More balanced and AI-friendly promotion-system

[Bernd-das-Brot]

Combat is by far and wide the weakest part of the AI. It has huge problems with the 1upt-system and far as I know it is not to be expected that this will be changed any time. So to make combat a bit more interesting and challenging - without just giving the AI more units that we (experienced players) can slaughter - we should try to make the circumstances of the combat system more AI-friendly.

And by far the easiest but very effective change would be to make the promotion-system non-ground-based, because the AI has absolutely no idea whats going on in regard to this. And this a huge advantage to a human player!

Lets see the following example:

AI has a unit with Combat Strength 40 and Drill III (+45 % CS in rough terrain). If this unit would be attacked on a hill or forest (another +25%) it would be a tough encounter with CS 68.
But an experienced player will not do this. He will attack when the AI moves that unit in the open (rare situational exceptions possible). Here its promotions are absolutely useless (no bonus) and it gets another -15% for fighting in open terrain. So it has a meager CS of 34! That's only the half of what it should have!
And the AI is not aware of this! This is nearly like cheating.

So my proposal to solve this unfair AI-handicap is as follows:

Remove the Shock/Drill (and Barrage/Accuracy for ranged units) line of promotions completly and add combat I-V(?) promotions which each give +10% CS anytime. This would help the AI immensly without simplifying the game. Ground still matters (+25% for rough, better shooting positions), but does not cripple the AI completly.
AI promotions would really matter and veteran AI units wouldn't get slaughtered like green boys just because they are standing on the wrong type of tile.

This combat promotions can act as prerequisites for the more specialised promotions (e.g. Combat III for March or Combat I for Sentry and so on).


Sadly I have absolutely no modding skills to do this myself, but I imagine that this would be an comparable small task (compared to the scope of this great mod) with a huge influence of the fun and excitement combat in Civ5 could provide.

So what do you think?



Added on 01.11.2014:
I have opened a new thread in the balance subforum and posted a detailed proposal how I think the promotions should work for a more challenging and fun game.

Here is the link: http://forums.civfanatics.com/showth...2#post13538212

 

[Funak]

Am I the only one who actually likes 1upt? The civ4 system of my pile attacks your pile was just plain silly and the current system creates more strategic decisions besides where to place my unitpile.

Back to the original point I personally really enjoy the terrain-based promotions in civ5 and found the straight up 10% ones in beyond earth extremely boring, so I don't really see the point in such a change. Perhaps something could be done about the promotionsystem, since most people, me included, usually just goes for the same promotions every time.

 

[lockstep]

Am I the only one who actually likes 1upt?

No, you aren't. 1upt (even though the AI handles it badly) is one of the main reasons why I prefer Civ5 to Civ4.

 

[Bernd-das-Brot]

I like 1upt, too. But thats absolutely not the point.

The point is to make it better and combat against the AI challenging and not so easy to exploite as it is now.

 

[Stalker0]

Bernd-das-Brot makes a number of good points.

First, I will also admit that I like 1 UPT combat more than the old system. But that said...the AI is truly horrible at it. There are a number of computer programming reasons why, but in quick summary 1 UPT is just plain more complex to code than SOD AI was in Civ IV...and lets be honest the Civ IV AI still could have been better.


So if there are quick fixes that can help the AI they are worth considering.

So let me ask this, CIV IV used unit based promotions primarily (you get bonus against X type of units) instead of Civ V's ground based system. Would the AI fare any better in such a system? Or would an AI friendly promotion system need to be +10% CS each time?

 

[Gazebo]

Bernd-das-Brot makes a number of good points.

First, I will also admit that I like 1 UPT combat more than the old system. But that said...the AI is truly horrible at it. There are a number of computer programming reasons why, but in quick summary 1 UPT is just plain more complex to code than SOD AI was in Civ IV...and lets be honest the Civ IV AI still could have been better.


So if there are quick fixes that can help the AI they are worth considering.

So let me ask this, CIV IV used unit based promotions primarily (you get bonus against X type of units) instead of Civ V's ground based system. Would the AI fare any better in such a system? Or would an AI friendly promotion system need to be +10% CS each time?

I have to agree. The rough/open terrain isn't a terribly exciting choice anyways, and would benefit the AI greatly if it were flat combat bonuses instead. I'm going to have to argue for 'AI needs this' over 'Player Flavor' in this case.

G

 

[Bernd-das-Brot]

So let me ask this, CIV IV used unit based promotions primarily (you get bonus against X type of units) instead of Civ V's ground based system. Would the AI fare any better in such a system? Or would an AI friendly promotion system need to be +10% CS each time?

Civ4 had a mixed system. Combat I-V with +10% CS was the base of it. For a melee unit for example you needed combat I to choose shock (+25% against melee) or cover (+25% against archers) and combat II for formation (+25% against horses).

The same way would the modded promotion-tree look if we replace shock/drill with combat I-V. I cannot say if the AI is aware of the unit class of its enemies and so can make use of these promotions. Perhabs somme of the modders here knows. If not they would the AI only help perchance on the defense.


In this case the most AI-friendly variant would probably be to only use the combat promotions and the very specialised promotions (march, medic, sentry, extra flanking boni and so on) and no unit- or ground-specific promotions.

But both variants would surely be better than the actual way combat and especially promotions work.

 

[Lord Tirian]

So let me ask this, CIV IV used unit based promotions primarily (you get bonus against X type of units) instead of Civ V's ground based system.

I doubt the AI would do much better, especially because it's like... playing delayed rock-paper-scissors without the human bluffing element. I'm really not a fan of rock-paper-scissors balance because... is RPS really such an entertaining game? Without the human element? It's just memorising the army proportions the AI builds.

I'd say: keep the terrain based promotions but make them partially "flat", so something like Shock I provides "+10% combat strength, +5% extra combat strength fighting in open terrain". It means the player can feel still "smart" about using the right unit and right promotion (plus multiplayer, I guess) but that would help a lot with lessening the gap without throwing out the gameplay element entirely.

 

[Gazebo]

Civ4 had a mixed system. Combat I-V with +10% CS was the base of it. For a melee unit for example you needed combat I to choose shock (+25% against melee) or cover (+25% against archers) and combat II for formation (+25% against horses).

The same way would the modded promotion-tree look if we replace shock/drill with combat I-V. I cannot say if the AI is aware of the unit class of its enemies and so can make use of these promotions. Perhabs somme of the modders here knows. If not they would the AI only help perchance on the defense.


In this case the most AI-friendly variant would probably be to only use the combat promotions and the very specialised promotions (march, medic, sentry, extra flanking boni and so on) and no unit- or ground-specific promotions.

But both variants would surely be better than the actual way combat and especially promotions work.

The AI calculates its strength before attacking an enemy (including promotions), but not when moving to a potential location for defense. That's the problem, and why I think the +10% CB would be ideal, even if it is boring.
G

 

[Bernd-das-Brot]

I have to agree. The rough/open terrain isn't a terribly exciting choice anyways, and would benefit the AI greatly if it were flat combat bonuses instead. I'm going to have to argue for 'AI needs this' over 'Player Flavor' in this case.

G

I think the players will gain much out of it (more challenging combat without need for dull carpet of doom) and so the AI and the players win.
As the AI nearly always favors the rough-terrain promos the decision is nearly always for "kill them in the open".


@Gazebo: By the way thanks for your efforts on this great mod!

 

[ThorHammerz]

@ Gazebo:

Implementing a check for which tile a unit should move onto based on only combat modifiers would not be horribly complicated to do, but my question is, how would an AI decide when it is useful to move to a tile with bonus modifier or not?

When would it decide where moving to open terrain is better (for the sake of getting to target faster, such as a 0 HP city), and when would it decide when tap-dancing along the forest/hill to gain said bonuses (i.e. sees enemies around)?

Would it limit its check to:

- check all tiles that can be moved to;
- of selected tiles, check which ones are moving towards target/are next to target (perhaps something like capping the tangential paths to no more than 3x the minimum number of turns it would take by the shortest route? to avoid running around in circles or moving backwards)
- check selected tiles for combat modifiers
- move to tile

 

[Gazebo]

@ Gazebo:

Implementing a check for which tile a unit should move onto based on only combat modifiers would not be horribly complicated to do, but my question is, how would an AI decide when it is useful to move to a tile with bonus modifier or not?

When would it decide where moving to open terrain is better (for the sake of getting to target faster, such as a 0 HP city), and when would it decide when tap-dancing along the forest/hill to gain said bonuses (i.e. sees enemies around)?

Would it limit its check to:

- check all tiles that can be moved to;
- of selected tiles, check which ones are moving towards target/are next to target (perhaps something like capping the tangential paths to no more than 3x the minimum number of turns it would take by the shortest route? to avoid running around in circles or moving backwards)
- check selected tiles for combat modifiers
- move to tile

The AI already checks tiles based on combat modifiers and decides where to go (I've done a lot of work on this to make it better, and it is marginally better now). What the AI doesn't do, however, is check to see what nearby enemies would benefit from this because of promotions. That'd be pretty complex to code, and would drastically increase the operations needed for the AI to move units (and likely result in a very passive AI.

G

 

[ThorHammerz]

What the AI doesn't do, however, is check to see what nearby enemies would benefit from this because of promotions. That'd be pretty complex to code, and would drastically increase the operations needed for the AI to move units (and likely result in a very passive AI.

G

Would a fuzzy/approximate check per war-theatre be appropriate?

- Run check on all visible(!) enemy units in "local" area (same continent, or some other arbitrary condition, maybe put a cap on 15 tiles radius limit or something if playing on super-huge Pangaea or something)
- Tally promotions of shock and drill, and apply that as a "weight" (i.e. 6 "shock" promotions total on this island, maybe tack on a 1/(1 + 60/200) = 0.76x consideration for rough terrain... )
- Apply weight as a co-efficient/factor/multiplier/exponent/etc. into the movement consideration onto the tile.

Should not be too memory-intensive unless playing on a super-huge-many-battle-areas map, as you would only need to create/store, process, and delete 1 terrain-consideration-object per area, and only if at war (or if done at peace, with a much lower/softer consideration).


(note: all numbers/math are purely conjecture, for illustrating the example)

 

[TrystanCT]

I like this idea alot. Anything to make the game more AI friendly is a step in the right direction i think. Are there any other AI friendly changes we can think of?

 

[Deluxe]

I have to say that with the exception of the super early game, or on some really specific maps (like starting in the middle of a flat desert) I usually tend to take both promotions anyway ...
As a human player in Civ, you tend to save your units, and as they are going to fight everywhere, this ground typed promotion system is not so interesting.

That's why I would have absolutely no problem removing those promotions. It's so easy to switch from a hill to a plain that they did not have a big impact nor interest.

If you want to add promotions based on ground types, it may be better to create more "flavored" ones : bonus in forest, on desert, on snow, etc. Some unique units (if I remember correctly) already have those kinds of promotions. And they have a real impact, because snow/forest/desert zones tend to be big enough to have a meaningful impact on a player's strategy.

 

[mitsho]

On a player yes, but wouldn't then the AI try to move a "desert archer" towards the desert which means onto the other side of the city it wants to take, through enemy infested forests...

So I do see the point in having straight up strength bonusses, way better for the AI to understand. In a perfect world, wouldn't the optimal choice be between higher strength and more lifepoints? That's better (but similar) to a attack/defense split and still independent on the tactical map. That'd be a total change of the combat system so not viable nor desirable, I know

(I didn't read the thread indepth due to lack of time, so sorry for any repetition)

 

[Bernd-das-Brot]

I have to agree. The rough/open terrain isn't a terribly exciting choice anyways, and would benefit the AI greatly if it were flat combat bonuses instead. I'm going to have to argue for 'AI needs this' over 'Player Flavor' in this case.

G

If you are intending to give the terrain-independent combat system a try, I would make a promotion tree based on it. So we have something specific to discuss about the details.

 

[Funak]

If you are intending to give the terrain-independent combat system a try, I would make a promotion tree based on it. So we have something specific to discuss about the details.

Make a thread in the general balance subforum with your suggestions and I'll drop some suggestions aswell, have a few decent ideas for promotions.

 

[Gazebo]

If you are intending to give the terrain-independent combat system a try, I would make a promotion tree based on it. So we have something specific to discuss about the details.

That sounds good – let's make a thread for it, and if you want to do the XML legwork, that'd be wonderful.

Cheers,
G

 

[Bernd-das-Brot]

That sounds good – let's make a thread for it, and if you want to do the XML legwork, that'd be wonderful.

Cheers,
G

As I wrote in the opening post I sadly have absolutely no modding skill or experience. So I fear the programming must be done by you or somebody else who is more talented in this regard and likes the idea.

I offered to "shape" a promotion tree and post it here for discussion. I will open a thread (in the balance subforum?) and post the tree once finished. Hope I get it done tonight. At latest it will be ready tomorrow.