AI Questions?

[amr35]

I apologize if this has been asked before, but I need clarification

I saw that the CMx2 AI is scripted, but how scripted is it? If things are going poorly, will the AI try to attempt to adjust, or will it just blindly follow the original plan without any awareness of the situation above the squad level?

Also, does the CMx1 version include AI that reacts to the player, or is it scripted as well?

Thanks

[Paper Tiger]

There was a discussion about the restrictions of CMx2's AI scripting in this thread a couple of months back which might answer your questions.

http://www.battlefront.com/community/showthread.php?t=103215

The CMx1 series of games had an AI system that reacted to VP locations switching control. This gave it the appearance of reacting to the player's moves but it also led to some utterly horrible behaviour that it was almost impossible not to exploit.

[Chris Ferrous]

The AI is fairly complicated to 'program' well but even then, as stated already, it has no real reactivity at all - no 'what ifs'!

The only reaction the AI can make is at the individual unit level. For instance, a tank may take a panzerfaust hit and it'll usually back-off to avoid the threat and then come forward again when it is no longer thought to be threatened. Sometimes it may even try to go around the enemy threat; although vehicles seem more likely than foot units to do this.

On the attack, the AI plan for a particular unit/ group can be fought to a standstill in that unit morale may well drop to the point that once a unit is under fire again it either will not go forward or it even runs away. This gives the appearance of reactivity I suppose but it is a result of unit cohesiveness which affects your units too.

What the AI definitely cannot do is to exploit or pursue which is a great shame. And of course this is the great advantage you have over the AI.

The biggest change from CM1 is that the AI does not care about 'objectives' / current 'score' at all, and it will continue to follow the plan till prevented from doing so. Also, it is my impression, not confirmed b testing that unit command grouping is less well preserved in CM2 than CM1 which is disappointing, and that the waypoint system, unless nodal, has no facing element so that a unit advancing on the left to waypoint 1 may suddenly switch to the right, and vice versa, when it comes to try to get to waypoint 2. It would have been great if having arrived ay WP1, the unit then went for the first WP2 directly in front of its current facing.

[amr35]

Will each AI plan always play out exactly the same? Suppose one plan tells the AI to attack the left flank -- will the same units always attack, or will you see different units taking that approach? Because to me the scripted AI sounds as if it would reduce each single-player scenario from a conflict to actual puzzle solving, kind of like the Brothers in Arms series..not that that's necessarily a bad thing lol. It would just make it difficult to play a scenario again if you can still remember what the AI will do and where it will be.

[sburke]

It is a computer, it isn't thinking. It is simply doing what it has been told. Even if we had a reactive AI it would still be a simple matter of learning the trick to make it react and preparing yourself to counter that. It will be a long time before we see an AI that will give you the HTH feel.

Until then for the full experience you will simply have to react with other humans. And unfortunately they are wargamers- grumpy, smelly, childish, lacking in almost all basic social skills, prone to fits and tantrums, crying foul over every perceived flaw in the game and immediately citing it as grounds for invalidating your hard fought victory..... so much to look forward to!

Seriously the AI is what it is and after you have played the scenario the umpteenth time you will generally know what is gonna happen. The upside is there are so many quick battles that you will never get to the point where you remember all of em. (SDP just announced some 200 in CMFI) and if you want a better challenge, you don't need to figure out who in a 20 mile radius might actually be interested in the game. The world is your oyster. (and the human opponents out there are actually a decent bunch.)

[womble]

There isn't a fat lot of replayability against the AI. Less even than against a human, and I don't rate this game very high for replayability, since my own preferences are for the FoW to be maintained for both sides. Once you know what the other side has to throw at you, half the tactical puzzle and 75% of the tension has gone from the game.

[Chris Ferrous]

Some scenarios may only have one AI plan, but if it's 'painted with a big brush' it can still produce some variety. Others may have the full works, 6 plans all carefully tweaked, and there can be quite a bit of variability as a result.

Ironically it is possible for the simplest plans to produce the most variability. For instance, an order for a single platoon to target an advance indicated by a single line right the way across a map from flank to flank, could, by random chance, result in the entire platoon advancing down one flank one time, or spread all the way across or straight down the centre another time.

Essentially the designer can try to choreograph the entire game, or he can try to put in a bit of randomness and variety. I prefer the latter because that way I can play a scenario I made myself without knowing exactly what, where and when everything is going to happen.

[amr35]

Thank you very much for the answers everyone!

[mjkerner]

There isn't a fat lot of replayability against the AI. Less even than against a human, and I don't rate this game very high for replayability, since my own preferences are for the FoW to be maintained for both sides. Once you know what the other side has to throw at you, half the tactical puzzle and 75% of the tension has gone from the game.

Damn straight!

[ss11955]

Great info ---thanks