Implementable improvement suggestions for the military AI

Talk about on-going development of Supreme Ruler 2020 here. What would you like to see in updates or in a future Supreme Ruler title?

Moderators: Balthagor, Legend, Moderators

Locked
madjack
Captain
Posts: 122
Joined: Aug 31 2005
Location: Istanbul

Implementable improvement suggestions for the military AI

Post by madjack »

I’ve been thinking about ways to improve the military performance of the AI and put together some thoughts. So here’s one of those long and boring posts.

1. First of all, economy. The AI has too much trouble maintaining its economy during war, so the first step in impoving the AI war machine should be programming it in a way that gives utmost priority to maintaining a positive treasury balance. It should increase taxes, reduce social spending, and should never increase defcon above 3 if it has a budget deficit. The sinews of war are money, money and yet more money.

2. Maybe the most implentable and efficient solution for “decisiver” battles is changing the deployment of units. Most of the time the units (both starting and freshly built) are sent into the reserve of the closest barracks. And as far as I could observe, the barracks, or bases, act as a independent (regional) “army unit” – say, when a hostile unit enters into a certain range, the base deploys units to fight back. The point here is, if all the units are in the reserve of the base which is the closest to the enemy, we would be seeing massive battles along the borders.

So I suggest moving all the reserve and newly built units into barracks which are closer to the (potential) enemy. Most of the fabrication facilities are close to the capital, which is almost always close to the center (the safest point) of the AI country. This is an even more significant drawback when thinking about larger countries like Russia and China. So, let’s not let the AI send units into the reserve in the safer parts of its country. Those “safe” bases could as well be left empty or with 2-3 units.

3. Deployment of units along the border would be even better (but probably harder to implement). The likelihood of war is quite easy to measure, there’s even a solid value for it, which can be taken advantage of by the AI. Let it deploy most (nearly all) of its forces next to the border of its (possible) enemy (or enemies).

4. Some suggestions for army composition: The AI is not using “stacks” and I won’t be saying “it should use stacks” one more time – I the devs would have implemented it long ago if it was easy. We all know that the AI uses “swarms” instead. It’s also nice, as long as it has enough units. And to make progress beating garrisons and capturing territory, it should have enough “infantry” units. I noticed that tank units have the most priority in the contruction queue. It simply does not help enough. The AI should be building up lots and lots of mobile infantry, and the quality does not matter. Producing and using units like, for example Vodnik of Russia would work wonders when thinking about assaulting cities. Keeping in mind that only 11 days or so is needed to build one unit of Vodnik, it’s quite easy for the AI to produce hundreds of such units in a few months and deploy thousands of troops to the front.

5. And there are some things to avoid: First is the low supply areas. I think the AI should keep away from low or no supply areas as much as possible. The second only applies to tank and AT units. These units should not enter adjacent hexes of enemy cities (or all low visibility hexes), and only be used to assault military & industrial facilities, and to fight in the open.

6. Of course, such an extensive use of infantry would a major drain on military reserves. And I don’t think raising defcon is the best way to increase it. It’s just another burden on economy. The AI should be disabling facilities to increase reserves (especially naval & air fabrications). Reducing the number of garrisons is also a must. After that, if still losing, it should back off, but only after deploying all it’s artillery and tanks to the front as a defensive measure (or distributing them to multiple fronts). Losing a significant amount of units and ground from this point on should cause the ai to seek peace in any way possible.

So, is this better or worse than just “fix the ai”?

Regards.
User avatar
Ruges
General
Posts: 3408
Joined: Aug 22 2008
Location: Nearby, really I'll see you tonight when your sleeping
Contact:

Re: Implementable improvement suggestions for the military AI

Post by Ruges »

1). I think you are wrong here. Military always comes before economy, This is how it is in RL. Becouse if you get conqured militarily, Your not going to need an economy. And somtimes this way of thought does destroy itself. USSR is probly a pretty good example of this. However it is probly easier to come back from an economical collapse instead of a military conqure.

2/3) I agree here. No country is going to fight a loosing front with so many units held back in reserve. However there needs tobe balance between the current system and all out border protection. The AI needs tobe able to reconize if it should be defending its entirior, or if it should be protecting its front. Which would probly the the hardest of the task.

4) I will sugest stacks and battle groups. I think they can do it, but have not desired to do so becouse they dont believe it will make the AI any better. Becouse they believe that the current way the AI treats its units is better then what it could do if they where grouped. However that is just false thinking. The units have tobe grouped inorder to succesfuly take defensive positions.

5) While I do agree with this, I dont put a high priority with it. Its kinda suprizing how effective the AI is at in using tanks to take cities, as a player I am not that effective in using tanks.

6) And again I disagree here. Defcon 1 has allot of advantages to it. Techs build faster, Units build faster, and units fight beter. It does cost allot, However your fighting force is so much greater at defcon 1, I think it nearly out weights having it at 3. Just going to throw some random numbers that are in no way correct, but kinda give you the gist of the way it works. You have a tank that cost $10 to maintain at defon 1. and $7 at defcon 3. It fights at 100 percent at defcon 1 and 70 percent at defcon 3. So lets say you had 10 tanks with an attack of 10. SO 100 attack at defcon 1. You would need 15 tanks to reach that 100 attack at defcon3. So your army would cost you $100 at defcon 1, where it would cost you $105 at defcon 3. And that is not even considering the advantages to faster build and tech.

Of course this asumes that you are going to acualy use your units. If 90 percent of your army is siting around doing nothing. Then yea it is stupid to maintain defcon1. And that is where we have the issue with the AI. Its trying to support a massive military that it does not use.
madjack
Captain
Posts: 122
Joined: Aug 31 2005
Location: Istanbul

Re: Implementable improvement suggestions for the military AI

Post by madjack »

Thanks for the reply Ruges.
Ruges wrote:1). I think you are wrong here. Military always comes before economy, This is how it is in RL. Becouse if you get conqured militarily, Your not going to need an economy. And somtimes this way of thought does destroy itself. USSR is probly a pretty good example of this. However it is probly easier to come back from an economical collapse instead of a military conqure.
Well, the economy's sole function is to support the military here - there have been many occasions where the AI goes into debt, and is not able to purchase military goods or petroleum, so it's better to have military goods and petroleum than to have high efficiency through high defcon.
Ruges wrote:2/3) I agree here. No country is going to fight a loosing front with so many units held back in reserve. However there needs tobe balance between the current system and all out border protection. The AI needs tobe able to reconize if it should be defending its entirior, or if it should be protecting its front. Which would probly the the hardest of the task.
Giving priority to the bases closer to the borders when distributing units seem implentable to me. AI just defends its "interior" way too much. Of course, it would be a lot better if the AI could recognize things like you said. But it can only happen in a future title, and i don't want SR2020 to be left like this.
Ruges wrote:4) I will sugest stacks and battle groups. I think they can do it, but have not desired to do so becouse they dont believe it will make the AI any better. Becouse they believe that the current way the AI treats its units is better then what it could do if they where grouped. However that is just false thinking. The units have tobe grouped inorder to succesfuly take defensive positions.
Well, i'm sure they appreciate the fact that some independent operational capability for a group of units would make the AI better. It's just way too much of work, needs too much testing, and i'm not sure the current game engine is suitable for such programming. It's quite a large scale change - if what I understood is correct.
Ruges wrote:While I do agree with this, I dont put a high priority with it. Its kinda suprizing how effective the AI is at in using tanks to take cities, as a player I am not that effective in using tanks.
How? Using such low close combat values for such purposes is just wasting resources and time. It's only advantage is not losing too many men. The tanks are very useful in capturing facilities, but not cities.
Ruges wrote:6) And again I disagree here. Defcon 1 has allot of advantages to it. Techs build faster, Units build faster, and units fight beter. It does cost allot, However your fighting force is so much greater at defcon 1, I think it nearly out weights having it at 3.
Well, I won't argue about it's advantages - there definitely are some - but as I told, if the AI has no military goods or petroleum, than it's a waste. Please note the "if it has a budget deficit" part. I think it should not even research anything if there's a budget deficit.
Locked

Return to “Development - 2020”