I have no more than 10 Units the entire game and by a mix of Diplomacy and AI Reactive Play (see above) I have never been in danger from invasion. It's these very issues which are at the very core of the musings about Programmed Irritation.

It does come as a bit unusual, but it does come as possible enough that you can't necessarily infer that the programmers programmed the game to mess with your psychological state.

Messing with a player's psychological state is a very important aspect of programming, never underestimate the minds of game programmers! I have no doubt they don't use the exact word 'Irritation' when deciding on 'balance' issues and 'challenge' issues, but Programmed Irritation is often the end result!

You would need to have a thorough and long list of battles of exactly what happened in what situations.

Now this is an idea I definitely LIKE +1 !

More artillery type units generally implies fewer hitpoints for the AIs when attacking them which generally implies less irritation.

As you write this I've just finished a Monarch Domination in 800AD for 6500 points where I didn't use a single piece of Artillery as the key to the success and huge point scoring rested not on waiting for Catapults to slowly plod round the map but because I pumped out non-stop Chinese Riders who could eliminate half an empire in one turn. Thanks for the advice and all, but this thread isn't about the 'best way to win in this that or the other scenario' it's about the method behind the madness of the AI.

If you looked at a much larger data set here

The large dataset you refer to is called 'experience'. If people notice something on a regular basis, they make it a point for discussion, this is when tests are required to confirm or deny suspicions. Examples from both sides are required, both yay sayers and nay sayers and by simply discounting out-of-hand you are being as irrational in your rejection of the idea as someone, such as myself, who is probably over stating the case.

who needs a news paper in the morning with a Thread like this one.....

Now THAT should have been the 100th post! LIKE +1

 

[Raliuven]

Once we've established that the AI is not playing to win, but can win (or at least destroy the human or stumble upon a victory) if the human player does not play 'well', then the AI must have a purpose otherwise the game would just be a lumberjack game of cutting down static Units and cities.

The interesting thing would be to consider what would happen if the AI were programmed for a specific victory. If the victory condition was randomly applied to a civ (after all, pre-programmed VC by civ would be boring), then you would not only need to race against any AI that has decided to follow your victory path but also hope to win before the other AI complete any other victory condition. Of course then we would be discussing how easy it is to tell which VC they were going after and how easy it is to defeat them once it has been identified.

Ultimately this comes down to processing power. In order for the AI to be truly competitive each AI would need to evaluate its situation each turn (or X # of turns) and decided which VC is the most possible. It would need to evaluate its terrain and the potential of its empire (both of which we know it is terrible at) and its current position for each VC relative to every other civ. Since it calculates who to attack based on relative strength and aggression settings, it can't even formulate a plan to kill off Civs competing for its VC (for example, to purposely capture the UN to force a vote). It may have a program to 'attack the top dog' in attempt to simulate this, however. While this may seem true it would be difficult to test without bias.

I think we can summarize the AI's purpose: simply to survive in the presence of the human maniac that has invaded their world. Anything they do in support of that goal is valid, even if as a byproduct it irritates the human.

The end result is a process whereby almost every encounter with the AI in any respect seems to be of an irritating nature, as opposed to ineffectual (Cheiftan level excluded) or, conversely, actually challenging. Even on the harder difficulties the element which makes the game harder is enhanced irritation rather than improved intelligence, such as even more severe city/Unit spamming and even worse auto-tech-trading-between-AIs-which-the-human-isn't-invited-to, at least not in any logical way as any human player will refuse 25 Gold Per Turn in exchange for a few saved tech turns, something the AI won't refuse, though this is an element the human player can manipulate to their advantage in some circumstances, but this wasn't the intention.

There are a lot of things we know the AI isn't good at. There are many threads that bemoan that fact. But it is generally accepted as a lack of programming, not over programming. Because of these deficiencies they are given a production bonus at higher levels to help them compete with the human player who has a brain to help them devise complex ways of manipulating and exploiting the AI weaknesses. As this does nothing to improve the programming and only multiplies what is already present, it is not strange that if the game causes you irritation then a higher level game will produce enhance irritation. Or, if the game gives you a challenge at a lower level it will produce an enhanced challenge at a higher level. Sounds insane, I know.

So with the programme being so irritating in design, it suggests that, for some events, an Irritation modifier is directing AI motivations as well.

No, it suggests that the AI is limited in its programming and responses. Once again, processing power. This is not in any way new territory. Players can and do anticipate these events. I am surprised because I would have thought the complaint would be about how utterly predictable the AI responds.

When that AI settler gets to your spot quite literally 1 turn before you do, is it all simple coincidence or does the AI 'see' what you're building, where your moving said built Unit to and 'prepare' accordingly to intentionally get there that 1 solitary turn before you do?

Or, on an even more paranoid level, does it just 'magic up' a Settler out of the fog in order to add drama to the game?

Or is it all simple coincidence?

If any of this was true, why does the AI suck at amphibious invasions? Rather than dropping a single Longbow on my shore to attack my railed nation full of cavalry and infantry it should drop a legion of cavalry supported by muskets and cannons.

The more likely reason is how the AI calculates its movement.

Consider:
1) It won't move the transport out of its home port without a unit loaded. As soon as at least one unit is loaded, it will leave when its movement comes due in the movement cycle (or maybe it checks at the end of the turn and moves transports last). Because a single longbow could make it to the transport the boat sets sail (even if knights or cavalry are due to build next turn). Twenty turns later is washes up on my shore as tribute for leader fishing.
2) The AI is well aware of my tech level and the fact that my nation has use of steam power as well as the iron and coal to build rails. It knows I have 30 cavalry units stashed around my nation as well as infantry to hold border cities and artillery to ward off defenders. After consulting the Irritation Program it conjurers up a single caravel/galleon with a lone longbow to stage an invasion.

Although I don't know when the ship or longbow was created or when they set sail, I don't believe they were 'fabricated' by the AI. If the AI were fabricating units it would be pretty obvious, particularly in the REX stage.

And there are hundreds of little Irritations like this throughout the game, from awkward battles to least ideal terrain placement (you know, when a square is just in the wrong place to be used by any city without putting a city in a stupid spot, etc.). All of which seem to happen too often to be sheer coincidence but are not of enough significance to be, individually, evidence of 'blatant' programming.

Well if I were designing a planet then I wouldn't place large oil reserves under the deep oceans or in Alaska, but what can you do? On the other hand, being able to identify the mountains and hills that contain gold is much easier.

 

[Spoonwood]

Please use quote tags when quoting someone Buttercup.

Messing with a player's psychological state is a very important aspect of programming, never underestimate the minds of game programmers! I have no doubt they don't use the exact word 'Irritation' when deciding on 'balance' issues and 'challenge' issues, but Programmed Irritation is often the end result!

Programmed irritation can only come as the end result if programming lead to the situation which caused the irritation. You haven't shown that programming here has caused anything, nor does there exist enough information to suggest that programming has done anything with respect to your irritation here.

As you write this I've just finished a Monarch Domination in 800AD for 6500 points where I didn't use a single piece of Artillery as the key to the success and huge point scoring rested not on waiting for Catapults to slowly plod round the map but because I pumped out non-stop Chinese Riders who could eliminate half an empire in one turn.

Um, so what? The point scoring isn't huge there for a standard monarch map (if that's what you played), and plenty of XOTM games played at monarch level finish well before that. Second, what I wrote came with respect to the battle you fought with swords, ancient cavalry, and horses, NOT some sequence of battles with riders. So, what you've referred to here comes as irrelevant.

The large dataset you refer to is called 'experience'.

No, that's NOT a (relevant) dataset, because experience relies on memory and selective use of information. You could easily discount anything in your experience which doesn't fit your pet hypothesis, only remember the things which fit with your pet hypothesis, and only report the things which fit with your pet hypothesis. You need someway to systematically record everything that happens, or have a truly random and large enough sample. So far as I can tell, you don't have either here.

Examples from both sides are required, both yay sayers and nay sayers and by simply discounting out-of-hand you are being as irrational in your rejection of the idea as someone, such as myself, who is probably over stating the case.

No. The default hypothesis comes as the null hypothesis, in other words that no relationship exists between the variables. Examples which would seem to suggest the null hypothesis correct with respect to "programmed irritation" come as very easy to find. See basically any the succession game writeup, such as this one.

Second, I didn't simply discount out of hand with respect to your information. I provided an analysis of the defense values of the spears (3.7 at least) you attacked which implies that you had a low probability of winning out in one turn with the force you had. In other words, it didn't actually come as likely, or barely likely, that you would take that capital in a single turn. And you didn't. I also indicated that you didn't have all that much data.

Additionally, in effect, I mentioned that we no reason to think that you have a relevant data set.

Honestly, one can discount the implication of the original report by sampling bias. Is Buttercup (or Darski for that matter) just as likely to report an instance of where the RNG behaves more in his (her) favor? No.

Do we find threads like "wow! the RNG really has favored me today! How did that happen? Can anyone explain how I had such good luck?" No, we don't. And thus, given a person writing a report about the RNG producing unfavorable results, that sampling bias exists with respect to such a report comes as likely.

If you want to make a claim which goes against the null hypothesis here, and claim such a claim about the game as rational, then you'll need the balance of information to support your hypothesis over the null hypothesis. You don't have anywhere close to such, and some of the information you've given actually fails to reject the null hypothesis... that is, once you understand the defense values of the units you attacked (and how many hitpoints did they have? and how many hitpoints did your units have?).

 

[Lord Emsworth]

Also, and the irony here is quite amusing for me, you actually inadvertently highlighted another Irritation in the programming -

Hmmm, I think you are misunderstood. I did not point out the manning of the fortress "inadvertedly". I was fully aware what I did. I can give you more: the Longbow might have been just passing through; these were the forces that the AI was throwing at me. I can't remember. And the Pikeman would have served a much better purpose elsewhere.


On the other hand, at least you are not complaining about cheating.

 

[vorlon_mi]

Vorlon_mi wrote:
One of the prime aspects of the AI programming is to identify a rival tribe which is weak, and attack it.

Buttercup replied:
Not at all. Though I enjoyed all of your post, this line stands out as suggesting a reply is vital. The most common complaint I read on the forums in this respect is that the AI attacks you *when you are ahead* and NOT when you are behind.

So I will respond by saying, "you must be reading different forum postings than I am." Or we've both got the selective memory thing going on. Nearly every month, an inexperienced player starts a thread (usually over in "Strategy") about how they can win at Chieftain, but always lose at Warlord. The AI comes crashing in with more military than they can handle. The player will upload a save, and experienced players (such as vmxa) will load the save, and note that the newbie did not build enough military.

In two of my most recent games, I was expanding rapidly, with just a few warriors around to pop huts and fog-bust. Lincoln, my "friendly" neighbor, kept demanding tribute. He had more archers, and had hooked up his horses. I kept giving tribute, because I wasn't ready to fight him yet. But a thousand years later ... I've also had weakly defended towns located on little islands attacked, even though I was ahead in tech and military overall.

So, while I will grant that I am sometimes attacked when I am ahead, I would submit that I have been *consistently* attacked when I am perceived as weak. For those players who play at the higher difficulties, where the human starts from a position of weakness starting at turn 1,
isn't that consistent with your observations? That the AI will attack a weaker rival?

 

[PaperBeetle]

Well, I for one. I have about 10 years of experience. But I agree that your posts seem to be from someone who only has experience to play the game well enough to boastfully post about how someone else is wrong.

I have made no boasts, no claims at all about my level of skill or length of experience. A boast would be a comment such as "Well, I for one [do not have limited experience]." It is a rather silly boast, I think, given that 10 years is obviously not an unlimited amount of time.

 

[Buttercup]

After consulting the Irritation Program it conjurers up a single caravel/galleon with a lone longbow to stage an invasion.

Now you're talking my language!

You haven't shown that programming here has caused anything, nor does there exist enough information to suggest that programming has done anything with respect to your irritation here.

And you haven't shown anything to disprove it!

Um, so what? The point scoring isn't huge there for a standard monarch map (if that's what you played), and plenty of XOTM games played at monarch level finish well before that.

This post says a huge amount about your fractured relationship with a lot of posters

what I wrote came with respect to the battle you fought with swords, ancient cavalry, and horses, NOT some sequence of battles with riders. So, what you've referred to here comes as irrelevant.

Ooooooooooh, reeeeealy. Ok mr assumptive pooh-pooh, tell me, your entire army consists of 5 Units, two Ancient Cavalry, Two Swordsmen and a Horseman, now, which ones do you replace with Catapults in order to definitely effect a better, quicker result against three Spearmen?

No, that's NOT a (relevant) dataset, because experience relies on memory and selective use of information.

Well deary, the reason I and everyone else doesn't 'remember' having weaker Units defeat stronger Units is because *a human wouldn't make the decision to do that in the first place, the human quits the game when 30 Cavalry march into their 4 Cities that are defended by 3 Spearmen each and a human wont even bother attacking unless they have 'superior' tech or at least 'equal' forces*... duh.

You could easily discount anything in your experience which doesn't fit your pet hypothesis, only remember the things which fit with your pet hypothesis, and only report the things which fit with your pet hypothesis.

Which is precisely what you are doing with your responses.

You need someway to systematically record everything that happens, or have a truly random and large enough sample.

Yes, I know, I've already agreed this is a good idea!

So far as I can tell, you don't have either here.

And neither do you!

I provided an analysis of the defense values of the spears (3.7 at least) you attacked

And yet three Horsemen can breeze these stats suggesting the stats mean didly squat, and YOU seem to Forget I had Five Units, so each unit only needed to do 60% percent damage each, whereas the three Horseman, by you own stats, were having to do persistent 180% damage EACH. Yes, those stats really do mean squat don't they.

I also indicated that you didn't have all that much data.

I have acres compared to you.

Do we find threads like "wow! the RNG really has favored me today! How did that happen? Can anyone explain how I had such good luck?" No, we don't.

Well deary, the reason I and everyone else doesn't doesn't create such posts is because *a human wouldn't make the decision to do that in the first place, the human quits the game when 30 Cavalry march into their 4 Cities that are defended by 3 Spearmen each and a human wont even bother attacking unless they have 'superior' tech or at least 'equal' forces*... duh.

the information you've given actually fails to reject the null hypothesis...

According to you... who is desperately inventing reasons not to provide evidence.

(and how many hitpoints did they have? and how many hitpoints did your units have?).

I never field a Unit with less than Veteran Staus and, as you well know, Ancient Cavalry *start* with a bonus Hit Point. And, yes, I had 'Elites'. Them? The Usual two Veterans and one Regular, what else would they have, you've played loads of games, why don't you know this off by heart? Sorry dear, but no matter how you spin it, it's still a stat mockery, which, ergo, suggests another programme is at work...

On the other hand, at least you are not complaining about cheating.

Ah now, how about this Civ classic: I watched an enemy Civ Unit try to 'escape' a Settler/Spear Unit while I was invading them. Because I had the whole continent, they had nowhere to go and they were literally walking past my Fortified Units for the whole journey. Eventually it stopped and made a city. Upon my turn I attacked it. In the Space of half a turn (or what we call the 'interturn') this new town had 'miraculously' amassed *two* Spearmen. This can be referred to as 'cheating' because the human player does not have access to this kind of immediate purchase, this is a trick only the AI is allowed to pull. Further more, how did it 'pay' for this Unit? I could have sworn my Diplomacy screen showed him to have 0 Gold on the turn he produced the city... Just more of that Programmed Irritation. I suppose some might call it cheating though

The AI comes crashing in with more military than they can handle.

Did they refuse one of the AI's Irritating demands? Or did they cave in? When you refuse an AI demand it will do a risk/aggression/random/Irritating calculation.

"you must be reading different forum postings than I am."

If there are varied forum posts, it proves that neither strength nor weakness are the sole cause. In one of my 20k games it was a Large Continents game and there were 5 other AI civs on my continent none of which even looked like attacking me even though my Military Advisor told me 'Compared to these guys we have a weak military' for the entire game all the way to 1820AD and tanks aplenty. Maybe it's because most of them were in 'Gracious' states of Diplomacy...

 

[Cyc]

I have made no boasts, no claims at all about my level of skill or length of experience. A boast would be a comment such as "Well, I for one [do not have limited experience]." It is a rather silly boast, I think, given that 10 years is obviously not an unlimited amount of time.

Well, I'm done. Paper must describe your depth, PaperBeetle. I gave an honest answer to your question, yet you debase my answer because you feel you are so knowledgeable about universal standards. I bow to your greatness.

 

[Spoonwood]

You haven't shown that programming here has caused anything, nor does there exist enough information to suggest that programming has done anything with respect to your irritation here.
And you haven't shown anything to disprove it!

I didn't claim I did, and I don't need to do so. One doesn't need to disprove that there exists a teacup orbiting Jupiter, or disprove the claim that Mitt Romney is a unicorn. You do NOT act rationally when you make such a comment as this.

Um, so what? The point scoring isn't huge there for a standard monarch map (if that's what you played), and plenty of XOTM games played at monarch level finish well before that.
This post says a huge amount about your fractured relationship with a lot of posters

No, it doesn't say anything about my relationship with other posters one way or another. And if you haven't checked, I have two threads here with over 20,000 views. Do you remember the responses to my posting about your game, and how many people agreed with me there? On top of this, I once got contacted about a Polycast Focus on Civilization III.

what I wrote came with respect to the battle you fought with swords, ancient cavalry, and horses, NOT some sequence of battles with riders. So, what you've referred to here comes as irrelevant.
Ooooooooooh, reeeeealy. Ok mr assumptive pooh-pooh, tell me, your entire army consists of 5 Units, two Ancient Cavalry, Two Swordsmen and a Horseman, now, which ones do you replace with Catapults in order to definitely effect a better, quicker result against three Spearmen?

You do NOT replace any of them. You take catapults along with those units, or get more ground units total to fight such a battle.

No, that's NOT a (relevant) dataset, because experience relies on memory and selective use of information.
Well deary, the reason I and everyone else doesn't 'remember' having weaker Units defeat stronger Units is because *a human wouldn't make the decision to do that in the first place, the human quits the game when 30 Cavalry march into their 4 Cities that are defended by 3 Spearmen each and a human wont even bother attacking unless they have 'superior' tech or at least 'equal' forces*... duh.

Wrong on both counts. I have played to the bitter end before, and I've seen some other players do it, generally from much older threads. Also, if, for example, you look at the screenshots here on page 11 on, you can see someone fighting nukes, mechanized infantry, modern armor, stealth bombers and fighters, among other things with artillery proper, cavalry, and infantry.

You could easily discount anything in your experience which doesn't fit your pet hypothesis, only remember the things which fit with your pet hypothesis, and only report the things which fit with your pet hypothesis.
Which is precisely what you are doing with your responses.

No, because the null hypothesis is NOT a pet hypothesis of mine. The null hypothesis comes as the default hypothesis for the purposes of discussion.

You need someway to systematically record everything that happens, or have a truly random and large enough sample.
Yes, I know, I've already agreed this is a good idea!

Then get to work.

So far as I can tell, you don't have either here.
And neither do you!

I didn't claim to have anything which goes against the null hypothesis, so I didn't need either. On the other hand, you did have something which goes against the null hypothesis, so you did need something here.

I provided an analysis of the defense values of the spears (3.7 at least) you attacked
And yet three Horsemen can breeze these stats suggesting the stats mean didly squat, and YOU seem to Forget I had Five Units, so each unit only needed to do 60% percent damage each, whereas the three Horseman, by you own stats, were having to do persistent 180% damage EACH. Yes, those stats really do mean squat don't they.

Yes, the three horseman mean diddly squat, because that ONLY consists of three battles. It comes as expected that rather unusual events will happen at some time if you have enough trials... like people getting two consecutive SGLS. A unit can't always do 60% damage, so that number doesn't come as relevant. Sometimes a unit can only do damage in 1/3 increments, sometimes in 1/2 increments, 1/4 increments, and 1/5 increments. And no, the horseman had to do 100% damage each. And I repeat, those stats don't mean all that much, because you have far too small of a sample there.

the information you've given actually fails to reject the null hypothesis...
According to you... who is desperately inventing reasons not to provide evidence.

You claim to have a hypothesis which works differently than the null hypothesis, then you need to provide the evidence. Not only does you claim lie outside the ordinary understanding of veterans of the game like VMXA and myself, thus requires extraordinary evidence, you also make the positive claim of a relationship here. So, you need to provide the evidence.

I never field a Unit with less than Veteran Staus and, as you well know, Ancient Cavalry *start* with a bonus Hit Point. And, yes, I had 'Elites'. Them? The Usual two Veterans and one Regular, what else would they have, you've played loads of games, why don't you know this off by heart? Sorry dear, but no matter how you spin it, it's still a stat mockery, which, ergo, suggests another programme is at work

No, I don't know that by heart, and I don't think that comes out always in the same way. I rarely play with barbarians, though apparently you do play with them sometimes, and that can affect things.

And no, that the spears had a defense of 3.7 at least is NOT spin. If you use the combat calculator here, a 4 hp sword vs. a 3 hp fortified spear in a city (this means a size 1 capital... a size 7 capital counts as a metropolis) on flatland has a 55.4% probability of winning. If such a spear has 4 hp, then the sword has a 38.7% of winning. A 5 hp 3 attack unit vs. a 4 hp spear of this type (fortified and such) has a 51.8% probability of winning. How many swords and how many ancient cavalry with how many hitpoints exactly? What was your battle order? Again, there exists NO stat mockery here, because you have 5 units with attack values less than the defense values of the units you attacked. That your units had an attack value less than the effective defense values of the units you attacked, even if you did not attack a capital on a hill, even if you did not attack across a river, and even if you did not a size 7 or above capital, comes as a fact.

Unless you have overwhelming hitpoints (via sufficiently more attackers than defenders, healthy many hitpoint armies, or sufficiently weakened defenses due to artillery fire), even if the odds favor you slightly due to greater hitpoints on your end, if you attack a city with defenders with defense values higher than that of your attackers (there exists no question that the defenders had effective defense values higher than your attackers), they don't favor you enough to warrant the sort of inference concerning a "programmed irritation" that you've tried to make.

 

[MysteryX]

It seems to me that AI war decisions are influenced by relative strength and attitude. They tend to bully the weak, and they tend to target those with whom they are furious. The tendencies, however, are not absolute. I suspect that on any given turn, the chance that an AI player will start a new war against any other player is always something above zero percent.

The tendencies can be successfully manipulated. For me, especially early in the game, I am determined to turn the first war in which I am involved into a world war. I have noted that ones two AIs have done enough damage against each other to get to a fairly permanent furious state, they keep starting wars again each other over and over throughout the game.

I tend not to build a strong military early in the game. Those who do, however, have reported on these boards that it not only seems to deter attack from AI players, but they have figured out that units with higher attack values do a better job of deterrence.

I don't know if, when measuring relative military strength, an AI player considers the strength of its allies though mutual protection pacts. One thing I have noticed is that an AI player will often enter into a mutual protection pact with another AI player right before launching an attack against me. When I see two AIs form a mutual protection pact, I do brace myself for trouble.

I have noticed in some of my games, late in the games when I am doing well, AIs start to drift to negative attitudes toward me even though I am doing nothing to provoke them. I assume that, similar to Master of Orion (a very similar game), AI attitudes go negative toward whatever player seems to be winning the game. So it's possible that when the human player (or an AI player) starts doing really well, a "jealousy factor" kicks in with the AI to make them more aggressive despite not having the strongest military.

I don't know how the math for these factors works. I do have a pretty good sense of who may attack me when. AI players who have been Polite toward me have started wars with me- early in games when I have been militarily weak. AI players with weaker military have started wars with me- later in games when they are furious with me for some reason. I don't recall a polite, weaker AI ever declaring a war against me (though I am sure there is some small percentage chance that it could happen).

 

[Buttercup]

*sighs*

you can see someone fighting nukes, mechanized infantry, modern armor, stealth bombers and fighters, among other things with artillery proper, cavalry, and infantry.

That's a tad different to fending off 4 towns with Spearmen against 30 Cavalry. As per usual, I give an example and you give a twisted response.

You do NOT replace any of them. You take catapults along with those units, or get more ground units total to fight such a battle.

Ooooooh, right, so you delay victory and score less points, ok, gotcha. Good plan mr 'it's all about the points'.

No, because the null hypothesis is NOT a pet hypothesis of mine. The null hypothesis comes as the default hypothesis for the purposes of discussion.

So you're just nay-saying without proof. That's nice, please feel free to write 'what a load of crap' after every single post on the thread, that will be really helpful. In fact, why don't you also do it with skewed, twisted and outrageously arrogant and assumptive posts, just to hammer home the fact that you don't have to prove anything.

Then get to work.

I have been, virtually every one of my posts carries examples, what a shame you prefer to piss in the wind over just one of these examples which proves nothing either way.

A unit can't always do 60% damage

Excuse me? You have knowledge about what damage Units are capable of? Are you telling me you actually KNOW how the Random Number Generator works? If there even is one? Well now, looks to me like your basing YOUR replies on an UNPROVEN NULL HYPOTHESIS. Understand?

Not only does you claim lie outside the ordinary understanding of veterans of the game like VMXA and myself, thus requires extraordinary evidence, you also make the positive claim of a relationship here. So, you need to provide the evidence.

Again, you talk about knowledge of the RNG with the confidence of someone who knows EXACTLY how it works, yet give no proof in your ceaseless ramblings. Tell me, do you know how the RNG works? Or are you just going to give like-for-like 'experience' evidence as well?

Unless you have overwhelming hitpoints

And here you are getting in all kind of argumentative loops of hell confusion. On the one hand you're saying I'm dumb because I didn't attack with VASTLY OVERWHELMING forces (even though I did) while at the same time saying I'm dumb because I suggest no human player would bother attacking unless they had OVERWHELMING forces. What a card you are.

Either the stats matter or they don't, you suggest the three Horsemen to be an 'expected' outrageous occasional result, but then suggest it's possible for Cavalry to take out Mechanised Infantry like it's 'more than possible and something people should 'bother' to do', which, by absolute default, implies the RNG, if there is one, means didly squat and has no relevance to the point structure, and yet my five superior forces were 'expected' to fail because they had a vastly better statistical chance of winning. What a card you are.

Wow, I hope someone 'interesting' posts soon, these circles you force us to run in take a hell of a long time to unravel and reply to and I'm not getting any younger...

Don't tell me, you're going to reply with something along the lines of "Oh, no I don't" so I can post another "Oh yes you do" to which you can reply an "Oh no I don't" to which I reply "Oh yes you do" to which you reply... ah heck, you get the picture. Have you ever watched a classic pantomime?

 

[Buttercup]

MysterX: Spot on! I genuinely couldn't have written that post better myself

 

[Spoonwood]

you can see someone fighting nukes, mechanized infantry, modern armor, stealth bombers and fighters, among other things with artillery proper, cavalry, and infantry.
That's a tad different to fending off 4 towns with Spearmen against 30 Cavalry. As per usual, I give an example and you give a twisted response.

You said

Well deary, the reason I and everyone else doesn't 'remember' having weaker Units defeat stronger Units is because *a human wouldn't make the decision to do that in the first place

. People sometimes do make the decision to have weaker units fight stronger ones.

You do NOT replace any of them. You take catapults along with those units, or get more ground units total to fight such a battle.
Ooooooh, right, so you delay victory and score less points, ok, gotcha. Good plan mr 'it's all about the points'.

No, you do NOT delay your victory this way, because you go through fewer losses, and the AI doesn't produce as many units that you need to defeat. You also pick up promotions and great leaders easier, since you have fewer losses. You have to think in terms of the wars, not just battles. And honestly, if you think your strategies better, then take part in a competition like the HoF or one of the XOTM competitions.

No, because the null hypothesis is NOT a pet hypothesis of mine. The null hypothesis comes as the default hypothesis for the purposes of discussion.
So you're just nay-saying without proof.

No. I actually ended up proving that the example you've given doesn't work out as well as you made it seem. I didn't prove the null hypothesis, but I did prove that your example doesn't really work in your favor.

Then get to work.
I have been, virtually every one of my posts carries examples, what a shame you prefer to piss in the wind over just one of these examples which proves nothing either way.

I doubt it. You've spent plenty of time responding to my posts here and others. And where lies your large data set? What method do you have for making sure that you have systematically recorded information, and not selected information which only favors your pet hypotheses?

A unit can't always do 60% damage
Excuse me? You have knowledge about what damage Units are capable of? Are you telling me you actually KNOW how the Random Number Generator works? If there even is one? Well now, looks to me like your basing YOUR replies on an UNPROVEN NULL HYPOTHESIS. Understand?

No, I don't claim to know how the pseudo-random number generator works in detail. Battles work according to hitpoints. AFAIK, for each hitpoint either the attacking unit wounds the defensive unit, or the defensive unit wounds the attacking unit. Say you have a 4/4 sword attacking a 4/4 spear. Suppose the sword wins the first sub-round. So, the spear goes to 3/4. Then say the spear wins. So, the sword goes to 3/4. And so on until one of the units goes to 0/x in general. AFAIK, and here lies the key point, there exists NO fractional values of hitpoints. We don't have a 2.5/4 spear, or a 2.1/4 spear. So, excluding what numbers we need for armies, we can only do 1/2, 2/2, 1/3, 2/3, 3/3, 1/4, 2/4, 3/4, 4/4, 1/5, 2/5, 3/5, 4/5, 5/5, 1/6, 2/6, 3/6, 4/6, 5/6, or 6/6 damage. Thus, if a spear starts at 4/4, we can only do 25%, 50%, 75%, or 100% damage to the spear in a battle. We can't do 60% damage, since none of {1/4, 2/4, 3/4, 4/4} equal 60%. That there consists of my point here.

Not only does you claim lie outside the ordinary understanding of veterans of the game like VMXA and myself, thus requires extraordinary evidence, you also make the positive claim of a relationship here. So, you need to provide the evidence.
Again, you talk about knowledge of the RNG with the confidence of someone who knows EXACTLY how it works, yet give no proof in your ceaseless ramblings. Tell me, do you know how the RNG works? Or are you just going to give like-for-like 'experience' evidence as well?

No, I haven't spoken with confidence to know exactly how it works. If I had, I would have made a claim that I could predict the exact result of a particular battle with confidence beyond that of a probabilistic guess. What I've written in general concerns sets of battles, which works out far differently. I don't know how this pRNG got programmed, but see here. And no, I've not just provided experience evidence... did you see where I referenced probabilities using a battle calculator?

On the one hand you're saying I'm dumb because I didn't attack with VASTLY OVERWHELMING forces (even though I did)

You did NOT attack with vastly overwhelming forces. Your biggest problem here comes as that you fail to recognize that you attacked with only marginally better forces, if indeed you did attack with better forces (it comes as hard to tell without knowing the exact hitpoint values with which you attacked, and without knowing if you attacked a river, the capital lay on a hill, and what size the capital stood at). You four 3 attack units and a 2 attack unit vs. three units with a minimum of 3.7 defense. As far as I can tell, it doesn't qualify as all too meaningful of a statistic, but you if we combine all the attack values your attack units come out to 14, while the defense units come out to 11.1. That doesn't come as much of a difference!

Again, go here, or find another battle calculator and find out actual probabilities of winning battles once you factor in all defense bonuses for units, and you factor in hitpoints also. Your best battle probably came as a 6/6 ancient cavalry vs. a 4/4 spear. If that spear had a defense value of 3.7 (if the capital sat at size 7 on flatland, and the spear was fortified, the spear has a defense value of 4.7 since the capital has defense bonuses of a metropolis), which may not hold true... I may slant the numbers more here in your favor by assuming the capital below size 7, your ancient cavalry had only a 63.4% chance of winning. That comes as your best battle. Plenty of them don't even hit 50%, and you only had 5 units taking on 3 defenders.

while at the same time saying I'm dumb because I suggest no human player would bother attacking unless they had OVERWHELMING forces.

There exists no conflict between the first part of your statement there and the second part.

Either the stats matter or they don't, you suggest the three Horsemen to be an 'expected' outrageous occasional result, but then suggest it's possible for Cavalry to take out Mechanised Infantry like it's 'more than possible and something people should 'bother' to do', which, by absolute default, implies the RNG, if there is one, means didly squat and has no relevance to the point structure, and yet my five superior forces were 'expected' to fail because they had a vastly better statistical chance of winning.

I did NOT suggest that people fight cavalry against mechanized infantry. I pointed to information which indicates that such has happened. That consists of a description, not a prescription.

And NO, so far as I can tell from the limited amount of information you've provided, your five units did NOT have a vastly better statistical chance of winning. Did you check probabilities on a combat calculator?

And NO each individual unit did NOT come as superior. A horseman with the same number of hitpoints as the spear, is NEVER better than such a spear anywhere, since the spear always has a terrain bonus. A sword, with the same number of hitpoints as the spear, attacking such a spear in a capital is NEVER better than such a spear, even if the spear is not fortified. An unfortified spear in a capital on flatland, due to the minimal 50% bonus a size 1 capital always has, and minimal 10% flatland bonus, has a defense of 3.2. You almost surely attacked fortified spears, making them defend at least a value of 3.7. So, you only had superiority in terms of an individual battle if you had units which had more hitpoints than the units you attacked. If we keep that in mind it doesn't come as all too surprising that you lost 4 units the first turn, and the AI only lost 2 units.

That the spears you attacked, given all of them fortified which almost always happens with defending units in any city, had a defense of 3.7 is NOT an opinion. I see no indication that you have acknowledged it.

 

[Buttercup]

Yes spoonwood, that probability calculator is fascinating, it gave this result:

Probability of a Horseman Defeating a Spearman Fortified in a City:

Attacker wins 20.1%
1-0 8.3%
2-0 6.4%
3-0 3.9%
4-0 1.5%

What a shame it can't calculate the probability of three Horsemen IN A ROW defeating three Spearmen, WITHOUT BARELY A SCRATCH.

But, of course, with a 'probability' of greater than 0.00000001% then it is, for you, an 'expected unusual result'.

Here, try my link:

http://www.youtube.com/watch?v=GUGt1lbXaxk

 

[Spoonwood]

Yes spoonwood, that probability calculator is fascinating, it gave this result:

Probability of a Horseman Defeating a Spearman Fortified in a City:

Attacker wins 20.1%
1-0 8.3%
2-0 6.4%
3-0 3.9%
4-0 1.5%

What a shame it can't calculate the probability of three Horsemen IN A ROW defeating three Spearmen, WITHOUT BARELY A SCRATCH.

But, of course, with a 'probability' of greater than 0.00000001% then it is, for you, an 'expected unusual result'.

Here, try my link:

http://www.youtube.com/watch?v=GUGt1lbXaxk

You haven't specified how many hitpoints the horseman has, nor how many hitpoints the spearman has. We also don't know what the city sits on, or its size, or if it has walls, etc. so we don't know the exact defense value of the spearman. From the numbers you've given, I'll guess that you point in a 4 hp horseman not crossing a river vs. a 4 hp fortified spear on grassland in a city.

Calculating the probability of success of three consecutive horseman attacks against three spears comes as easy to compute, given that you understand the multiplicative rule for independent events of probability theory. The rule says that P(a^b)=(a*b) where "^" indicates the intersection of two events (in other words, both events occuring), and "*" indicates ordinary multiplication. So, the probability of three consecutive victories here equals ((.201*.201)*.201)=.00812... or about .812%. We can also figure out other possibilities such as the probability of three consecutive victories of the horseman without a single scratch which, ignoring the problem of significant digits, equals about .00000338 (4-0 indicates the horseman doesn't lose any hitpoints), or the probability of the first victory with only a single hitpoint lost, and then no hitpoints lost for the last battle.

In other words, if we have the probability of winning for each battle, computing the probability for a set of such battles isn't a problem using the multiplication rule for probabilities.

And no, I haven't made any claim about specific probabilities. You waste time by trying to put words into my mouth, as you don't even have credibility with basically anyone around here.

 

[Buttercup]

So what you're saying is that the chance of the encounter what I did encounter (and have experienced enough to know it's not an isolated incidence) is that the probability of the result is somewhere between:

0.812% and 0.000338%

depending on how many Hit Points was lost by the one Horseman which lost Hit Points.

?

 

[MysteryX]

If you play enough games, something that happens to you 0.8 % of the time (roughly 1 in 125) is going to happen to you quite a lot.

Has anyone actually figured out with certainty how the combat calculations work? I've seen several "combat calculator" utilities, but even with those, there was no absolute certainty they had the math right. They were created just by people running a lot of different battles in different situations and doing there best to pull out what they could from the data based on known factors.

 

[PaperBeetle]

http://www.civfanatics.com/civ3/infocenter#military

Welcome to the Civilization III Info Center! The purpose of this Info Center is to index ALL known Civ3 information to make it easier for you to find the info you need on any aspect of the game. The info in the Info Center are gathered from numerous previews, reviews, interviews, articles, strategy guide, manual, and forum posts, both official and unofficial.

Military
Combat:
Each combat round has a chance of inflicting a single hit-point loss on a unit, which is simply a probability based on the relative attack and defense strengths of the combatants. If a unit has an attack factor of nine and attacks a unit with a defense factor of one, it has a 90 percent chance of inflicting a hit-point loss on it during that round.

 

[Raliuven]

I've thought about how to test this. Just for fun (yeah, for fun) I tasked excel with making 1 million coin flips. The bell shaped curve was no surprise but I was somewhat surprised by the results. I had expected a more gradual bell curve but instead what I saw was the expected clump near the 50/50 split and then it quickly tappers off. The most heads flipped in a set of 100 (that would be 10,000 sets) was 71 heads (to 29 tails) and the least was 30 heads (to 70 tails). That means that a deviation of more than 20% from the average is rather rare.

If the RNG is a straight statisitical result it should have a comperable deviation, right?

I think that if a set of 100 attempts was made in a test game as PaperBeetle suggested and the results deviate by more than 20% from the combat calculator, then there might be something to Buttercup's hypothesis (though extensive testing would still be needed to quantify it including many 'sets' to establish a solid pattern). In addition, I think it would also need to be determined if the combat calculators are reliably flawed (predicatable) or if the RNG is not playing fair. The problem is that this might be extremely hard to quantify (requiring many, many sets to establish either a shift of the bell curve or true errors).

In the above example a run of three is not conclusive; it is an extremely limited dataset. Even with the limited dataset the .8% result is within the 20% deviation (20.1% to 0.1%) though it would admittedly be a rare event. Of course we don't have all of the information so it hard to say the actually expected and observed results.

My statistics are rusty, but would that make sense?

If there is a gross error it may be immediately observable. If it is more elusive it could be nearly impossible to detect. On top of the fact that Vorlon_mi says the customization people already beat this horse to death so it may not be worth the time.

 

[Spoonwood]

So what you're saying is that the chance of the encounter what I did encounter (and have experienced enough to know it's not an isolated incidence) is that the probability of the result is somewhere between:

0.812% and 0.000338%

depending on how many Hit Points was lost by the one Horseman which lost Hit Points.

?

No, I did not say that.