Videogamer GDC Article

[nokmirt]

Off-topicking, but since part of the thread is dedicated to managing stacks I would like to suggest this idea (don't know if it will work for CiV, tough): stacked units to increase maintenance exponentially -->so, one-unit stack maintenance is 1, 2-unit stack maintenance is 3 (1+2), 3-unit stack mainenance is 6... and so on.

It will only take for the AI to calculate the "monetary" value of losing units against keeping them protected in a stack. I think it could cope with this well (at least, after some tweaking), and will allow for flexibility at a certain cost.

Good thought there, but I am not sure I care for that idea. It ties your hands too much, tacitically and strategically. (I have to defend my empire, yet I cannot afford to stack units right now to defend it properly. NO GOOD! !) Besides its hard enough to get cash and keep happiness sorted out, rather than adding burden of figuring out the extra maintenance for stacking two units together, which you may have to do in several tiles. There is no sense overcomplicating the game. In any case they are going to have to come up with stacking rules, if you stack no attack, something like that.

I do know this, there have been other threads in the past dealing with stacking, 1UPT, and 2UPT. Game designers I am sure have looked those over. There were some pretty good ideas, and heated arguments there, which make interesting reading. I think I even submitted some boardgame rules via Avalon Hill games, that I used to play.

 

[aziantuntija]

More rules=bad for AI?

Nope. What I meant was: More possibilities = more challenging for the AI. I belive that it is pretty much a obvious consequence.

Also you must realize that if you add a new possibility in the game, then you must make sure that the AI is as capable of utilizing that new possibility at least as well as the human player does. Since otherwise the gap between human player and AI just grows even further. I will quote what I said before:

If that is your only logic to backup limited stacking, then you must realize that due to the fact that then the human player can also just stack melee units in the bottleneck and behind them just create multiple stacks of ranged units. Human player can also recycle units very effectively to keep fresh units in the frontline and thus maintain the position. I don’t see how this makes it easier for the AI to overcome such a strongly defended spot no.

So yes I belive that your idea would be in many ways more challenging for the AI than the current 1upt we have right now.

 

[AriochIV]

The point of the Civ I system (where all stacked units are destroyed when the defender loses) is to allow stacking, but discourage it. In this system you are taking significant risk by stacking, so you are much better off not stacking. This fixes the problem where any limited stacking means you HAVE to have X number of units in a stack or it's not "full." And yet it allows the option of stacking when there's a logjam.

Another option is to treat melee units and ranged units as different types, and allow one of each in a hex. This wouldn't be the end of the world either, but it also wouldn't fix logjams.

 

[Monthar]

Forget about the choke points for a moment. Instead think back to how many times the AI will surround a city with its ranged and siege units, leaving no room for their melee units to reach the city in order to capture it. How many turns does that city get blasted down to 1 hit point because of this problem? Why does this happen time and time again? Because the AI doesn't understand it has to move one of those ranged units out of the way for a turn. Couple this with hills, forests and marshes around that city and the problem compounds itself. Furthermore, add in the wacky unit cycling and you'll find more often than not the AI's melee units move before the ranged/siege units, but because they can't move up they fortify in place instead. The AI doesn't even seem to understand the ability to have the two units switch places.

Now add in the ability to put just 2 combat units on one tile and the AI can finally get those units that can capture the city up to it and still fire off its ranged attacks. Forget about adding any complicated rules to discourage stacking 2 units and you'll find, that while it is a slight advantage for the player, it's a huge advantage for the AI.

 

[nokmirt]

Forget about the choke points for a moment. Instead think back to how many times the AI will surround a city with its ranged and siege units, leaving no room for their melee units to reach the city in order to capture it. How many turns does that city get blasted down to 1 hit point because of this problem? Why does this happen time and time again? Because the AI doesn't understand it has to move one of those ranged units out of the way for a turn. Couple this with hills, forests and marshes around that city and the problem compounds itself. Furthermore, add in the wacky unit cycling and you'll find more often than not the AI's melee units move before the ranged/siege units, but because they can't move up they fortify in place instead. The AI doesn't even seem to understand the ability to have the two units switch places.

Now add in the ability to put just 2 combat units on one tile and the AI can finally get those units that can capture the city up to it and still fire off its ranged attacks. Forget about adding any complicated rules to discourage stacking 2 units and you'll find, that while it is a slight advantage for the player, it's a huge advantage for the AI.

I have seen it many times as well, where the AI has plenty of units melee and ranged to take a city. What does it do? Attacks with one archer and a spearman. Next turn with an archer, while the spears just sit there. Then it will attack melee again from the same spear, that is still healing, from its first combat. The city could have been taken, and should have been on the first turn. There was once instance, even worse, with Siam trying to take a city state to the south. One hundred turns later it was still fighting and still did not take the place. I mean there were four of those Narusan elephants, plenty of pikes, archers, all milling around for no purpose. Moving back and forth from tile to tile, as if confused. If the AI had the intelligence of a plant it could have taken that city a thousand times. It was very frustrating to watch that and have fun playing. So there are some examples of the current state of the AI.

The one thing we did work out is that needs to be more space between cities. Two to three tiles seperation causes too many bottlenecks. You need at least four tiles, five tiles is my preference though. Still, the AI was so bad it did not have enough resolve to attack and destroy its enemy, even though it had plenty of room for units. Continuously, it attacks piecemeal, like a general with no backbone. The AI needs to act more decisively, and be able to gain the objectives it supposedly covets.

 

[Buccaneer]

Forget about the choke points for a moment. Instead think back to how many times the AI will surround a city with its ranged and siege units, leaving no room for their melee units to reach the city in order to capture it. How many turns does that city get blasted down to 1 hit point because of this problem? Why does this happen time and time again? Because the AI doesn't understand it has to move one of those ranged units out of the way for a turn. Couple this with hills, forests and marshes around that city and the problem compounds itself. Furthermore, add in the wacky unit cycling and you'll find more often than not the AI's melee units move before the ranged/siege units, but because they can't move up they fortify in place instead. The AI doesn't even seem to understand the ability to have the two units switch places.

Now add in the ability to put just 2 combat units on one tile and the AI can finally get those units that can capture the city up to it and still fire off its ranged attacks. Forget about adding any complicated rules to discourage stacking 2 units and you'll find, that while it is a slight advantage for the player, it's a huge advantage for the AI.

I would be in favor of different stacking rules for the AI and keep the current rules for human players.

 

[aziantuntija]

@Monthar, Im not saying that the limited stacking would not help the AI at all, im just saying that the limited stacking would, –in overall- help more the human player than the AI.

Well first of all, of course we cannot assume that this problem you described would just disappear with limited stacking, since I would see that it would still be very much possible (at least in cities where the roundings are limited like with mountains) for the AI to clutter the city roundings with ranged units. Sure it would be more unlikely scenario with limited stacking but its still would be a possibility. Limited stacking would surely help the way that the AI player conquers another AI players cities (is that even a huge problem in civ5? well anyway..), but I wouldn’t be so sure it would in -overall- actually help the AI against human player, wich I belive is the actual point here. If im right now having an advantage over the AI with my more prudent troop positioning, then I don’t see how it would help the AI if we would allow limited stacking. As far as I see it, limited stacking would only mean that then I can hold key points even better than before, by stacking two melee units or melee unit and a ranged unit in the same tile. The AI would still be tactically as dumb as before, but the only difference would be that now I can concentrate my troops more powerfully in even more smaller areas (key points) much more effectively than before. That surely cannot be a good thing for the AI.

 

[glaivemaster]

I would be in favor of different stacking rules for the AI and keep the current rules for human players.

I would not be happy with that at all. I already get very annoyed when the AI does its thing of building a ranged unit in a city that has one and then firing with both. Sure, the unit gets kicked out of the city at end of turn, but the human isn't even allowed to do that.

Find a combat system the AI can use (I'm not going to pretend it's as simple as "make the AI cleverer"), but allow the human to use it as well, otherwise you're not even playing the same game any more.

 

[aziantuntija]

I agree with glaivemaster about that the tactical combat rules must be the same for the AI and for the human player.

 

[nokmirt]

@Monthar, Im not saying that the limited stacking would not help the AI at all, im just saying that the limited stacking would, –in overall- help more the human player than the AI.

Well first of all, of course we cannot assume that this problem you described would just disappear with limited stacking, since I would see that it would still be very much possible (at least in cities where the roundings are limited like with mountains) for the AI to clutter the city roundings with ranged units. Sure it would be more unlikely scenario with limited stacking but its still would be a possibility. Limited stacking would surely help the way that the AI player conquers another AI players cities (is that even a huge problem in civ5? well anyway..), but I wouldn’t be so sure it would in -overall- actually help the AI against human player, wich I belive is the actual point here. If im right now having an advantage over the AI with my more prudent troop positioning, then I don’t see how it would help the AI if we would allow limited stacking. As far as I see it, limited stacking would only mean that then I can hold key points even better than before, by stacking two melee units or melee unit and a ranged unit in the same tile. The AI would still be tactically as dumb as before, but the only difference would be that now I can concentrate my troops more powerfully in even more smaller areas (key points) much more effectively than before. That surely cannot be a good thing for the AI.

The AI as it stands needs a severe rework anyway. They have to start someplace and I am sure they are testing their new combat system, whatever it is. Hopefully we will find out more soon.

I believe you are right in the sense that the current AI cannot do the simplest things correctly. So adding more for the current AI to take on would not work. The designers know this thats why they are changing things to make the AI better. This probably would entail coding I would imagine, so it can take on more responsibilty due to rules changes, and play much better. Whether they can do this remains to be seen.

So reworked AI with added rules etc.= Better at combat? I hope so

current AI with added rules etc.= NO GOOD!!!

 

[CYZ]

Nope. What I meant was: More possibilities = more challenging for the AI. I belive that it is pretty much a obvious consequence.

Also you must realize that if you add a new possibility in the game, then you must make sure that the AI is as capable of utilizing that new possibility at least as well as the human player does. Since otherwise the gap between human player and AI just grows even further. I will quote what I said before:



So yes I belive that your idea would be in many ways more challenging for the AI than the current 1upt we have right now.

It is indeed more challenging, true. As I said, they will need to be able to judge when to stack and when not to stack. However, I believe AI will profit more from this idea than humans. For humans chokepoints are not a problem, we can easily see how to either avoid/bypass them or what the best way and order is to move through our armies. The AI fails horribly at this however, and that's exactly where this change will help them alot, while only slightly helping the human.

So more challenging for AI? I agree. However that's only looking at the downside, if we look at the upside we see it's also much more profitable for the AI.

 

[Gedrin]

I am quite amused at the discussions of 1UPT vs 2UPT.
I play with a mod that provides 3UPT. I can say it is way way better. I will not use the word "good" it is still an AI and its not very good... but it is MUCH "better"

Those of you who have actually tried something other than 1UPT... what do you think? Did it help the AI?
Those of you who have not and like to share your opinions of the validity of 2UPT or 3UPT or 99UPT... I would encourage you to actually validate your point of view with experience. It's really worth it IMO.