AI any good?

[bssybeep]

I seem to recall reading that the primary focus for this game was multiplayer, is that true? I am interested solely in single player mode. So, I am mainly interested in how is the AI and replayability of the game.

*AI any good?
*Can custom scenarios be played in single player mode?
*Do the inculded scenarios have random factors or are scripted?

Thanks in advance

[06 Maestro]

oops

[Redmarkus5]

OK, I have now played several scenarios, both as attacker and as defender.  The AI needs some work for sure:
[ol][*]AI divisions seem to chase even minor unit remnants all over the map, instead of concentrating on maintaining their advance.  At one point I saw 2-3 AI divisions converging on a single sub-unit of mine.  Then, after they'd killed that, they raced off to the side of the map to get at another small unit, totally ignoring their axis of advance.  Weighting of these priorities needs urgent adjustment in a hot fix IMO.[*]When defending, the AI is too aggressive - it leaves its prepared positions and assaults your troops right from the get go, even before you have advanced to contact.[*]The AI is a kamikaze pilot - it continues attacking until it's well and truly dead.  It needs to slow down, pull back and recover after taking losses, use artillery PRIOR to attacking, probe the flanks, etc.
[/ol]These points need to be fixed before the game will be truly ready for single player use.

Markus

[GIveloper]

[font="microsoft sans serif"]Hi Mark,[/font]
[font="microsoft sans serif"] [/font]
[font="microsoft sans serif"]Right to the line.. we know the AI is not a conservative gamer. Too aggressive… we are working on that to provide tools or something to average it.. [/font]
[font="microsoft sans serif"] [/font]
[font="microsoft sans serif"]Victor.[/font]

[Redmarkus5]

That's great Victor.  I have some ideas [:D] about what is needed in terms of 'rules' for the AI decision process, but first I have a question.  Does the AI log the number and strengths of the units it can 'see'?  I am thinking that the starting point for an AI decision to defend or attack should be its level of Intel.  This should include FoW as a factor affecting the AI - it should not see everything - but the strategic situation should be coded in as well, so that the US never attacks on Day 1, for example (or even Day 2!)

[GIveloper]

[font="microsoft sans serif"]Hi Mark,[/font]
[font="microsoft sans serif"] [/font]
[font="microsoft sans serif"]That is a good constraint, but as you will suggest more thinks, let me tell you how we are working now for the next AI version. [/font]
[font="microsoft sans serif"] [/font]
[font="microsoft sans serif"]We are thinking to set up an AI at 4 levels.[/font]
[font="microsoft sans serif"] [/font]

[font="microsoft sans serif"]High commander: it commands at level of division and support units (engineers, heavy artillery and airforce). Strategic AI engine.[/font]
[font="microsoft sans serif"]Divisionary: it coordinates the Regiment (of the Division), basically a first level tactical AI engine.[/font]
[font="microsoft sans serif"]Regiment: it coordinates the battalions inside the regiment placing them in the best tactical position.[/font]
[font="microsoft sans serif"]Battalions HQ: mostly for isolated Battalions, very limited as the other three levels will carry on with most of the decisions.[/font][/ol]
[font="microsoft sans serif"] [/font]
[font="microsoft sans serif"]That is very hierarchical organization; obviously gamers have more freedom creating their own combat groups but clarify the so complicate development of AI for this game.[/font]
[font="microsoft sans serif"] [/font]
[font="microsoft sans serif"]Scripts: we do not use script at all… but perhaps the master plan could be designed as initial instructions more than fix scripts. Indeed is what you do in the deployment phase, the key moment of the game, played even before to click start!![/font]
[font="microsoft sans serif"] [/font]
[font="microsoft sans serif"]Now the question is define the limits of each one and what could be those decisions..[/font]
[font="microsoft sans serif"] [/font]
[font="microsoft sans serif"]All the best, Victor.[/font]

[Redmarkus5]

Thanks Victor - that makes perfect sense.

I think one of the key words should be 'Posture'.  For example, if the posture of the Strategic AI is 'Defend', then the lower level AIs can only launch very limited local attacks, and only if they have already defeated an enemy assault with high casualties.

Markus

[GIveloper]

[font="microsoft sans serif"]Hi Mark,[/font]
[font="microsoft sans serif"] [/font]
[font="microsoft sans serif"]In GC a well commanded army can beat a twice or bigger army, so if we constraint the defense army to limit counterattack probably we will see other non understandable actions, as stop a successfully counterattack that can almost evolve the total attack army (no all human generals left troops on rear to avoid these situations..). The AI can not be predictable or gamers soon will see those limitations. Probably we can agree to use the term, “risk”. So attack army needs to risk more than defensive army as its victory counter is negative at the beginning of the game.[/font]
[font="microsoft sans serif"] [/font]
[font="microsoft sans serif"]Obviously you are designing the AI for someone that really knows what he is doing… but AI must be adapted to any gamer, also those that play GC like a “another RTS”…. And let him know, GC it is not an RTS when a AI counterattack get you rear lines, and your troops need to retreat using the latest supply to open the lines losing all victory possibilities…[/font]
[font="microsoft sans serif"] [/font]
[font="microsoft sans serif"]Victor.[/font]

[Txema]

How is your work on the AI progressing?

As it has been discussed in this thread the AI has very important problems now in the game...


Txema

[sabre1]

I was wondering the same thing. Is this game still being worked on? From what I have followed in this forum the game has huge potential. I was hoping for a patch to address some of the issues that were brought up. I almost always wait for the patch before jumping in on a game.

[SlickWilhelm]

Me, too. Even though I'm not thrilled by the Bulge setting, after reading the positive remarks by folks here on this game's potential I'm going to jump in and give it a try once the patch is released. Any game that tries something innovative deserves to be supported.