Jump to content

Recommended Posts

Originally posted by Blashy:

I would say that most mods will fair MUCH better vs. humans than vs. the AI. Unless you HEAVILY script to help it out.

I agree, as its much harder to write a generic AI that accounts for the features of various maps and victory conditions.

The biggest change I would like to see is the AI has access to a variety of strategies. In war, most nations pick an initial war fighting strategy to follow and then execute it.

Example - the Western Allied AI may choose to only invade France or the Nordic Countries or perhaps it invades North Africa followed by Italy. Its choices affected by its relative strength in each theater of operations. In Sc1 it would only launch a serious invasion of France.

Example: Invade Italy via Sicily

IF AI relative Strength is above 50% then it will consider this strategy. If AI Relative Strength is below 50% then it will not consider this strategy.

Control Gibraltar +10%, Control Libya +10%, Control Malta +10%, Naval Superiority in the Med +20%, Control Vichy France +10%

A few strategies well executed is better than many poorly executed.
I would like to see the AI able to execute a few, perhaps 3 strategies well in the standard SC2 War In Europe Campaign Scenario. (as opposed to operational and tactical movements).

For the Axis it would be Russia First, Control the Med, and UK First.

For the Allies it would be Liberate France, Italian Invasion, Strategic Bombing, and Nordic Invasion.

[ October 15, 2005, 06:01 PM: Message edited by: Edwin P. ]

Link to comment
Share on other sites

  • 4 weeks later...
  • 2 weeks later...

Edwin I hate to be the bearer of bad news, but the AI competency is directly proportional to the game complexity. The easier, less complicated the game parameters there are, the greater the AI challenge.

Unfortunately, since we have asked, and will probably receive a more in depth SC2 engine, the AI will, more than likely, border on the line of complete ineptitude.

Not to be interpreted as pessimism, as I'm sure we will be able to create some really challenging custom scenarios that the AI will excel at. Referring to the versatile editor.

Now please forgive my thoughts to the creators of the default game scenarios as I'm sure they are very capable. Its just that the future scenario designers will have the advantage of hindsight and forum feedback on a scale the originals weren't the recipients of.

Link to comment
Share on other sites

SeaMonkey, I agree with you. AI competancy is linked to the complexity of the game and the scenario.

As game complexity grows so does the time it takes to develop a good AI. And as you said, the shorter/simpler the scenario the easier it is to develop a good AI as the permutations of possible actions are greatly reduced.

The good part about SC2 is that HC has most likely(?) used some of the AI routines that he developed for SC1 as a starting point, and that there will be to some degree (to what extent I do not know) AI scripting for modders.

That said, I would be interested in knowing if the new AI will select from multiple strategies in waging the campaign game OR will it follow one basic war strategy as it did in SC1.

I should point out that, from prior posts, in the new editor it will be possible to prevent the Major AI nations from researching units that it will never build and that it will also be possible to prevent the Major AI from building specific types of units.

For example, you can tell the Italian AI to never research or build strategic bombers or armor.

[ November 24, 2005, 09:08 AM: Message edited by: Edwin P. ]

Link to comment
Share on other sites

It would be interesting to see if the Axis AI will vary its gameplay by having access to routines for taking Egypt, or conquering Vichy Algeria or focus on defending its holdings in Libya.

So:

Option 1 - Take Egypt

Option 2 - Take Vichy France

Option 3 - Defend Libya & Withdraw to Italy if outnumbered.

Option 4 - Take Malta

In Sc1 the Axis AI acted only to defend Libya, even if the Allied player withdrew all forces from Egypt, and would never withdraw forces to Italy even if he was outnumbered by the Allies.

Of course, the AI chance of executing each strategy should be influenced by balance of Axis vs Allied Naval strength in the Mediterrean.

ie.

Most UK fleet in Mediterrean = Sea Lion / Norwegian / Icelandic Invasion Chance Increases

Most UK fleet in Atlantic = Chance for North Africa Land Offensive increases

[ November 24, 2005, 06:26 PM: Message edited by: Edwin P. ]

Link to comment
Share on other sites

  • 1 month later...

For those interested in what it takes to develop an AI here is a link to an article covering the recent developments in the Galciv AI:

GalCiv AI Journal

One of the things I notice is that the AI is not good at adapting its build strategy.
This observation could also apply to SC1.

This was one of the flaws in its strategies I had tried to rectify -- to have it "think" about what techs it should get a bit more and away from randomness.
Another weakness in Sc1, that I am expect will be addressed in SC2.

Another thing I modify is that AI ships need to worry less about maximizing their fleet sizes and go for targets of opportunity more. They need to THINK MORE and follow scripted directives less. So this change is in though a bit too late for the Drengin this time.
In SC2 terms I wonder how much the AI will think and how much it will use scripted directives. From my perspective the SC1 AI did not use scripted directives enough. Example: Move French Corps in Beruit to Egypt after Italy enters war. Every human did it. Why not the AI? Especially, when there was no downside to this scripted response.

[ January 23, 2006, 01:57 PM: Message edited by: Edwin P. ]

Link to comment
Share on other sites

Another Update on AI from GalCiv that I found interesting:

The AI in GalCiv scales based on its intelligence. As you increase the intelligence, new algorithms get unlocked.
The hardest part is having the AI research the right technologies. A good human strategy gamer often has a "build order". But we don't want to get down with "Research this, then this, then this". That's not AI. That' s a script. We want the AI to choose good technologies based on analysis of the map, the situation its in, and then conclude that they need X. So it's easy to tell the AI "Get weapons, then get invasion, then build transport and take human out." That might work for the particular situation, but it's not useful for the thousands of other scenarios out there. That's why most single player games get boring -- scripted AIs.
This time, the AI not only has multiple C++ personalities (our AIs aren't scripted). They have multiple random sub-classes to call upon in order to deal with different situations.

Link to comment
Share on other sites

  • 1 month later...

Well, naturally, I "want" the AI to be "ok".

That's my concession to the guys that will want to play the game solo against the AI.

But I am mostly just wanting the game's core mechanics to be solid. The AI doesn't have to be invited into the game.

I miss playing my board games solo, always such a hassle finding a decent place to leave A3R set up for a long time hehe.

SC2 might be the perfect no dust collecting on it grand strategy option smile.gif

Link to comment
Share on other sites

I'll share a morsel of information. I've been working on an adaptation of the old Tactics II game and spent some time with a few AI scripts. Being about as perfectly balanced a game as possible, it's interesting to watch what the AI does. I was playing on Intermediate +0.5 experience and found the first year on the defensive and then a blitz for a win later in the second year. Same for both sides. Higher difficulty settings would be more challenging.

I included a couple of planning scripts for each side to use a left flank or right flank strategy in addition to the straight up the middle default strategy. And yes, the AI did go for the flank. With FOW; you can't see what it's up to until it commits. I was pleasantly surprised on more than one occassion. A few things still need work with the AI, and once they're done I expect to see a pretty good challenge.

This was a useful exercise for identifying a few things that needed fixing. Sometimes you don't find bugs unless you go looking someplace different. Since the Editor and custom games is all part of the SC2 package, this all requires testing. But I'll tell ya, today it was fun playing RED and BLUE all over again. :cool:

Link to comment
Share on other sites

×
×
  • Create New...