Report Questionable Behavior

[Chris Withers]

A country is at war with me, sends one medium stack of attackers which I waste, and then only sends 1-2 more attackers soon after. I build up a bit more and then mass to take a border city. Seems like if they weren't going to prosecute the war then they ought to have sued for peace (not that I'd have taken it). Then after taking that city one would think they should sue for peace, but again nothing is heard from the idiot.

Also, they quickly pushed 2 ships up my coast and sat them outside my closest coastal city to them. Not much effect if the coastal spot is not a resource. A country ought to continue up the coast until it finds a resource spot. Or since it sees my coastal border city poorly defended maybe they should put greater weight to maybe building a galley and doing a coastal invasion.

Also, I agree with Yakk 2 posts up. The AI is stupid, stupid, stupid when it comes to manipulation of forces. They usually have more units (I'm at Monarch), but use them so poorly that I can always beat them in battles & wars.

 

[Roland Johansen]

Also, they quickly pushed 2 ships up my coast and sat them outside my closest coastal city to them. Not much effect if the coastal spot is not a resource. A country ought to continue up the coast until it finds a resource spot. Or since it sees my coastal border city poorly defended maybe they should put greater weight to maybe building a galley and doing a coastal invasion.

If the ships happened to be caravels, then that is their optimal use. They can't transport real units and can't pillage, so the only thing they can do is kill your naval vessels and block the use of your coastal tiles. If the ships weren't caravels and they knew you had coastal resources, then it was stupid.

 

[Wodan]

well i guess 3) is the good answer where should russian have grabbed land south ? theyre surrounded ^^
plus they claim the second iron and thats a good point.
anyway i dont say i would have build a city there, i just mean this is not so stupid

One tile EAST, not south. The iron is irrelevant.

Wodan

 

[kniteowl]

How about a on topic post.

Here's a screenshot of my current game. It shows how the BetterAI is still not quite there, in regard to city placement. I've reported this before. I don't know if the BetterAI team intends to tweak this some more before releasing 1.0, but I hope so.

Anyway, take a look at where Stalin put StPete. A total of 6 tiles of overlap, and he gave up fresh water. All he gained was a shot at an extra spice. (The spice is disputed with Thebes.)

In addition, he makes impossible to work several tiles. (not resource tiles, normal tiles)

My guess is that the resources are still driving a huge modifier. The AI builds cities to get resources. Everything else is secondary. (I don't think that's the best way to have them do it.)

Wodan

Personally If I were Russia, I'd settle on the coal hill 1 tile SE of it's current location and grab all the current resources plus the cows, have zero overlap with the capital and still get the fresh water bonus and is still a coastal city.

I'd get a work boat to the clams immediately and work it for growth and whip a monument. I Would Settle on the Coal because Coal has not been revealed by research yet as I Assume your starting the game from the Ancient Age.

I assuming that tile is Coal, It's hard to tell sometimes when the resource button isn't on lol.

 

[Khamul]

The AIs attack tactics are questionable. A stack should know if it has a decent chance to take a city before it actually tries -- and I've had an attack stack that was cleary insufficient to kill more than 1 or 2 units attack a city over a river rather than move around it.

Next, the AI doesn't seem to want to sally that much. I can often move a mediocre small stack near an enemy city, and the AI doesn't smash it -- even if it has a decent chance of being able to kill 2 units and only end up with at most 1 unit dead (and quite possibly none).

Same observations here. Last week my spy watched AI vs. AI wars. First there was Peter vs. Hyuana, and then Churchill vs. Hyuana. Peter and Churh had a tech lead compared to incas.

I think there was two problems. First it started well: Peter quickly captured 2 cities, but then Peter's stack stopped. The second captured city was large, but was totally surrounded by incan culture, so the city was useless. Peter's stack had over 20 units, mainly artillery, infantry and cossacks but also a few tanks (versus incan riflemen, cavalry and cannons). Definitely they should have taken the next (80% culture) city. The war finally ended, and Peter had one useless city inside incan cultural border. Not very smart, I think.

Second problem was wandering alone units, mainly artillery. Incan twice-flanked cavalry always destroyed those wandering units without real casualties, so Peter lost several artillery for nothing. Perhaps those units were pillaging, but still it was stupid.

Later Churchill too had alone unit problem. First Churchill too captured 2 cities. Then Churchill's main stack stopped, but several alone infantry tried to attack 80% city! I think Churchill lost about 10 units, and Hyuana lost none. After that engish stack easily took that city, but they suffered lot of unnecessary casualties.

I have also noticed that AI doesn't always counter-attack my small stack, even if they have decent number good units to do so.

 

[Nor Me]

Look at this Ireland city. It was founded in 300BC and now it's 2015AD.

Firstly, there's no worker. So England has no pigs. I know now that a major reason for this is all those 2 food, 3 commerce coastal tiles. It tries to be big enough to work all 9 of them. But without the culture slider and those happy buildings, the 9th citizen would be unhappy and the city would be starving at size 9. The governor is now sensible enough not to do that. Even without this, island cities are reluctant to build workers. For some reason, sometimes they build more when they've already got one.

So it needed buildings. This seems to be a problem for low production cities in general. I suppose that Hastings has the right buildings apart from the barracks and there total cost does represent a reasonable portion of the city's 2-production-for-2300 years output. But a whipped forge, library and university would have paid off quickly. With 55 base commerce, a bought library or university would pay off in 30 odd turns, assuming 1 gold = 1 beaker.

Low production cities tend to build only very cheap buildings because others take too long. Instead, they build units. Ironically, they rarely finish those either because they like to switch production every time a new tech is discovered. Spending 45 turns on a forge or library would obviously be better.

 

[Wodan]

Personally If I were Russia, I'd settle on the coal hill 1 tile SE of it's current location and grab all the current resources plus the cows, have zero overlap with the capital and still get the fresh water bonus and is still a coastal city.

I'd get a work boat to the clams immediately and work it for growth and whip a monument. I Would Settle on the Coal because Coal has not been revealed by research yet as I Assume your starting the game from the Ancient Age.

I assuming that tile is Coal, It's hard to tell sometimes when the resource button isn't on lol.

Sorry, yes that's coal.

Well, perhaps. as a human I would either settle on the hill (1SE) or the tile 1E. The hill is better defense, better cultural push possibilities. However, given the damn Greek settlements, I plan to go to war soon to get Thebes. Given that I'm facing Alex and also that he's cornered me in, I'll be at war with him all game, do or die. So, I would probably intelligently plan on having all his cities, and the hill is dubious, at best. I'll be on the offensive, so the defensive possibilities are null.

As an AI, I would prefer to be further away from the greek borders, not closer to it. If I have a premeditated war flag set, then see above... I'm going to war, and will plan to own all his cities soon. (Otherwise, I'll be dead, and it won't matter.)

If I don't have a war code set, then the hill might be good. Might. I could prioritize temples and theatre, to push borders. And, I could build walls on the hill, which will be all but impregnible until cats show up. That's about the best case I can build for the 1SE spot.

In any event, the spot he picked was suboptimal from any angle I can figure.

I've been watching this closely and my whole conclusion after looking at hundreds of AI city placements in multiuple games is that the current BetterAI has a predilection for 3-away cities rather than 4-away. I think that should be reversed.

Wodan

 

[ese-aSH]

Sorry, yes that's coal.

Wodan

hm i always wondered, do AI know where the ressources are before they reveal them ?
that could explain things

 

[Roland Johansen]

hm i always wondered, do AI know where the ressources are before they reveal them ?
that could explain things

In the original game they knew about the tile bonus of the resource when deciding to construct a city (they didn't actually know about the resource itself). This was fixed in an early patch.

 

[Yakk]

Ya -- a single worker (ideally ferried to the city) would be able to generate:

+3 food
+2 production (lumber mill)
+6 production -2 trade (water wheel with railroad -- those are 3 each IIRC)
+3 production -1 food -3 trade (railroad workshop) or +5 trade +1 hammer (town)

So an overall change of:
-5 trade +11 production +2 food

which changes the city from:
+3 food 20 commerce 2 productin
to
+5 food 15 commerce 13 production
or
+6 food 20 commerce 11 production

a ridiculously massive improvement in city quality.

...

This makes me think that there needs to be a "city needs workers" and "empire needs workers" subengine whose job is to figure out where workers would do the most good, then figure out a way to get them there (either by getting a connected city to build them, or using a transport to carry them there, or airlifts).

 

[Quagga]

Regarding the build governor:

1. It is too lilkely to build workers, especially during conflict. It seems as though it builds them because there are improvements not getting done, but often this is because the area in question is a war zone and workers won't go there.
2. The priority of jails is too low.
3. Although barracks are important, it builds them when I feel it should be concentrating more on building cultural and happiness buildings. I typically don't use newly captured cities to build units, so building barracks is not a priority for such cities.
4. In the last game I played, it built a workboat in a city that was blocked in by ice. The workboat was stuck.

 

[Quagga]

I recently finished a Monarch/Epic/Standard/Pangea game with Aggressive AI turned on, playing as Hashep. I'm happy to report that I won with a Domination Victory, though that outcome was extremely doubtful for a time.

What threw it into doubt was, after gathering my second vassal, the other big power in the game, Cyrus, and his vassal, Peter, declared on me. Cyrus sent a stack of 32 SAM Inf, 15 Cavalry, a bunch of Infantry (can't remember how many), and a few Artillery. Something on the order of 60 units total. By all rights, I should have been wiped out.

Fortunately, I had a stock of Artillery (all with CR promotions) and a few unpromoted Tanks. I also had better production, which I threw 100% into more Tanks and Artillery. I gave them all Barrage and kept inflicting collateral damage on this huge stack. Also, the Tanks usually killed the top unit.

So what's the questionable behavior? The attacking stack stopped in its tracks. I guess it decided it needed to heal all that collateral damage. It did slowly advance, but only as it was able to move forward and kill my wounded counter-forces. Mostly it just sat there. Over maybe twenty turns, I reduced the stack to nothing.

If the stack had continued on to my nearest city, it could have taken it, despite the collateral damage. I'm sure it could have taken a few cities with that stack.

If the AI had fanned out the stack, I couldn't have stopped it. Just pillaging with the 15 Cavalry would have been impossible for me to stop.

Also, why does the AI produce so many SAM Infantry? Nobody even had Flight.

 

[Uncle_Joe]

My guess on the SAM infantry is the same as for the Spearmen...the units has a BIG bonus against something and that makes it more attractive to the AI's algorithms.

I also agree that the AI stacks 'stop' too often and heal even when pushing on would topple the nearest city with ease. Even the unhurt units tend to just sit in the same stack while the most damaged of them heals.

 

[Yakk]

And you can really exploit it with aircraft. If they don't have SAM yet, just send bombers at nearby menacing stacks of AI units. The AI will choose not to attack ground forces that they could possibly overwealm, and instead pull back and/or heal up.

Spearmen and SAM infantry are in many ways amoung the worst units to spam build. Both are good at fighting units that aren't all that good at defending, which means they are both defensive units.

Large numbers of defensive units don't win wars against smart opponents. Offence can be concentrated -- so even if your units are each twice as strong as your opponents attack units, the opponent can concentrate a relatively small number of offensive units and smash your defences in a small area. (overwealm spearmen with concentrated cavalry attacks)

And the opponent can change what they are building to defeat your defensive units. If you turn out to correctly build the right defensive units, all that happens it the opponent fails to conquor you in the short term -- but if you pick the right offensive strategy to overwealm them, you get to conquor and pillage them!

Hence: defensive units should be built in a mix, biased towards an educated guess what kinds of units your opponent will throw against you (ie, instead of 33/33/33, go 50/25/25). It makes sense for the AI to build a stack containing SAM infantry, but it should contain a mix of SAM Infantry (anti-air), Infantry (anti-gunpowder), Marines (anti-armor) and (if in the field) Choppers (anti-armor).

 

[Roland Johansen]

I also agree that the AI stacks 'stop' too often and heal even when pushing on would topple the nearest city with ease. Even the unhurt units tend to just sit in the same stack while the most damaged of them heals.

Now, you wouldn't want the healthy units to move on and leave the wounded units to be slaughtered?

It's actually a very hard decision when the stack needs to move on and try to conquer something and when it should stop and heal. It depends on the strength of the opponents army and the amount of units in the stack and the degree in which the stack got wounded. The strength of the opponents army is often an unknown factor making the decision even more complicated.

If you have a stack of 30 units of which 20 just got wounded by collateral damage attacks (some of them seriously wounded, some of them barely wounded), should you move on? And which units should stop to heal and which ones should move on? And if you decide to move on with only part of the stack, then how many units should you leave with the wounded units to protect them? If you only have one healing unit in the stack, should it stay with the wounded units or move on with the attacking part of the stack? Should you retreat the stack to neutral/allied territory to heal?

All of these questions need to be answered and are hard to answer without exact data on the opponent. A human mind can make an educated guess on the opponents strength, but guessing is impossible for the AI. You have to write some code that makes the AI decide on the above issues.

If you think you could write some pseudo-code to make these decisions, then do so. I think it's very hard and whatever code you'll write, the AI will make the wrong decision sometimes. I hope that Blake and Iustus can improve this AI decision making, but it won't be easy.

I would suggest to just keep the stack as one and let the decision to move on or heal depend on the average degree of hitpoints lost in the stack and a random factor. The random factor is important because we don't want the AI to become predictable. If the human player finds out that the AI stack will stop to heal when it has lost 20% of its hitpoints, then we will learn to exploit that.

 

[Quagga]

If you have a stack of 30 units of which 20 just got wounded by collateral damage attacks (some of them seriously wounded, some of them barely wounded), should you move on? And which units should stop to heal and which ones should move on? And if you decide to move on with only part of the stack, then how many units should you leave with the wounded units to protect them? If you only have one healing unit in the stack, should it stay with the wounded units or move on with the attacking part of the stack? Should you retreat the stack to neutral/allied territory to heal?

I've realized that healing is for sissies. Heal while your artillery bombards the next city. Or don't heal. With 30 units in a stack, it's okay to take some losses. Fight, live and maybe earn a promotion and some instant healing. I haven't normally played this way, but the AI's stupidity showed me that I probably have been too timid.

 

[Uncle_Joe]

It's actually a very hard decision when the stack needs to move on and try to conquer something and when it should stop and heal. It depends on the strength of the opponents army and the amount of units in the stack and the degree in which the stack got wounded. The strength of the opponents army is often an unknown factor making the decision even more complicated.

Which is what I was saying all along back when the AIs were spamming tons of units left and right. If its not going to use the units 'correctly', then simply having the AIs build more and more units doesnt necessarily make it 'BETTER'.

I *know* its going to be hard to code proper algorithms for things like this because even as human players, most cant agree on exact what is the best way to go about certain decisions. So, its largely impossible to expect that an AI can 'decide' based on the situation.

So, thats why I am so opposed to having the AIs spamming too much military. Its still easy to beat them if you match them and all it does it slow the AIs down (and the game in general).

Note that at the moment, I think the number of units that the AIs are tending to build is just fine. The larger 'issue' at this point is one of usage.

 

[Wodan]

The best place to stop and heal is inside an enemy city.

Wodan

 

[Yakk]

The AI should have an estimate of what the enemy will be defending his or her cities with. A mixture of scouting, civilization power, and an assumption that the enemy will place much of their defence near where you are attacking should do it.

You also know how large your opponents productive base is, if the enemy is running slavery (which means food can be considered production), and the enemy's tech level.

From that, you can work out what the rate the enemy can replace dead units (and if you get fancy, a heuristic saying how expensive that replacement is), and how fast you can replace and produce dead units.

That gives you a relative unit value for both sides.

You can also work out how many of each type of enemy unit is required to destroy and/or harm your units assuming (full) or (damaged) health on your part. You know the cost of those units. You know the cost of replacing your units.

You know how long it will take to force march to the enemy city.

So you know how efficient enemy catapults are at damaging your units. You know how efficient enemy units are at attacking a damaged stack. You know roughly how many units you have to fight in order to reach your opponent, and roughly how tough the city will be when you get there, and if when you get there you can (economically) afford to fority and heal up and/or take the city while the enemy produces more units and throws them at you, vs your ability to build a new stack and invade again.

If the result says "you probably can't take the enemy city", you shouldn't be mass-stack-attacking: you should be economic raiding your opponent. Hitting your enemies roads, destroying their villages and economy, cutting the opponents empire up by severing roads, and moving around in small "expensive to kill" raiding stacks. Then if you spot an underdefended city that you can stack-attack and isolate via road cutting, go and win.

Ie, in the fuedalism era, a stack with a crossbow/pike/mace/knight/elephant (all x2 in some cases) is a sick pillaging monster (or even just xbow/pike/knight).

Many such combinations exist -- the goal being, you want the cost of taking out your stack to be higher than the cost of building the stack, and you want to destroy the enemy economy by taking out roads/mines/villages/farms.

Move and pillage. Pillage and move. Pillage and move. Pillage and move. Repeat.

 

[Roland Johansen]

(better AI, version 12 feb 2007)

1) The artillery type units that seem to be defending a city never attack my advancing city attack stacks.

I've been capturing some cities lately and each and every one of them has 1, 2 or 3 barrage upgraded cannons in them. They seem to be there to use collateral damage attacks on my attack stack. But they never do.

On the other hand, I have seen the AI use catapults, cannons and the like to attack single units walking through their lands pillaging in AI-AI wars. Of course, artillery type units are not very good at that. The normal units with combat upgrades however stay in the cities and let the units pillage as they like.

To me, this sounds like something went wrong in the code. Something like:
If attacking stack > 2, then attack with normal units, if attacking stack <= 2, then attack with catapults, instead of if attacking stack <= 2, then attack with normal units, if attacking stack > 2, then attack with catapults.

Of course, the problem is likely to be more complicated then that. But my observations are still that the AI never attacks my large (20 units or so) city attack stacks with artillery type units and does attack single units with artillery type units. It is quite weird and counterintuitive.

Has anyone been attacked by the defending artillery type units (catapults, cannons, artillery) that are present in the cities when you attack with a large mixed attack stack (say more than 15 units)?

2) In my last game (aggressive AI, huge fractal map), the Germans attacked me with a stack of 124 units: 4 marines, 55 infantry, 57 cavalry, 8 machineguns. Missing something? Yes me too. There were no artillery or cannons in their stack (and no panzers because I kept sabotaging their oil patch). It was a bit weird as I've seen the AI use plenty of artillery type units to attack other AI civilizations. The AI got the technology for artillery type units shortly before the war declaration but it did have plenty of cannons (barrage upgraded to defend the cities).
During the turn that they attack my city, they move 3 barrage upgraded artillery units towards the same city. These units are completely unprotected and won't help in the attack on the city because they are too late.

3) Units placed upon critical resources reduce the chances of sabotaging the resource with a spy. However, the AI never placed a single unit on top of their only oil resource while I pillaged it 3 or 4 times in a row. Maybe something should be coded that the AI starts protecting their critical resources with units after the first sabotage act against their civilization.


Here is the save before the large stack that attacks me. It is lightly wounded because I just bombarded it with 11 barrage III artilleries.

http://forums.civfanatics.com/uploads/31106/Battle.CivWarlordsSave

Here is a zip file of the mod that I'm using. It is betterAI + some minor interface enhancements. The mod name is still betterAI.

http://forums.civfanatics.com/uploads/31106/BetterAI.zip