Engine rewrite .. single player mode

[Barleyman]

Let's make a separate thread, maybe I'll provoke someone to comment. I feel CMBO and CMBB were and are wonderful fun as PBEM, but the single-player experience leaves something to be desired.. Looking at what people have ben requesting for CM2, nobody brings up solo-play. I think most people would agree it's sort of boring against the AI as it is. Baby+candy sort of way.

How about some single-player improvements?

</font>

• Chaining scenarios to mini-campaigns with pre-defined reinforcements in-between. Maybe relative to victory level?</font>
• Allow date to change in above. Personally upgrading dinky Pz-IVFs to Pz-IVGs was always damn nice in Steel Panthers.</font>
• Allow scenario designer some simple scripting for AI. Think "attack vectors", "target area" and maybe "defense zone". This would tell the poor AI where to set up for a fight. Fallback/regroup-area would be nice, too.
  </font>

Anyone?

[Mr Byte]

the AI scripting is a wonderful idea !! and would make for very challenging games but also it could be used in games by the player to order platoons to fall back etc and not have to micro manage every aspect, done right it would allow the player to take on a 'commanders' role and would make larger battles more manageble.

[Captain Wacky]

There was a big discussion about scripting AI while back but I can't remember how it turned out. Steve chipped in though, and you may get some answers if you do a search.

[Egan]

I would love to see some sort of AI scripting, even if it's no more detailed than" go down that road, not that one." Just think of the possbilities!

[Michael Emrys]

Although I can't personally recall BTS ever quite explicitly discussing it, my feeling overall is that they are not interested in scripting as a resolution of the AI problem, but are more in the way of searching for more and more sophisticated forms of emergent behavior. I really can't say if they would fall back on scripting if their search doesn't bear more fruit in the short term though.

Michael

[Lt Bull]

Barleyman,

As much as I too would like any AI/CPU opponent to be more challenging (not just in CM), I have basically resigned myself to the reality that it will be a long, long time before it is practically possible to code an AI/CU opponent in a game like CM that will be anywhere near as challenging as it is playing a competent human opponent. I believe that the AI/CPU coding that goes into CM, although it the best of any AI/CPU coding of any game of its kind, is in the category of "most difficult kind of computer game for which to code AI/CPU opponent". For me, the most important AI in CM is the TacAI. The StrategicAI is what is coded for the CPU opponent in single player games, and it is THIS StrategicAI that is the most DIFFICULT AI (and I would guess, most time consuming) to code so as to result in a competent AI/CPU opponent.

Because I almost exclusively play vs human opponents, I could even say that if BTS ditched the StrategicAI altogether (and therefore the AI/CPU opponent) to focus on improving and expanding on other aspects of the game (like the TacAI, orders, game engine etc), I would not care less. I might even welcome it. Perhaps we are 10yrs or more away from seeing a game like CM having an AI/CPU opponent that is even in the ball park of what I consider the "best" AI/CPU opponents in any computer game, and they are the ones in computer chess games, like the ChessMaster series. 99.99999% of the population may probably never even get to meet anyone who could even come close to beating the best AI/CPU opponent in ChessMaster, it is that good.

It is interesting to note that despite the complexity of the game chess, coding an AI chess opponent for is comparatively MUCH easier to code than a game like CM. The analogy is that chess is like CM with only one map, one scenario and one fixed OOB for each side every time. If this were the case in CM, then I would imagine that the StrategicAI would be much better than it is now. It becomes quite apparent when you consider the range of units, the range of maps, terrains and the range of situations and conditions for which BTS are trying to code a competent StrategicAI for, it is a great achievement that they have even got the AI/CPU opponent to the level they have.

So in ending, I think that of all the aspects of CM, the area which I believe is the most challenging and most difficult to improve, it would surely be the AI/CPU opponent (or StrategicAI). Unfortunately, I see it as a case of lots of effort needed for very little gain/improvement . I hope I can be proved wrong.

BTW, you suggestion for scripting and attack vectors etc, is similar to what the creators of TacOPs do for the AI/CPU opponent in TacOps. However, this means that only the selection of maps/scenarios which have been designated as "playable against the CPU" can be played against the CPU, as the AI/CPU opponent has been INDIVIDUALLY scripted for each of those maps (I think the AI/CPU has something like 4 or so different battle plans it can choose from for each map/scenario). This means that any custom maps/scenarios created by users can not be played against the AI/CPU.

Lt Bull

[Crimguy]

Scripting is done in SteelBeasts (highly recommended by tankheads everywhere) to great effect, but it takes a lot of skill to utilize scripting tools properly. As a result, a lot of mediocre scenarios are created that are relatively easy to beat.

I think AI needs to be told more clearly by the mission designers what is important in terms of terrain as well as what the most likely paths of advance/contact exist for any given map, in addition of course to it's knowledge of VL's. It might be a bit too much to ask the AI to figure that out for itself.

[Cameroon]

Unlike Lt. Bull, I play almost exclusively vs the AI for various reasons. I play the occasional PBEM, but mostly I just like to play a quick game when I've got the time. I don't think there's any research concluding that there are more or less of either variety (that is, mostly solo or mostly multiplayer), thus there's every reason to believe I'd not purchase CM without its solo play . Of course, I don't believe even for a second that BFC would head that direction.

It seems that there is a focus on the AI being "one or the other" as either scripted or on its own. Personally I don't see why both can't be used together. Allow the designers to provide hints and set the flexibility of those hints (such as, "Might be a good idea" and "Do this or else!") but don't force the designers to do so.

Sure, one could build bad scenarios that way, but one could build bad scenarios anyway

[Eden Smallwood]

Originally posted by Cameroon:

It seems that there is a focus on the AI being "one or the other" as either scripted or on its own. Personally I don't see why both can't be used together.

Absolutely. That jump to a bad conclusion has been made before in this topic. Noone has advocated making scripts a mandatory part of scenario design, just an added tool.

What we SP people would like to see for the rewrite is simply more power of all kinds in the hands of the designers- it's up to them to use that power wisely.

Like Luke using the force.

Eden

[laxx]

"it's sort of boring against the AI as it is. Baby+candy sort of way"

eh ? excoose my engleesh but are you saying Solo against AI is too easy ?

[Barleyman]

The point was exactly to give some more options to the scenario designers, which they can choose to use or not. Single-player scenarios would have limited replayability, not necessarily a big problem.

I saw the BFC comments which was basically that only the BFC programmer is allowed to comment on AI

And my other suggestion, to chain scenarios for mini-campaigns.. That'd take minimal changes to the game engine, being front-end stuff. That would also create extra dimension of losses .. Pyrrhic victories?

[Barleyman]

Originally posted by Michael emrys:

Although I can't personally recall BTS ever quite explicitly discussing it, my feeling overall is that they are not interested in scripting as a resolution of the AI problem, but are more in the way of searching for more and more sophisticated forms of emergent behavior.

That's pretty universal engineer mindset. When I work on a design, sometimes I have to swallow my pride and do things un-elegant way, which does what it's supposed to do. Customers do not pay us for a product which makes especially clever use of technology. They pay us for delivering a product which performs favourably compared to other products in the market.

In this case, scripting or hinting the AI during the scenario design might be too easy a solution.

[Terrapin]

I'd also like some scripting ability. If nothing more than "wait until THIS group gets to THIS point, then MOVE this other group". I find it hard to believe Charles can't program this in his sleep.

As for the AI intelligence, I find it decent enough. If I don't do anything, I get my a$$ kicked, and that's usually the sign of at least competent AI.

It's also surprisingly adept at selecting clever locations for defensive obstacles. It likes to put those minefields right where my tanks would like to go!

[peckham]

Wait--so are we saying that it's impossible for Joe Average Consumer to actually script AI in CMBO and CMBB? I've never spent any time in the scenario editor to speak of, but I just assumed there was at least some basic interface to do things like direct a unit to move to a certain spot by a certain turn.

I did a bit of work on HPS' Squad Battles series (I'm the webmaster for World at Wargaming) and while the 2D paradigm certainly simplifies basic AI scripting in that series, you are at least able to "direct traffic" on a more or less turn-by-turn basis. It is absolutely essential to coordinating assaults, setting up fallback points, and so on.

Am I reading correctly that every user-designed scenario out there is designed for multiplayer only? Does the computer AI somehow auto-prioritize movement paths and objectives, and auto-initiate advances on its own when you play vs. it?

This would be pretty disappointing. In spite of the fact that computer AIs generally tend to be vastly inferior to uber-human-opponent, I'm weighted 90/10 toward the single-player experience due to my fascination with AI design and theory. I also prefer the idea of replaying something with strategic/tactical historical value, as opposed to merely a historical setup that deviates sharply from history after turn one. And it can be a pain in the butt trying to set up reliable multiplayer matches with random strangers.

[ December 08, 2002, 09:33 AM: Message edited by: peckham ]

[cbb]

Originally posted by Cameroon:

Unlike Lt. Bull, I play almost exclusively vs the AI for various reasons.

Same here. I get nervous when I hear people say: "It's impossible to program a really good AI so let's just leave it out altogether." ...

Even if the AI forever remains inferior to a human opponent, my hope is that Battlefront will continue to improve the AI. I think with improvement to the AI (scripting, waypoints, whatever), scenario designers will be able to design human vs. AI scenarios that are fun and challenging.

[Lou2000]

The AI isnt going to be an easy thing to change to make it a challenge for most human players ... If BFC manage it then they can probably retire ... most of the computer and scientific world will be knocking at their door !

Scripting may work to some extent .... events could be activated by triggers ... it seemed to work ok in Op Flashpoint .... yeah i know totally different game type and gameplay .... Probably not workable in QB's but could play a part in user designed battles.

Lastly .... I've said it before, but never seen much response either for or against.......

I'd like to see the option for a single player to just control a percentage of units on a side ... say 25%, 50% or 75% with the AI controlling the rest. Meaning the player would be cast in the role of 'helping' to achieve the objectives rather that in total control. Able to support the AI controlled units, but not issue individual orders.

Added commands for 'request support' 'request artillery' 'request armour' support could allow the player to call for assistance ... but the AI may not respond, leaving you to get on with it as best you can ........ All of which is not totally unrealistic !

I think that would certainly add something to single player games. Hell you could almost 'hotseat' a game against yourself without being confident of what side would win.

Any comments ?

Lou2000

[Tarquelne]

Scripting may work to some extent..

I think that's most likely to be BFC's attitude (if they go with scripting at all). I don't think anyone wants to replace the AI, but even a few basic script-like tools could make a big difference in scenarios.

The new AI, if it does indeed incorporate "memory", will be a big step foward, but I bet it could still use a few nudges in the right direction.

I'd like to see the option for a single player to just control a percentage of units on a side ... say 25%, 50% or 75% with the AI controlling the rest. /snip/ Any comments ?

Sounds fun! (If frustrating at times: "The enemy is in THAT direction.... come, you can do it... you can do it.... NO!") Maybe even have multiple players on the same side, who (via the honor system) don't "chat" more than they'd realistically be able to... Which, come to think of it, is pretty much the multiple player per side feature already requested by the TacOps fans, with the AI allowed to control some units.

[peckham]

Originally posted by Lou2000:

The AI isnt going to be an easy thing to change to make it a challenge for most human players ... If BFC manage it then they can probably retire ... most of the computer and scientific world will be knocking at their door!

When I interviewed Soren Johnson at Firaxis on the Civ III AI, as well as two of the Codemasters AI developers, the sense I got from the AI community in general is that it's not nearly as difficult as most believe to code a challenging AI that doesn't cheat. Most developers aren't out to create such a beast, however. They only spend a fractional percentage of total development time on coding the AI, first of all, and even then most of them are aiming for a fairly wide demographic.

Most non-grogs that play CMBO or CMBB will be heartily challenged by the computer AI because they won't be doing things like looking for perfect hull-down/maximum-LOS spots for a tank, or using HQ with binocs as spotters for mortar teams sitting on the other side of a hill somewhere. For this crowd, the AI will be plenty challenging.

My point, which was really Soren's (and shared by the folks at Codemasters and Bioware who contributed to my article on AI for CGM) was that it's not nearly as difficult to create a challenging AI in these sorts of games as some might have you believe.

Scripting may work to some extent .... events could be activated by triggers ... it seemed to work ok in Op Flashpoint .... yeah i know totally different game type and gameplay .... Probably not workable in QB's but could play a part in user designed battles.

Absolutely. Many other games do this, 3D or otherwise.

Speaking to your last point, I'd love the option to play cooperatively at different levels, with the computer AI. A year ago I was engaged in a lengthy discussion with several grogs over at the now-RIP Computer Games Magazine forums on various ultimate wargames in terms of realistic C2 vs. engaging gameplay. One of several possible variations on that theme included a game like CM where one could elect to have several battalions engaged in a battle, with you assuming the role of one battalion, company, or platoon, with commensurate C2 issues sorted out in terms of FOW and your ability to receive orders from your HQ (and so on...the list of issues here can be exhaustive depending on your level of interest).

It's low on my request list for future CM games, just given the other things I'd like to see and which I think are more tenable given BTS' development capabilities, direction, and goals. I think it's something that will factor into a certain sort of future wargame, eventually.

[ December 08, 2002, 10:41 AM: Message edited by: peckham ]

[Michael Emrys]

Originally posted by Lou2000:

Lastly .... I've said it before, but never seen much response either for or against.......

I'd like to see the option for a single player to just control a percentage of units on a side ... say 25%, 50% or 75% with the AI controlling the rest. Meaning the player would be cast in the role of 'helping' to achieve the objectives rather that in total control. Able to support the AI controlled units, but not issue individual orders.

Added commands for 'request support' 'request artillery' 'request armour' support could allow the player to call for assistance ... but the AI may not respond, leaving you to get on with it as best you can ........ All of which is not totally unrealistic !

I think that would certainly add something to single player games. Hell you could almost 'hotseat' a game against yourself without being confident of what side would win.

Any comments ?

Sounds good to me. I'd give it a whirl. Might also be good for those players who like to fight really big battles but who don't like to micromanage.

Michael

[Cameroon]

Originally posted by Lou2000:

Lastly .... I've said it before, but never seen much response either for or against.......

I'd like to see the option for a single player to just control a percentage of units on a side ... say 25%, 50% or 75% with the AI controlling the rest. Meaning the player would be cast in the role of 'helping' to achieve the objectives rather that in total control. Able to support the AI controlled units, but not issue individual orders.

Added commands for 'request support' 'request artillery' 'request armour' support could allow the player to call for assistance ... but the AI may not respond, leaving you to get on with it as best you can ........ All of which is not totally unrealistic !

I think that would certainly add something to single player games. Hell you could almost 'hotseat' a game against yourself without being confident of what side would win.

Any comments ?

Lou2000

I'm with you and Emrys, that sounds like something I'd definitely get a kick out of. Given that there are hints of multi-multiplayer (i.e. more than one person on each side in control of a portion of the force), I would think we might have a fair shot at getting this They could even use the same interface, just allow us to substitue a computer player for a human.

Oooh, now I'm excited.

[Major Jerkov]

Thats something that really bugged me in OFP, borg spotting to the 9's, bastages see through bushes, trees, anything... and all you would see is a little black speck opening up on you with extreme accuracy... gah. Ohh, scripting, yes. Its usually been a way in most games to compensate for less than satisfactory AI, I could see it being a valuable tool in this game as well, as long as it doesnt kill replayability, ie alternate AI "plans". Also, the cooperative games with the "AI" could be fun but also very frustrating as someone above pointed out.

[peckham]

It doesn't have to be to compensate for poor AI. In wargames, it's usually meant to supplement a tactical AI by offering a historical strategic directive. The strategic AI in Combat Mission, while impressive enough for what it brings to the table, doesn't allow a scenario designer to build historical imperatives into each side's strategic initiatives.

In the smallest scale scenarios, it's not as important because the size limits the amount of gray-area-between-strategic-and-operational flex you can inject into the design. In the larger scenarios, of course--battalion level--it makes an enormous difference to the deployment of forces and the specific groups that you assign (in various strengths) to attack specific objectives.

This is really only a problem for folks interested in recreating certain battles based on historical record and eyewitness testimony. It's not a criticism of the game engine proper, but a request to have the capability to have some operational control (directives-through-scripting) over a given side.

Realizing of course that the best laid plans go to hell as soon as they encounter combat, it would still be fun to have as optional for purists interested in attempting to tweak a battle toward historicism as much as possible.

[Pingu]

Jesus. You guys are so unhip it's a wonder your bums don't fall off.

The ability for the AI to play on any RANDOM map and have it surprise you every now and then is a brilliant achievement and this alone makes me doff my cap to the programmer.

Adding scripts would be banging the drum of defeat. Red alert was all scripted. How many played the same mission over and over again.

Yet in cmbb i have played Hill 312 a variety of times (especially when the patch came out to test the differences).

But then again if they brought out the perfect game, where would we all go to complain. The forums would be empty.

You can't please everyone. But BFC have done a fine job in trying.

[Cameroon]

Originally posted by Pingu:

Jesus. You guys are so unhip it's a wonder your bums don't fall off.

The ability for the AI to play on any RANDOM map and have it surprise you every now and then is a brilliant achievement and this alone makes me doff my cap to the programmer.

Adding scripts would be banging the drum of defeat. Red alert was all scripted. How many played the same mission over and over again.

Yet in cmbb i have played Hill 312 a variety of times (especially when the patch came out to test the differences).

But then again if they brought out the perfect game, where would we all go to complain. The forums would be empty.

You can't please everyone. But BFC have done a fine job in trying.

I think you're missing the point, we want to be able to give it HINTS through scripts (or whatever you want to call them). In fact, I think everyone here is quite impressed with what the AI in CM can do.

What people are asking for is the ability to provide some external direction, not to gut the AI. My own personal suggestion is to be able to provide hints to the AI and set the strength of those hints. Thus providing a way for the AI to determine how important a given directive/hint is and if it should abandon or ignore a given hint.

Read the thread again, no one wants the AI to lose what it has now.

[aka_tom_w]

there is a whole "other" thread about this if you want to read more:

click here its just been bumped to the top

From the first post of the old thread.....

"Zarquon posted

P.S. : it's been said before : what we really need is the possibility to add scripting to

scenarios. It would not help QB gaming vs. the Artificial Incompetence, but imagine pre-made

scenarios where the designer could mark different paths for an AI attacker, set movement

modes (cautious, overwatch

yes/no, sneak, rush...), designate reserves and their trigger conditions etc.). Even simple

scripting options would vastly improve solo gaming.

If anyone is interested, please open a new thread."

Jörg

OK that sounds good to me.....

That would happen in the editor and not be "open source" code.

I think this suggestion is (perhaps ) a workable way (in CMII) for scenario designers to get the most

out of the AI by scripting some opening moves. Maybe SOP's could be added. I have heard alot about

SOP's in Aiborne Assault (I have not played this game but other's here have raved about the SOP's in

that game) maybe SOP's plus scripting in the editor could provide tools (in CMII) that would give

scenario designers what they need to enhance the AI and the Single Player experience so it could be

more fun and more challengeing??

Just a thought.

-tom w"

[ December 08, 2002, 07:18 PM: Message edited by: aka_tom_w ]