Am I *really* that good, or does the AI *really* suck?

[leakyD]

Ok, it's been a while...like, 3+ yrs or something since I've played CM.

So I grabbed the demo, and for the most part, well done. Hat's off for a significant step up in the engine. It's only going to get better...

BUT, what's up w. the AI?

WTF, Chuck? Am I really that good that I've only 4 KIA and 10 WIA on the 3rd scenario @ elite. It can't be true. Sh*t, it's been 3+ yrs....

So, I know you guys are busy. There was a HUGE "to do" list for the new engine, and lots of other stuff to deal with...y'know, w. folks whining about M113's n' crap...but...

I guess the question is: Can there *ever* be a decent AI? Y'know...one that doesn't charge right into fire? Is it possible...at all?

Please don't tell me to just play online, or PBEM, because that isn't always feasible.

I have this fantasy of a *decent* AI...one that will at least put assets in tactically redeemable locations, or simply not rush right into kill zones....

I did a bunch of searches for AI issues, and didn't find much. Either I suck at searches, or, the thread fell off the DB or, heck....maybe I *really* am that good (heh...NOT!)

Thanks for listening....

[ChrisND]

.

[ May 28, 2008, 09:49 PM: Message edited by: Normal Dude ]

[M1A1TC]

The quality of AI in-game depends on the scenario designer. Crappy designer - crappy AI

[leakyD]

Does that mean the 1.07 demo sucks? From a design standpoint?

I can't seem to accept that the demo is poorly designed... maybe so, but, then, why have it as the demo? For the newbies to cut their teeth on the CMx2 engine?

In CMx1 the AI was "OK" w. defense..that was about it.

...but the 2nd scenario was a total pushover as well..

will it have to wait until separate CPU core(s) can be allocated for AI routines? If so, well, fair enough...I guess it will be worth the wait...

[Guest Guest]

deleted per user request

[leakyD]

So, the demo is a cake-walk for CM vets (if played as US)?

That's not fair.

How the heck am I supposed to value the merits of the game for purchase?

Do I buy it, and simply play "AI only" crafted scenarios, unless I'm PBEM/online?

Or perhaps buy, and only play as Syrian w. less quality forces...?

Or a combination of the two?

That doesn't seem fair either....

*wah!* :-(

[Pandur]

The quality of AI in-game depends on the scenario designer. Crappy designer - crappy AI

overall thats right, but in my view that cannot make up the lack of reaction by the AI. lets say maybe plan X is perfect to counter the play of person X, while person X meight trash plan Y.

so you can gamble as scenario designer about what person will do when faced with whatever situation, but the more "active" the AI you create, means the more it comes out and tries to actually do something beside staying put, the more danger you face that a player meight sit there already and cut up that force wich can only be controlled by "time" by the scenario designer.

so if the AI thinks the time is right, it goes forward and gets itelf shot up badly if the player observes the area.

its a gamble as scenario designer, either you let them stationary wich is "boring" or you let em move wich can possibly look "stupid" if it is at the wrong time/place.

either way, the AI cannot react but just act if told to, and thats not enough on a modern day battlefiled in my view

EDIT:

totaly forgott; point is also the best srenario designer is limited, it depends on how well he can guess what players do at what time, and not to a small part how big the map is and how it is build. means if the map is narrowed down and channels movement to one or two possible aproaches, its much easier than on open maps with much room to come from diffent sides and do different things.

also objectives help, you can make the player go to somewhere, howevever that all just "help" and cannot take place for "good" AI.

[ May 28, 2008, 10:41 PM: Message edited by: Pandur ]

[Paper Tiger]

I have never played the Demo scenarios so I have no idea how difficult they are to beat. It's a fair bet that if you buy the full game and play some of the missions that come with the game 'Al Amarah' being a good one, you'll probably find yourself being a wee bit more challenged.

Then, from there, you can download some of the community designed content from CMMODS. Unless you're reaaly THAT good, I'd guess you'll find yourself getting your ass handed to you by the AI when it's both defending and attacking.

As mentioned by earlier posters, the AI must be programmed by the scenario designer to do something. Otherwise, it will just park it's units in it's set up area and wait for the time to run out. Even without triggers for AI behaviour yet, some designers have found ways to make the AI VERY effective on defence and occassionally on attack too. If anything, community designed missions are notoriously difficult for the community players here. You won't find much scenario feedback where people complain that say 'your mission was too easy'.

[leakyD]

Ok here's a roadmap:

CMx2.x : dual core support, 1 core for AI, other core for everything else. Game engine is "core-aware" so if 2 CPU cores are present, one is allocated to AI, one to everything else.

CMx3 : quad core support. 1 core AI, 1 core graphics, 1 core ballistics, 1 core....whatever else is needed.

Is this a technical possibility? If so, will it improve in-game experience enough to be worth the effort?

Can an AI *ever* be written/allocated resources to ensure a "near-human opponent" experience (or at least not a "total idiot charging into fire/kill zone" experience)?

Or am I just lost in wishful thinking here?

Just curious...

[Seanachai]

Originally posted by M1A1TC:

The quality of AI in-game depends on the scenario designer. Crappy designer - crappy AI

So we're all supposed to flutter about like Blanche Dubois, and say:

"I have always trusted to the kindness of strangers!"

I'd personally prefer that the unkindness of the AI come from the competency of the game designers.

Putting AI performance into the hands of scenario designers is like leaving the course of evolution up to the residents of a Texas trailer park.

Proliferation is inevitable. As is extinction.

Would the last player out turn off the neon beer sign?

[slug88]

In my view, when playing as the US, victory should generally be inevitable, which is realistic. You'd have to be completely incompetent to lose most of the campaign scenarios. The challenge, however, is not in winning, but in preserving your force while doing so. I'm not familiar with the demo scenarios, but I'd bet that those 14 casualties you suffered could have been completely avoided; it's quite possible to complete many of the challenging US vs Syria scenarios without suffering a single KIA.

Now, if you'd prefer a more traditional challenge a la CMx1, there are plenty of user made campaigns and scenarios that should more than satisfy you. Paper Tiger's Hasrabit campaign, which depicts a fictional Syrian civil war, is more difficult in places than any battle I've encountered in CM:BB. That is just one example among many.

[Pandur]

The challenge, however, is not in winning, but in preserving your force while doing so.

there is more luck involved than anything else your guys take the wrong turn once and you have some casualties more.

beside that i dont find that verry challengeing, or lets say the wrong challenge. its quite boring to play a kindergarten in uniforms and keep em from dieing.

so i personally play "full on" with the US whenever i play US wich is not often but i played a bunch nevertheless.

whos interested in some casualties when the enemy mounts up X times more casualties than me!? iam not, and fun thing is the scenarios i played on US side are all verry forgiveing with points, means you can leave quiet a bunch of US troops back on the battlefield and you still win(without a syrian surrender wich is a auto win).

that said i see currently no challenge in preserving force. the odd 15 guys you loose in a battle, they did their job well up to the point.

but it wont make you loose a scenario,...

[slug88]

Originally posted by Pandur:

there is more luck involved than anything else your guys take the wrong turn once and you have some casualties more.

I disagree. Luck is certainly involved, but careful management will minimize its influence. I think the art of playing as the US is in eliminating the role of luck, and in attacking so decisively that the enemy never has a chance to respond. This can be quite difficult at times, especially in MOUT. To me, the satisfaction of executing a perfect assault, of taking a city block without a single casualty, is just as great as that which I would get from narrowly winning a hard fought battle.

[Pandur]

minimize its influence

exactly, but "after" that, luck comes into play. thats why i said "more luck involved...".

"...than anything else" is over done

but i asumed a player already tries and did minimize other factory so just luck is left

also minimize does not mean neutralize.

however, i see where you are commeing from. but fact is there are many different players wich play different and to a verry different degree of skill.

so if a scenario designer should choose to score a scenario in a way the US can only win with casualties like 5 or less, than i guess its not winable for quiet some others. thats also not good, but than again if they score it more loose players like me are comming wich shamelessly abuse the qualities of the US units while not respecting the "artificial" limitation with wich the US is burdened(casualties).

that leads to a total "pawnage" by the US, however with casualties involved.

thats too easy than, also for the people wich care about casualties.

so to get a battle in CMSF challengeing/balanced for the "most" people one needs to be a tricky scenario designer or do a red on red or blue on blue situation.

thats how i see it.

[Panzer76]

Good AI? No. Atleast not in 1.03, I uninstalled the game after that. Perhaps its better now, but I doubt it is significantly better.

[George MC]

Originally posted by Panzer76:

Good AI? No. Atleast not in 1.03, I uninstalled the game after that. Perhaps its better now, but I doubt it is significantly better.

Hhmm... Bit of confusion here. The tac AI is how units react in game and there have been significant changes since 1.03 (now we are on 1.08).

Strategic AI and how well that does it mostly down to how good the scenario designer is at creating the plans. The old strategic AI in CMX1 was pretty predictable and I for one enjoy having a bit more say in what the strat AI does and how it does it.

My wish list for improved strategic AI would include triggers and such like but at the moment I reckon I can get the strat AI to do enough mean stuff

Cheers fur noo

George

[Childress]

There are some scenarios I've tried where I've just pushed the pieces forward and won decisive victories. Others, can't solve the puzzle and get plastered. The challenge is there with the right battle which, for me is, vs the AI, since the game hasn't grown on me enough -yet-to contact someone for a PBEM which was the only mode in which I played CM1. But it's getting there.

Winning in CMSF may require an anal-retentive mentality. Acute attention to unit placement is a must due to the violent and precise weapons environment. You have to depend on your guys, the TacAI, doing reasonable things. So the future success of CM2 depends, imo, on further refinements to squad behavior.

[C'Rogers]

Compared to CMx1 the AI is infinitely better as an opponent. In all the years of playing CMx1 I maybe played single player five times. Now I have no problem kicking up a single player game. But as people said the quality is up to the scenario designer.

[leakyD]

OK, I played Demo Scenario #3 as Syrian.

I got totally stomped by the AI

Cool. I can learn something!

So I re-played.

I realized US thermal/NVD tech was the issue (why else would 1/2 my units get waxed in total darkness, right?), so I dispersed/hid my units as much as possible until sunup. I also gave my ATGM teams/tanks keyholed covered arcs. This was proved better a solution. I still lost a bunch of armor, but in the end, had 4x T-72, and all my inf/ATGM teams intact.

So, a little more hope for the game, but the AI still ran right through my kill zones without so much as even a *slight* altering of tactics.

It'd be nice if the AI could at least be aware enough to realize: "Hmmm.. 50+% of unit "A" KIA/WIA. Therefore, do not send more units in that area, say a 100-300m diameter area, of where unit "A" just received fire" ( or something similar..not sure what the algorithm would be).

This would *at least* prevent AI tendency to blindly stomp through kill zones, and give a *slightly* greater sense of AI realism.

Further, playing US with the objective of 0 casualties doesn't seem interesting either (unless you play Inf-only). All you have to do is overwatch with heavy weapons/armor and as soon as you ID a unit, hose the area down w. overwatch, an mop-up with your Inf. Granted, you'll still have to fire/maneuver to target (and ATGM variables on your overwatch), but the "0 casualty" rule makes for only 1 (one) way of playing, and I think that would get old after a while...at least for me it would (very real, but not very fun).

It seems it comes down to how well a scenario can be crafted/tweaked to accommodate for AI shortcomings.

I dunno. I'd kinda rather BFC put off WW2 EU theatre in order to beef up the AI (if only a little). That may not be as technically feasible as desired, and certainly does not help w. short term cashflow, so I can understand the situation.

I'll say it again: *Wahh!*

[ May 30, 2008, 09:16 PM: Message edited by: leakyD ]

[MarkEzra]

Originally posted by C'Rogers:

Compared to CMx1 the AI is infinitely better as an opponent. In all the years of playing CMx1 I maybe played single player five times. Now I have no problem kicking up a single player game. But as people said the quality is up to the scenario designer.

Very true!

When you play a player designed scen in CM:SF you are far closer to taking on a human than old CMx1. With the AI coming at you with as many as 5 different battle plans, multiple unit orders and more, it might just hand you your ass. And when you replay it...all set to clobber the AI infantry that came down road X...The AI tosses another plan at you, and your fightin' for your life. About all a designer could do in Old CMx1 is set up re-inforcements in sneaky places or shell the hell out of artifically restrictive setup zones at the start. Big difference.

LeakyD:

Demo scen are simple, easy to play battles. They're designed for somebody to quickly pick up how to use the interface and have a bit of fun...not to be able to pass judgement on the state of CM's AI. "Wahhhs", not withstanding