Not a screenshot, but still cool!

[Guest Big Time Software]

Well, at long last the basic engine of CM is at a point where we feel it is time to start putting in "game" stuff! Just this week several major items have been coded up and made to work. The two big ones are Victory and Starting Locations. So I figured it would be nice to tell you all a little bit about each

Victory Locations - A simple flag (flag pole plus national banner) is placed by the scenario designer to denote an area that *should* be taken/held by their respective sides. There are Major and Minor markers, represented by large and a small flags. There are also three types of markers; Allied, Axis, and Neither. There are no limits as to numbers of markers right now, but we will probably cap it at some reasonable figure just so nobody gets silly on us

Like other games, the player that "captures the flag" gets credit for that location as long as it is held. Unlike other games this action is NOT the only factor that decides who wins and loses, and by how much. As we have stated before, and in the FAQ as well, there are many other factors that come into play. Enough of them, in fact, that it might be better for you to NOT try and take/hold something because of upsetting the other factors (ex: taking too many losses). More details on this later.

Starting Locations - Each side is allowed to setup their forces in any way they want, provided they follow the "rules" set up by the scenario designer. These "rules" come in the form of designated starting areas with designated forces for each. For example, the scenario designer might decide that 1 company and 2 tanks are to start in one part of the map, while 2 companies and 5 tanks are to start in another. You are free to deploy your troops anywhere you like WITHIN these areas, but are not allowed to mix and match between them. This gives the player a whole lot of flexibility, yet does NOT allow them to do really unrealistic and/or ahistorical deployments (ah, like piling ALL your tanks on the bottom edge of the map and driving them forward at top speed, like I always did in SP <g>). Here is how it works...

Each side has up to three color coded Deployment Zones, each of which can be subdivided up to 5 times. Forces are then allocated to a Deployment Zone (i.e. color) by simply placing them inside of one. The choice and location of these zones is left up to the scenario designer, NOT the player. So say the scenario designer wants to put 1/2 of the available forces behind a big hill while the other 1/2 start out some 1000m to the right on a road. A Deployment Zone is laid out behind the big hill and is given a color (say White). The second zone is made by the road (say Blue). Now forces are placed in each Zone freely. When the scenario is presented to the player, all White forces are placed in the White Zone, Blue in the Blue Zone. The player can NOT put a Blue Unit in the White Zone or vice versa. And NEITHER unit color can be placed anywhere else on the map.

The scenario designer has the further flexibility to subdivide up the Deployment Zones. For example, the White Zone could consist of two smallish areas instead of one big one. This means that the player will HAVE to divided up his forces between the two or risk overcrowding the starting locations. But the difference is that the player can decide which forces go in which sub zone. Maybe all tanks go in one, all infantry in the other, or perhaps a little in each. Also, the designer wants can choose to put down one or two Zone colors instead of doing all three. The size of each is also up to the designer. In short, LOTS of flexibility and power are left in the hands of the designer, while realistic, yet flexible, constraints are placed on the player.

A few other things have been added as well, but aren't complete enough to discuss at this point. More next week!

Steve

[Fionn]

Sounds great.

Thanks for putting some good thought into the deployment systems. It sounds really solid for scenario design purposes.

One thing I am curious about though is the cap on the number of victory locations. Please be sure to set the cap very high since it is better to allow some to put out crap scenarios than unfairly cap the one or two designers who might validly use this many.

Thanks,

Fionn

[Guest Michael emrys]

How many would you need for a single firefight? Remember, the largest force you are likely to have in action is about a battalion. I should think that 2 or 3 objectives would already be a lot. Or do you have something else in mind that I'm not seeing?

[Guest Big Time Software]

We believe Victory Locations should be STRATEGICALLY important, rather than tactically valuable. As the player you should take tactically important positions in order to best achieve the strategic objective/s assigned to you by the scenario, NOT because there is an artificial flag stuck there. Because of this CM will most likely impose a cap, probably dependent on map size, to make sure the scenario designers use the markers properly.

If we do a Victory Location cap it will be higher than generally needed, but low enough to prevent abuse/misuse. I have played scenarios from other games that produced really silly and confusing games because the designer put in too many Victory Locations. For example, a map 1000m by 3000m won't have more than a few strategically important locations. So is there a need to have dozens of markers all over the place? Nope. It can confuse players as to what is REALLY important, and also make them do tactical moves simply to take a location that they should otherwise be able to bypass. From a tactical standpoint a little bunch of trees might be important to take, but from a strategic (i.e. scenario) perspective it isn't likely.

The way we envision the marker system working is that the MAJOR markers are used for the main goals of the scenario. Put them in buildings, pillboxes, road junction, etc. The MINOR markers are used to get people to take areas of small importance along the way, like a small farmhouse, or to ensure that both sides take a bit more interest in keeping forces spread out. This last bit SHOULD be a natural result of sound tactical and strategic planning on the players' part, but a FEW Victory Locations to ensure this isn't a bad idea. We just don't want to see little Hansel and Grettle trails laid out for the Attacker to follow

And don't forget that the starting Deployment Zones are the SINGLE best way to ensure that BOTH players set up their forces in a sane and historically correct manner. In other games Victory Locations are used for this purpose, but as you can see this isn't necessary in CM.

CM will allow the scenario designer to do what is realistic, and then some, but won't allow silly placement of markers wilynily all over the place. After all, this is a simulation of war, not of slalom skiing competition

Steve

[Guest Michael emrys]

I think I understand and agree with what you are saying, but I want to stress the point that I think that a victory location should represent an objective that is important to a higher echelon of command: regiment, division, corps, whatever, in pursuit of *its* objectives. If I am right about this, it should be kept uppermost in the mind of the scenario designer when deciding how many and where victory locations should go.

[Marko]

Good point about "Hansel and Gretel". But:

I have always thought that whatever you can do as a game designer to help the AI, then you should do it, because AI usually sucks. That's why the scenario designer should also do her utmost to make sure that the AI will give a good accounting of itself. There are (at least) two ways to do let her do this:

1) The scenario designer advises the player which side he should play;

2) The scenario designer uses special objective flags (which are *not* visible to a human player) to help guide the AI side to its objectives.

It's this second idea that I would like to impress upon BTS. To reiterate, this would imply two kinds of objective flags:

1) Those that BTS has already mentioned above;

2) The AI-only objective flags which I have described.

Importantly, both types of objective flags are under the control of the scenario designer. 2) is optional, but desirable, I think.

What do you think?

[Guest Big Time Software]

The problem with markers and AI is that it tends to use these to a fault. If the markers are visible (i.e. Victory Locations) it is usually painfully clear after a game or two that the AI is going to go after them in a predictable (or nearly so) manner. If the markers are hidden, replay of the individual scenario can suffer. The key is that if the AI sucks, markers will only make it suck a little less

We are setting fairly high internal goals for our Operational and Strategic AIs. The Tactical one (the bit that tells units how to react) is already in and surpasses anything that has ever seen the light of day. Therefore, we feel confident that we can do quite well with the other two levels of planning and execution. Perhaps we will use markers in some way (certainly Victory Locations will), but we would like to avoid this if possible. What we intend on doing is coding up the AI first, then see what kinds of aids it may need to make it better.

Steve

[This message has been edited by Webmaster (edited 02-21-99).]

[Moon]

One somment on victory flags: I found it always annoying in SP that a flag would stay in the color of the side that moved through the spot last... I encountered several scenarios where I had annihilated the whole enemy force which was holding the location but didnt have enough time to move one of my own units to the hex because the turn limit was reached.

This leads to two questions:

- would it make sense to switch a flag to NEUTRAL when there is no units of either side close to it? Alternatively, the victory location could be assigned to the force which is closest to the location without the need to actually MOVE through it.

- Will the flags have an area assigned to it which represents the victory location (i.e. you take the flag by moving within X yards to it)? And how big will this area be (e.g. fixed size for any flag or variable size)?

[Guest Big Time Software]

When I think of what Victory Locations should NOT be, SP is one of the games I keep in mind

You shouldn't have to station units in Victory Locations simply to maintain control. If you feel the area is threatened, THAT should be the reason for leaving forces there to guard it. CM has many other factors for deciding victory, so it won't be a problem if the flag doesn't "revert" if not controlled. In other words, if you wipe out almost all of an enemy's force, you will most likely win no matter who has what markers. A possible exception might be if the enemy blows a bridge you are supposed to take intact. In such a case the location is much more important, and therefore the loss of it weighs more heavily against casualties.

I haven't really checked out to see if Victory Locations are proximity based, and Charles isn't around to ask, so I can't give a final word here. I would imagine that if there was a radius would be quite small, though. If you are tasked with taking a building you really ought to be forced to get inside it. Having proximity makes this much more difficult to enforce, and also comes up with several problems if there is enemy forces near the marker as well.

Steve

[Sten]

One thing I always found irritating is how several games show Victory Locations (VL) in a 'God'way. What I mean is that if none of my troops can see the (still) neutral VL on the other side of the hill, then I, as their commander, should not know when(if) the enemy takes the position. If I get that information, i.e. the neutral flag changes color to enemy flag, then the VL's will work like sono-bouys all over the map. I hope CM will only show the status of the VL's the troops can actually see. Maybe shade them in the last know color, if you move out of sight.

How will CM take care of this?

Sten

[Fionn]

So, if I understand what you are saying about Victory Objectives (kind of a poor name given the slightly different role they have in CM) they work as follows:

1. Major objectives are used to simulate regimental/battalion level orders handed to you as Team Commander. Take this crossroads, or that village.

2. Minor objectives function as pointers to remind you of tactically important positions along the way.

[Thomas Davie]

**When I think of what Victory Locations should NOT be, SP is one of the games I keep in mind**

This is the truth. The AI suffered from what I call 'the dog in heat' syndrome. It would pursue objective hexes to the point that it wasn't in my opinion worth replaying. At least not .vs. the AI

I hope that you manage to avoid the use of 'invisible' markers for the AI for the reason that you mentioned, as well as the fact that I see no way of implementing them in a random map. Not to mention that this would just be partially obscuring a poor AI. There's probably no on else on this board that would echo my stance, but if you can't do the AI decently, take another year....or don't bother. But I'm not worried; I think that BTS will make a good AI, one that is tactically sound w/o fudging.

Tom

[Fionn]

I'll second that ..

I've gotten very tired of AIs which slavishly run for VPs or hidden objectives. Give me an AI which has a GOAL of reaching the objectives BUT takes into account the tactical situation and I'll be happy. In Europe (where we pay for local calls on a per minute basis) internet head to head (or whatever they call it) isn't that popular.

Make a nice AI and good PBEM interface and you'll keep me happy. I have already been challenged to a CM game hehe (now there's optimism for you ).

[Marko]

Thomas, you missed my point, slightly. I am not proposing that these invisible AI markers be set by the programmer. Rather, that they be set by the scenario designer. Once the scenario designer has created the map (whether by her own hand or randomly), and set some general objectives, she then considers how she would go about winning this scenario. After some thought, she places the invisible markers in such a way as to assist the AI in meeting its goals.

It up to the BTS programming guy to make sure that his AI does not slavishly follow the invisible markers like an ant follows a scent trail. I consider the invisible markers as merely a helpful guide for the AI. And you know what? If you don't like them, don't use them - they could be made optional!

But you know, we're kind of pissing in the wind, here, because we're not the ones doing the programming.

[Guest Big Time Software]

If the AI is good enough that markers aren't needed, then there is no point in having them. But if the AI *does* need them, we certainly wouldn't encourage people to NOT use them. They are either necessary all the time, or none of the time. There is no middle ground here.

But as Thomas said, the real problem is that AI props (like invisible markers) really become a problem for random maps. Yet another reason to try and avoid needing them in the first place.

Steve

[Ken Talley]

I am looking forward to seeing the CM AI. Although I must admit I have some reservations. Most wargames have truly atrocious AI's. I can only think of one wargame that has a really challenging AI. And that is Sid Meier's Gettysberg. Not to put you on the spot but do you anticipate an AI on the same level or better than SMG's AI?

PS: I really like the flexibility of your deployment process. I get a kick out of designing scenarios so those features are very interesting to me.

Ken

[This message has been edited by Ken Talley (edited 02-22-99).]

[Guest Big Time Software]

: Most wargames have truly atrocious AI's.

Hehe, tell us about it Well, at least it leaves us plenty of room to make improvements!!

Can't comment on Sid's game too much since I didn't play it all that much. But I think I can say for sure that CM's tactical AI (the bit that gives unit's their behavior) is much better from what I have seen.

AI generally stinks because it is hard to do and time consuming. Publishers and developers know that you will likely buy the game no matter what the AI is like, so why bust the bank and shipping schedule when sales will not be proportionally increased? They will only spend what they need to get you to buy their game, not what they need to get you to fall in love with every aspect of it. Simple dollar decision.'

Fortunately, we don't have to play that game with you! Since we aren't getting paid, it doesn't cost us anything more to do it right

Steve

[Moon]

Before playing Flight Commander 2 and Over The Reich I also had my doubts about the AI in CM based on previous "experiences"

However, these two games from BTS make the computer a though opponent - hey, they even have a menu option "weaker computer opponent" (or something like that)

Of course it is much more difficult to program an AI into a tactical ground warfare game like CM. I am not a programmer or anything (at least not beyond programming a database), but the task of making the computer coordinate his forces in a realistic and challenging manner seems an incredible task to me. So incredible, in fact, that I have always wondered why not make a 1-on-1 multiplayer wargame ONLY and skip the AI completely - I prefer playing against human opponents anyway...

But I am just thinking loud here, no real suggestion or something...

[Fionn]

Grumble, grumble, grumble... Don't talk to me about the Over The Reich AI. It still regularly kicks my conceited Nazi ass out of the sky (I play it as the Germans).

One tip about that game, if you've stopped playing it for a while and then go back to it do NOT start with your favourite experienced squadron. Your slowness and lack of recent gaming time WILL result in heavy losses.

The damned bomber defensive armament is really well modelled also. It's a real kick to HAVE to do initial head-on attacks. A 30mm burst into the cockpit area does wonders for a B-17s station-keeping ..

I normally treat AI promises with extreme skepticism but from previous experience with FC2 and OTR I expect a tough opponent.

[Guest Michael emrys]

>Since we aren getting paid, it doesn't cost us anything more to do it right.

Interesting philosophy. Interesting because it is my own. What will become of us in the hard, cruel world?

[Thomas Davie]

---------post from BigTime--------

If the AI is good enough that markers aren't needed, then there is no point in having them. But if the AI *does* need them, we certainly wouldn't encourage people to NOT use them. They are either necessary all the time, or none of the time. There is no middle ground here.

But as Thomas said, the real problem is that AI props (like invisible markers) really become a problem for random maps. Yet another reason to try and avoid needing them in the first place.

Steve

----------end post-------------

Normally I wouldn't specifically highlight and repeat a post from you But you bring up an interesting point; that of random maps. Quite frankly, I never thought that it would be picked up on, so hat's off to you. OTR, AS! and FC2 don't have random maps in the sense that we commonly think of them; they're free flowing quite fluid in nature. A battle where there is terrain becomes quite another beast entirely because all of a sudden the AI has to at least *have the illusion of being imbued with a terrain awareness*. Quite difficult.

Now, as the possibility of hidden flags/waypoints/signposts/etc is raised? Well, that can certainly help the AI give the illusion of being smart, but ti's not the real test of the AI. I can't comment on Sid Meier's Gettysburg, but I don't think it would have random maps.

Now, and this is straying off topic, but there is a game called Warlords 3 arklords Rising. Roughly speaking, it's an operational/strategic wargame with a fantasy bent. Set piece maps, campaigns AND random maps. So what? Well, IMO it has a sparkling AI, one that becomes readily apparent on random maps. 3 levels of play, with the AI cheating on none of them.

So, good AI is possible and on random maps too. Of course, tactical ww2 combat is infinitely more complex, esp when you consider that combat in WL3 is resolved on a stack .vs. stack basis.

I could be wrong, but I think that the SP series had waypoints. Haven't played any of them in a while though. And of course the waypoints are non existent in a random game.

Tom

[Marko]

I am a bit mystified by people, including BTS, saying that random maps are a big problem if you want to use invisible markers to assist the AI.

Oh, you know what? I think I understand now. By random map, you guys mean the *player* generates a random map, and *not* the scenario designer! Is that right? Phew! I'm pretty thick today!

Actually, I can think of another reason for those AI markers. What if the scenario designer wants his troops to *roughly* follow some historical pattern or tactically important path? For example, a scenario designer might want an AI unit to march up a bloody big mountain to do some long-range spotting. What that unit sees, will probably affect how the AI side goes about its subsequent business. Maybe that's what happened in the historical situation?

One could restrict the use of markers to the first few turns. And of course make them optional. At this stage I have some faith that BTS will create a good enough planning AI to make AI markers only marginally useful. The beta testers will probably be the judge of that, though.

Oh well, thanks for the interesting discussion, people.

[Guest Big Time Software]

Thomas, thanks for the votes of confidence! And yes, the Boys from Down Under (SSG) are well known for their AI opponents. Interesting to hear that it doesn't cheat, as the SSG guys are big fans of this (though they also feel the user should know).

Now on to Marko...

The problem with invisible markers is that they should not be used inconsistently. Either they are needed, or they aren't. If they aren't, then there they have no place in the game in any capacity.

If it is in the AI's best interest to march up a hill to get a good view, it should figure it out on its own or perhaps with some minor help from a Victory Location. But if something happens to make it think that there is a better, or more important, thing to do, then it needs the flexibility (and intelligence) to do it. In general we want the AI to make decisions based on the maps themselves (i.e. the terrain, enemy positions, friendly capabilities, etc.). What we don't want is for the AI to say, "ugh, me gots to follow dis here paths because my Master told me to" This is what AIs usually do, and it is why they almost always suck.

Hidden markers, in general, are bad news (especially for random maps). Even worse is "scripted" AI, which is "you WILL go here, here, and then here, but not over here. Do exactly as I say, even if everything tells you to do otherwise." This form of AI is so bad that we are pretending that it doesn't even exist It barely works in games like Warcraft and Quake, and totally falls down in a fluid wargame.

Steve

[Brian Rock]

"Even worse is "scripted" AI... This form of AI is so bad that we are pretending that it doesn't even exist It barely works in games like Warcraft and Quake, and totally falls down in a fluid wargame."

Thank the gods of war that you aren't going down that route. The only reason it semi-works in those games is that the player's path is channelled. What usually happens is that the scenario becomes a puzzle. Annoying and dull.

It's what put me off the C&C style game, as well as quite a few otherwise good flight sims.

I recall BTS saying you were using fuzzy logic algorithms for the AI. I think this approach holds a great deal of promise, as do neural nets and genetic algorithms. To date the games I've seen that have applied these techniques have been fizzers.

SWAT 2 springs to mind - supposedly very intelligent little AI agents, but in practice they'd stand there and refuse to return fire as the bad guys took them out one by one. I _never_ had the sniper take a shot. Big disappointment.

Not sure where they went wrong. The principles for fuzzy logic are fairly straighforward. Either their values or models were screwed up, or maybe the programming was just buggy. Dunno... but I'm a lot more confident that CM will be closer to the mark.

I reckon the first great test of an AI program is if the AI continues to surprise the designer/programmer.

Of course the other great test is that the surprises are sensible and effective ones...

The team working on the German panzer game (you know the guys - I've forgotten who they are) had that happen. I've read interviews with a few other programmers who have talked about this.

A good quality control mechanism to test for this is when the people who write the program want something that's good enough for them to play. Hey, isn't that you guys?

Rocky

[Marko]

I suppose you could say that I'm a computer AI skeptic! Or maybe too many games have disappointed me w.r.t. AI. That's why I'm a stickler for anything that helps. BTS is confident that they can overcome this enormous obstacle and produce something worthy of being called "challenging AI". If they do this, I'll have to change religions.