CEP: Armies

[mitsho]

For me it was mostly a AI-issue. We failed at teaching them how to use defensive units which made warfare too easy. Sit-back-and-bombard-them-into-oblivion-while-vanguards-storm your walls... Or let the vanguards attack you in the open...

So the strength promotions didn't help the AI because they were using them wrong... One aim of the mod is to make the AI play the same game, I don't want them to use different units than I do (because it makes them impossible to calculate/I need to learn two systems) and buffing vanguards for the AI only just creates two different combat systems...

Can't comment on the retreat-promo. Don't find time really to play the game right now...

 

[Drakir89]

One aim of the mod is to make the AI play the same game

I don't think we have the resources to create an ai that can stand up to emperor difficulty players without handicaps no matter how hard we try. I play on emperor/immortal mostly so as far as I'm concerned I will always play vs an ai that is not playing the same game. Even in civ4 which had a much simpler combat system the human player always had the advantage in combat just from knowing when to attack and being better at adjusting their forces.

To me combat is a puzzle to solve, and I want that puzzle to be as varied and interesting as possible, and I found Thals design goals (no automatic decisions etc) much better for this than those of firaxis.

But most of all, I really liked using those fast ITC vanguards as part of my armies.

 

[Ahriman]

I don't think we have the resources to create an ai that can stand up to emperor difficulty players without handicaps no matter how hard we try.

There's a difference between giving the AI passive economy bonuses, which we can't really see and don't mind so much, and clear active bonuses like different units which feel more like cheating, or mechanics that the human can use and the AI just won't understand.

 

[mystikx21]

A useless promotion that comes on an early ranged unit UU should not be even an option for mainline units. I could see a use for scouts and maybe for subs. I don't see the point for infantry and spearmen.

I agree it is nice to have variability. But if you give the AI something new and it doesn't understand how to use it, to the point of crippling itself in combat even further, something is broken, and not worth interesting unit functions like screening and recon. Those are very limited roles. Healing and anti-mounted are also, but at least they are easy to understand rather than defensive only promotions and ITC/extra movement on infantry units.

 

[Dunkah]

Is it possible to change the name of the Vangaurd line, to either light footman, footman, Heavy footman, maceman, light spearman, heavy spearman, Skirmisher? Anything other than Vanguard?

IMO Vanguard is a position in a battle line rather than a class of troops. Depending on which tactics you use the Van can be made up out of any type of troops from light conscripts to heavy cavalry. And they can be placed anywhere in the battleline depending on your choice of tactics. The Van typically just leads the way. In a Pincer Wrap type engagement they might be in the middle, but in an Oblique type of engagement they would be on one flank or the other.

It just doesn't make much sense to use that name.

Traditionally there should be a cheap lightly armored "Conscript" spearman type of unit, a more heavily armored and better trained "Regular" type of spear/sword/mace/axeman and a "Skirmish" type of unit on the battlefield. (Bows/Scouts)

If you made the following:
"Skirmish" Would be the Scout.
"Conscript" would be a Spearman. A " Low Strength, Low Cost.
"Regular" could be the Vanguard Class called Footman or Regular Spear. Med Strength Med Cost.
"Elite" would be the Strategic Units. High Strength High Cost.

If it is possible to make the retreat promotion work based on who is beside them? As long as you have a strong unit on either flank or two weak units on both flanks you won't retreat. Take away one or the other and there is a chance you retreat based on damage sustained.

edit:
Or how about a system where units are born with the retreat promotion and have a high chance of retreating and then when they earn promotions that chance can go down.
Conscript = 66% chance of retreat (if possible)
Trained = 50% chance or retreat (if possible)
Veteran = 33% Chance of retreat (if possible)

 

[Stalker0]

Actually, I don't really understand why people want to get rid of the vanguards from GEM in the first place.

So as not to rehash old arguments, here is a recent thread where we aggressively debated vanguard units:

http://forums.civfanatics.com/showthread.php?t=501375

 

[Drakir89]

If vanguard units cripple the ai (and that cannot be solved by adjusting ai flavours for units or such) then yeah, they have to go. However, I don't really remember such behavior from when I played GEM (at least not worse than g&k) but that was a while ago so I may be remembering things better than they were.

There's a difference between giving the AI passive economy bonuses, which we can't really see and don't mind so much, and clear active bonuses like different units which feel more like cheating, or mechanics that the human can use and the AI just won't understand.

When we gave the ai free promotions in GEM to make wars more difficult I found that to be an improvement, better than tuning so ai would always have larger armies or better tech. I found the "cheat" units less immersion-breaking than obvious resource advantages.

EDIT: Thank you stalker0, didn't see that one. I'll check it out.

 

[mystikx21]

Tuning the AI also involves giving it better decision making, not just raw bonuses. It is easier to do this with a building and for "what do I build/buy" effects than for "where should I put this unit" effects.

The unit AI in VEM/GEM was often much better, but still did lots of dumb things like move units around fruitlessly in front of cities it could not take, wander off into the water and get executed, and so on.

 

[Zarggg]

I don't understand these complaints about the retreat promotion. It's optional right?
If you don't want your tank vanguards to retreat, give them another promotion instead? Am I missing something?

There's a common adage when it comes to choices in games' ability trees: "An option that is always or never taken is not truly a choice."

 

[mystikx21]

Right. If the option is useless (even harmful), why have it at all?

 

[JE_McKellar]

Just for the sake of discussion, can we imagine a unit that the AI can use as well as, or even more effectively, than a human player? What exactly trips up the AI?

For example, it seems that the AI is particularly bad at lining up units in the field, it just tend to move them individually towards a target, without the patience needed to keep them safely in formation as they transit. The suicidal jumps into the ocean seem to be due to pathfinding bottlenecks, and the pointless milling about in front of cities seems to be something similar, with the AI wasting all its time getting into position when a human player would just assault.

ITC might help the AI deal with those issues, as well as range promotions. But what is the AI actually good at? Spamming?

One thing I do notice (at least in my vanilla game) is that the AI tends to be very enthusiastic about spending most of their gold, while I at least keep a hoard big enough to upgrade my units. Maybe a line of units with low upgrade costs, or a free upgrade promotion, would help keep the AI competitive.

Finally, the main source of challenge in the game should come from the human player being outnumbered by all the other civs. Imagine if all the UUs got a boost, while that would give the Human player a moment of dominance in the game, they would also have to endure every single AI civ having their own moment of dominance.

 

[Ahriman]

Just for the sake of discussion, can we imagine a unit that the AI can use as well as, or even more effectively, than a human player?

The AI does pretty well with units like longswordsmen, which it can throw at units and cities. It can produce and use large numbers of melee units, getting Discipline and flanking bonuses.

It's not too bad with mounted units like knights, the mobility helps it attack well.

It's not too bad with long range units like artillery. In some sense it's not bad with aircraft, except that it doesn't understand AA guns or similar very well, and will suicide its aircraft into the guns.

In contrast, it's not good with most ranged units (it doesn't really understand move-then-shoot), especially move-after-attack units. It doesn't understand any kind of reconnaissance or healing role, so these all favor the human player.

 

[JohnS]

Wouldn't it make more sense if the retreat option was available to ranged units, like the one it originated on? A high-end defensive promotion, there aren't many of those around for archers, are there?

 

[Tomice]

Just for the sake of discussion, can we imagine a unit that the AI can use as well as, or even more effectively, than a human player?

I experienced that AI's that AI's are most dangerous when they have strong melee units with multiple promotions. And - to a lesser extent - when they have siege within range of my cities.
I remember instances where some AI had longswords before I could field them (and a general tech advantage). My comp-bows barely scratched them, especially when their rough-terrain promos applied. It was pretty scary seing them approach my cities unhindered. It reminded me of the Terminator movies

On the other hand, the typical combination of vanguards and archers we commonly see now is rather easily disabled. We just need to focus-fire the weak melee units. When they're gone, the archers (and catapults) will continue to attack our cities pointlessly, while a single archer kills them one by one from the safety of the city tile. The pillaging is a minor annoyance quickly repaired afterwards.
Note that the AI doesn't cancel a city attack when it runs out of melee units, only after their losses get too high.

Conclusions (IMO):

• It should build as many iron and horse units as possible
• It should be eager to get access to more iron/horses
• It should research techs that unlock strategic ress. units soon enough.
• It should have at least 50% melee units
• There should be enough siege among the ranged units
• It should cancel a city attack if it runs out of melee units

Of course, much of this was in GEM already...

Maybe a line of units with low upgrade costs, or a free upgrade promotion, would help keep the AI competitive.

Lowering upgrade costs for the AI is surely among the better options to compensate for its shortcomings.

 

[Ahriman]

A high-end defensive promotion, there aren't many of those around for archers, are there?

Extra range, heal while moving and double shot are all very powerful high ranged unit promotions.

If a player wants the retreat on ranged units, then they can play Inca. If ranged units are attacked by melee units they *should* take serious damage, I don't think we would want it to be easy for them to retreat.

Lowering upgrade costs for the AI is surely among the better options to compensate for its shortcomings.

Agreed! This would be a really nice AI handicap. Much better than trying to have a line of units that was cheap to upgrade.

 

[BroOfTheSun]

Conclusions (IMO):

• It should build as many iron and horse units as possible
• It should be eager to get access to more iron/horses
• It should research techs that unlock strategic ress. units soon enough.
• It should have at least 50% melee units
• There should be enough siege among the ranged units
• It should cancel a city attack if it runs out of melee units

Lowering upgrade costs for the AI is surely among the better options to compensate for its shortcomings.

Hi, i just want to add the discussion. I think what you say sounds pretty good. A question with the last point on cancelling city attack when AI runs out of melee units. Do you mean with all melee units in their army, or just the ones that are part of the attack?

Also, can the AI retreat mechanic be linked to how much HP is left in their army? For example, let's say I'm attacking with 4 units. The total HP of my army is at 400. All 4 units take some damage, and the total HP of my army now goes down to 190. From a player's perspective, it makes sense to pull your troops back and heal them. Can this type of logic be coded into the AI? Of course, it will be more complex than this simple example (i.e. have to think about what the total HP is, and how it changes as units die).

 

[mystikx21]

Lowering upgrade costs was already in place for CS units (were they not free upgrades?). I could see this as a useful handicap for the AI as well.

I'd also agree it needs at least 50% melee in an assault group. This would assure some survive to threaten the city, and they can do respectable quantities of damage to the city itself along the way (and pillage-heal to keep going). It does not matter if this is spears/swords per se in the group, but the emphasis should be more toward strategics if it can get them and that spears should remain useful alongside. Both as a counter-defensive unit and as extra damage to have to deal out to prevent city capture. But damage that it is difficult to deal out quickly enough with just the city and garrison (without a big tech lead or lots of open terrain and other combat bonuses).

I'd also second that ranged units should not get the retreat promotion (and that they have several very powerful later promotions as it is, especially range and blitz). The AI seems to be a little better at trying to pull back wounded units in BNW already (from the few test games pre CEP I played). I think that's fine rather than retreat promotions. If we could teach it to use ranged units effectively (move then shoot, and actually use the double shot promotion), I'd be much happier.

 

[Tomice]

It does not matter if this is spears/swords per se in the group, ...

Since only a certain number of units can get close enough to the city to actually do damage, it is important for the AI to understand that quality beats quantity. A swordsman is also less likely to be dispatched within one turn, actually surviving long enough to reach the city for a melee attack.



On another topic:
The AI almost doesn't build ranged ships. This makes them quite good at defending their own coast, but they'd probably be more dangerous with a few ranged ships. Especially when taking an AI's final city, it tends to have several helpless melee ships swarming around the coast while I finish the city off by land.

Also, they're not very effective at hunting down my navy, even if they have a large numerical advantage. It seems like they are limited by the relatively low sight of ships (compared to their movement speed). If my navy retreats behind the fog of war, the AI seems unable to "expect" where my ships could be. This might be very hard to improve I guess.

 

[Ahriman]

The AI almost doesn't build ranged ships. This makes them quite good at defending their own coast, but they'd probably be more dangerous with a few ranged ships

This is probably something the army composition plan could tinker with?

Also, they're not very effective at hunting down my navy, even if they have a large numerical advantage. It seems like they are limited by the relatively low sight of ships (compared to their movement speed). If my navy retreats behind the fog of war, the AI seems unable to "expect" where my ships could be. This might be very hard to improve I guess.

Didn't we find that AI combat performance improved somewhat with free extra sight promotions for AI units? Maybe we should do that for naval units too (an AI bonus, the human player doesn't need it).

 

[mystikx21]

I'd agree it should get as many swords into the group as it can. But simply having enough melee units in the force mixture to be able to still capture a city it has beaten down would be a start. The spears should be an acceptable but less effective option for this.

AI did have a sight bonus, which seemed to help yes. I'd agree it would help with navies especially.