How's the A.I?

[JayA55]

I'm thinking about purchasing CM:BS, so was wondering if I could get a little feedback on how the AI is in this iteration. I've owned CM:SF along with a few expansions and while those games were great, there were some issues with how the AI behaved. I have the BS demo as well and enjoyed it, but it's not enough to really get a good feel for it, plus a lot of the past issues surfaced when the AI had to attack. I haven't played any of the newer titles so I'm very interested in hearing how the AI has progressed.

 

Any feedback is appreciated. Thanks.

[sburke]

Honestly the AI is really dependent on the designer. It is not a magic bullet to replacing playing a human, but in the right hands it can do pretty well. On it's it's own it is really dependent on circumstance as there are many things it can't do. It can still be fun, but playing against another human is where you will get a real competitive game. Kind of all depends on what you are looking for.

AI on the attack is honestly not going to be that much different. There are a few more options for a designer, but fundamentally it is still just an AI.

[c3k]

The latest engine (v3) has an AI which can have triggered events. This is a huge benefit to AI responsiveness. As akd said, it is just a tool in the hand of the designer. It's up to him.

 

In CMSF, the AI had fewer groups and was totally time driven. For example, given an AI on defense, the designer could create, say, 3 groups. (I really don't remember what the maximum was: it was greater than 3.). Group 1 would stay in position. Group 2 would pull back at 30 minutes, regardless of the situation, and Group 3 would counterattack from the Woods to the Big Building at 45 minutes.

 

In CMx2e3 (the "e" stands for "engine": which version of the exe is being run), the AI could have 8 groups. (Again, I'm not sure what the maximum is, but I'm sure it's at least doubled over the number in CMSF. Perhaps 4x?) Now, the designer can break the defense into many more units instead of CMSF's limit. The triggers are defined by geography. Think of trip-wires. That counterattack? It won't happen at 45 minutes. It'll happen when the Human player crosses the road. That might be at 15 minutes or, for a slower guy, at 60 minutes. Better than a "senseless" timed attack. Etc., with the other groups.

 

(Timer-based events can still be scripted in, and they can be stacked. Meaning, it can also be a triggered event, but not later than a certain time.)

 

Caveat: I have NEVER designed a full-up battle and certainly none with triggers. I can only relate the above from the user perspective, not the designer perspective. There are some MUCH more knowledgeable and skilled folks who can chime in.

 

It makes for a FAR better fight against the AI. It reacts (within the bounds of the designer's imagination) to the player.

 

It is far short of the experience you get playing a human, but it is much better than CMSF.

 

(Note that triggers were introduced post-CMBN. I forget when, exactly, in the line of games/modules that triggers were introduced. The battles that came before triggers don't have them, even if the game engine supports them. For example, having the latest version of CMBN (v3.11?) has triggers. The very first battle in the release version of CMBN, there is no AI plan which takes advantage of this support, for the simple reason that the support did not exist when the battle was created. It CAN be added in, of course, using the editor.)

 

Ken

Edited May 24, 2015 by c3k

[Wiggum15]

If you talk about the TacAI then its still bad and not really any better then in Shock Force.

[delliejonut]

Negative Nancy over here

[agusto]

The AI in CMBS is not dynamic but scripted by the scenario designer. As far as i know, it has not changed in any way compared to CMSF, except for the introduction of limited If-Then triggers. The triggers work the following way:

 

The scenario designer specifies a certain area as touch-trigger. All triggers are always touch triggers bound to a specific set of 8x8m ground tiles.

A unit that has an order dependend on that touch trigger will only excute it if the trigger is activated. The trigger can be specified to be either activated by enemy or friendly forces and (i think) it can even be bound to a specific unit.

 

There are no triggers like "If A is true, do B, if A is false, do C". All triggers always only allow for one possible decision. They mostly help to coordinate AI attacks etc, but they are not decision making tiggers.

 

Examples for how triggers can be used in CMBS:

 

Bounding overwatch.

Counterattacks.

Flanking maneuvers.

Corrdinated retreat.

 

Overall the introduction of triggers has noteably improved the way the AI behaves, but it still depends 100% on the scenario designers skills whether or not the AI in a specific scenario behaves tactically good or bad. It is absolutely possible to have one scenario made by a talented designer where the AI is a real challenge and to have another scneario made by a different designer where the AI is nothing but canon fodder. The scenarios that ship with CMBS are, as far as i can tell, all very, very well made. The experience may be different though if you play user-generated scenarios.

[agusto]

 

(Note that triggers were introduced post-CMBN. I forget when, exactly, in the line of games/modules that triggers were introduced. The battles that came before triggers don't have them, even if the game engine supports them. For example, having the latest version of CMBN (v3.11?) has triggers. The very first battle in the release version of CMBN, there is no AI plan which takes advantage of this support, for the simple reason that the support did not exist when the battle was created. It CAN be added in, of course, using the editor.)

 

 

 

AFAIK triggers were introduced with and are exclusive to the 3.0 engine upgrade. CMBS natively uses the 3.0 engine, the other games got them with the 10$ upgrade.

[Stagler]

Negative Nancy over here

 

Ha, I know im going to take flak for this but wiggum is correct at the end of the day,

 

JayA55, if your going to buy this to play singleplayer, you will get your moneys worth out of the campaigns and scenarios included in the game, but you will tire of the quick battle/skirmish mode quickly as the AI is easy to defeat.

Besides the campaign and scenarios, this game is best enjoyed multiplayer, and the only real way to play this at this time is via pbem which is worth a punt.

[JayA55]

Thanks, I think I may still get it for MP, but it's honestly a little disappointing that we're still stuck with PBEM and IP. Hopefully that gets modernized in the future. 

[Stagler]

Thanks, I think I may still get it for MP, but it's honestly a little disappointing that we're still stuck with PBEM and IP. Hopefully that gets modernized in the future. 

 

We don't even have fully functioning IP either.

 

And I hope so too, I really do.

Edited May 24, 2015 by Stagler

[sburke]

Ha, I know im going to take flak for this but wiggum is correct at the end of the day,

No flak, but I woulld humbly disagree. That is far too broad a statement. I have seen the tac AI do some pretty incredible stuff (anybody recall a post I did on one HerrProbst?).

http://community.battlefront.com/topic/95381-cmbn-screenshot-thread/?p=1296285

And that was in the first version of CMBN.

And has seen much refinement. That being said I have also seen it do some dumba** stuff, kinda just like real life.

[Fade2Gray]

If you talk about the TacAI then its still bad and not really any better then in Shock Force.

 

TacAI is quirky, but I've learned to deal with it.

Honestly the only things that really rustle my jimmies is how stupid and fat infantry is.  Whether its how they usually line up in nice neat columns or get exhausted real fast even with a downhill quick move over easy terrain, or even flat terrain, infantry is maddening to use most of the time.

[MikeyD]

Exhausted troops is often a function of 'user error'. If you're planning to send your men on a run its best to not max out their load limit with everything you can acquire out of the vehicles first. CMBN has few problems with exhausting troops. Why? No body armor and no rocket launchers for the men to lug with them. Oh, and check the temps. Mid July can get pretty hot out.

[agusto]

Exhausted troops is often a function of 'user error'. If you're planning to send your men on a run its best to not max out their load limit with everything you can acquire out of the vehicles first. CMBN has few problems with exhausting troops. Why? No body armor and no rocket launchers for the men to lug with them. Oh, and check the temps. Mid July can get pretty hot out.

 

If CM players were real life commanders:

 

"Here, Private, take another rocket launcher!"

"But i already carry 5, Sir, and i am dying from the heat."

"Yes, Private, and before i forget it, here are another 5000 rounds of 5.56. I expect you to deliver them to Outpost Charlie in under a minute. It' s not far, just 1000 meters. You can easily sprint that."

"But Sir.."

"But what? I will send you on a suicide mission if you fail me! Good luck!"

Edited May 25, 2015 by agusto

[Glubokii Boy]

One thing that i would like to se would be an increase in the number of avaliable AI-Groups.

 

I don't know if increasing the number of AI-groups will have any significant impact on FPS though...If not...I think this would be a nice improvement.

 

With more AI-groups the scenario designer could 'help' the AI to perform better...For example...A platoon with an HMG in support would do considrably better if the designer could afford to spend something like 3 to 4 AI groups on that single platoon...

 

AI1 - squad 1

AI2 - squad 2

AI3 - squad 3 + HQ

AI4 - HMG

 

With smaller scenarios this is already possible but with company + scenarios supported by a number of veichles the designer could soon run out of AI-Groups...

 

having an entire platoon, or two,  being a single AI group  puts quite a high demand on the AI i think..

 

.

[MikeyD]

The best way to play the AI is to forget the AI exists. Don't try to leverage AI quirks to your advantage, just play the situation like it was reality. One of the nice things about triggers in the game is they're often entirely invisible to the player. Unless its something obvious like being attacked as soon as you reach a crossroads. Usually the chain of events is more subtle than that. I'm of course talking properly made scenarios.

[kraze]

AI is acceptable. Triggers are a game changer, especially compared to CMSF. Before I knew they were in it looked like AI finally got that reactivity it always needed instead of timed events.

 

However my main gripe with AI is how it lacks reactions programmed vs. precision strikes. If I spot a tank somewhere and launch a precision strike and it doesn't penetrate the tank - I'm free to launch as many of them as I want until they do their job - AI tank will never move.

 

This is especially glaring vs. AI actually reacting to incoming fire and threats by hiding behind something. And ofc when arty starts pounding me I insta get out.

[MikeyD]

A lot of AI criticism comes from people who have played the game engine continually for years now. Its rather like asking someone what its like to be married to a supermodel and you hear complaints about how messy she leaves the bathroom. Sure, that's valid but it seems kind'a peripheral to the overall experience. 

[Los]

The best way to play the AI is to forget the AI exists. Don't try to leverage AI quirks to your advantage, just play the situation like it was reality. One of the nice things about triggers in the game is they're often entirely invisible to the player. Unless its something obvious like being attacked as soon as you reach a crossroads. Usually the chain of events is more subtle than that. I'm of course talking properly made scenarios.

MIkeyD really gives the best advice in his two posts.

 

You know what I do? I try most of the time to play the game in real time using less than a company. (I dont have the time or patience to spend four hours fighting a half hour battle, that's what ASL is for!    don't sit there and watch each minute from every angle and then try to lay out the best path that insulates me from trouble in every situation. In Realtime with no time to screw around and under time pressure I give simple yet effective orders and let the chips fall as they do. If the scenario has been developed correctly the AI  on the other side returns a reasonable result and play experience.

 

Los

[Glubokii Boy]

 

 

However my main gripe with AI is how it lacks reactions programmed vs. precision strikes. If I spot a tank somewhere and launch a precision strike and it doesn't penetrate the tank - I'm free to launch as many of them as I want until they do their job - AI tank will never move.

 

This is especially glaring vs. AI actually reacting to incoming fire and threats by hiding behind something. And ofc when arty starts pounding me I insta get out.

 

To limit this (atleast to a degree) i would suggest that scenariodesigners when making their scenarios considder adding some small movements to AI tanks that they fear might be targeted by precision artillery. With larger scenarios this might not be very easy to do but in small/medium scenarios i think it ought to be possible.

 

These moves will not need to be timed absolutelly perfectly to happen exactelly when the player artiller starts falling...If the AI tanks change possition a few minutes before the precision strike lands then the player will atleast be forces to retarget...loosing time...

 

 

I realize that this will to a large degree be pure guess-work on the part of the designer...when to move...but some movements of static veichles would be nice...

[Stagler]

To limit this (atleast to a degree) i would suggest that scenariodesigners when making their scenarios considder adding some small movements to AI tanks that they fear might be targeted by precision artillery. With larger scenarios this might not be very easy to do but in small/medium scenarios i think it ought to be possible.

 

These moves will not need to be timed absolutelly perfectly to happen exactelly when the player artiller starts falling...If the AI tanks change possition a few minutes before the precision strike lands then the player will atleast be forces to retarget...loosing time...

 

 

I realize that this will to a large degree be pure guess-work on the part of the designer...when to move...but some movements of static veichles would be nice...

 

To be fair, they should be able to react automatically to Russian FAC/FO teams lasing them. That would mitigate some of the problem at least.

[kraze]

That wouldn't be a solution if you want to put those tanks to ambush an attacker. Plus if they will keep moving it's a straight giveaway vs. stationary.

 

*SPOILER!* like in the last mission of US campaign where I found 4 enemy tanks in the ambush position in the forest. So I just took my time blasting them off with precision strikes one by one. And as they exploded their buddies didn't even blink.

 

It's a thing that only Battlefront can fix. If AI vehicle detects incoming arty fire hitting it - it should retreat.

Edited May 27, 2015 by kraze

[Los]

BTW lest anyone think these AI issues are limited to Combat Mission, anyone who's every spent time in, say the infantry, is well aware of what its like to have an observer with your platoon OR to be an observer going with another unit. It actually doesn't matter how well trained or elite the unit being observed. When you spend time looking at them or being looked at under scrutiny, you find flaws (Even if they are perceived flaws vs what you think should be happening) and go WTF? Of course the guys being observed go, "Tactics are like assholes, everybody has one"  That's before you takie into account that under various conditions humans are apt to not the right thing at the right times.

 

Los

Edited May 28, 2015 by Los