AI attacking in QB's - implications for CMBB

[James Crowley]

The general opinion on this forum, in regard to the AI's performance in attacking, especially in QB's, is that it less than stellar by a big margin. I tend to agree.

Given the overall performance of the AI, which is very good, this aspect has always been something of a surprise and a source of disappointment.

I use QB's far more than scenarios, especially as more and more scenarios are being constructed specifically for head-to-head play and/or are simply too big for my taste.

I did a little test whereby I set up various QB's but turned off the FoW, then cheked-out how the AI had set up. Oh dear :eek:

Firstly, the bulk of the units were set up almost as far back as they could be put; certainly nowhere near the optimal distance that presumably most human players would place them. So the AI has the immediate disadvantage of having to travel further, just to get started. In heavy terrain this aspect would be exacerbated.

Secondly, the AI seems to like to set up in cover. Unfortunately this leads to everything being scattered to the four winds. Very few squads were anywhere near their respective HQs and most of the heavy weapons seem to have been placed right back in the rear.

So with most squads out of C&C and the mortars and heavy machine guns hopelessly tucked away at the back, is it any wonder that the units that seem to show up leading the charge, as it where, are the tanks and HQ's. In a default 30 min game, many of the "tail end charlies" never really participate. And this seems to include FO,s; so no artillery support either.

My understanding is that the AI was tweaked to make it more aggresive but this does not seem to help in this situation; only those units that can move without command delays or are not "slow" will benefit. The other "stuff" pretty much gets left behind, so the result is a disjointed and ineffective attack.

If CMBB is putting more emphasis on C&C, which seems to be the case (and rightly so), then the Russkies in particular are going to be completely bolloxed if controlled by the AI in a QB.

IMHO it is the setup that needs to be tweaked, with greater emphasis on command raddii, rather than on cover, especially for the attacker.

Having said that, the layout of defence set-ups could also use a bit if a shake-up especially around the placement of wire and mines (and trenches in CMBB)

Please note that these comments are made by an ardent fan of THE greatest game ever and someone who wants to maintain the solitaire aspect of CM on at least an equal footing with head-to-head, through the use of QB's

[Jaldaen]

<BLOCKQUOTE>quote:</font><HR>Originally posted by James Crowley:

Please note that these comments are made by an ardent fan of THE greatest game ever and someone who wants to maintain the solitaire aspect of CM on at least an equal footing with head-to-head, through the use of QB's<HR></BLOCKQUOTE>

Hey James,

I've got a nice little single player scenario you might enjoy, check out the "Playtesters need(ed) for first scenario.." thread over on the Scenario Board. I've got some good feedback and would love to hear from a fellow single player junkie.

Good Hunting,

Jaldaen

[WWB]

I strongly second this motion. If the AI would just setup platoons together it would be much, much better. As it stands now, it scatters everything then spends the first couple turns re-organizing, etc.

Another thing that would help it alot would be if it could understand how to use support weapons such as mortars and mgs. Currently it uses them to charge with the rest of the light brigade, so to speak, where they would be much more useful hanging back in a support role.

And the other big thing it could use is some kind of instruction not to lead with armor. Now, the AI just moves the everything out, resulting in the faster units (i.e. vehicles) leading the way. And we all know what happens when armor leads. .

WWB

[ 06-22-2001: Message edited by: wwb_99 ]

[Jaldaen]

BTW- I agree with James and WWB on this one... and would like to add one more thing:

I noticed that the AI only seems to give move orders a turn at a time... there is almost always a 10 second delay in their movement at the beginning of each round... Not that I don't mind it seeing as it allows me to chew them apart for those 10 seconds, but it would be nice if the AI didn't have those movement pauses in there...

Take Care,

Jaldaen

[CombinedArms]

Great work, Mr. Crowley! The problem seems to be with the setup of attacks. A curious fact is that the AI does seem to do a reasonably decent job of setup on the defense--or is it just that the same pattern for setup works better on the defense? That is, setting up toward the rear and in cover wouldn't be so bad for a defender. Maybe somebody should check out what the pattern of AI setups is, in defensive QBs. Also in MEs.

I usually play prepared scenarios when I defend or in QBs tend to attack or play MEs. I'd recommend that if you like to defend against the AI, those prepared scenarios really are better. Since the AI has trouble organizing an attack, it's better to let a human do the initial placement. Some of the old Squad Leader conversions aren't too big.

[Jaldaen]

<BLOCKQUOTE>quote:</font><HR>Originally posted by CombinedArms:

Great work, Mr. Crowley! The problem seems to be with the setup of attacks. A curious fact is that the AI does seem to do a reasonably decent job of setup on the defense--or is it just that the same pattern for setup works better on the defense? That is, setting up toward the rear and in cover wouldn't be so bad for a defender. Maybe somebody should check out what the pattern of AI setups is, in defensive QBs. Also in MEs.<HR></BLOCKQUOTE>

From my scenario design experience, the AI likes to set up around the flags, and prefers trees to buildings (I have yet to see the computer set up a unit in the second floor of a build... has anyone seen this?). If given the oppurtunity it will split up platoons across the board and almost always in the same spots if you give them a wide set up zone, strangely enough the AI seems to like certain set up sites more then others and almost always sets up in these locations, as stated earlier trees are their favorite hangout... Note: this is a generalization, but seems to fit most of the AI's infantry set-up. As for MG crews, and sharpshooters they seem to like smaller buildings to two floors ones...

Just My Two Cents,

Jaldaen

[JasonC]

I have in the past suggested various improvements to the AI routines. This is a good addition. The obvious reason for the assymmetry on defense is more room to set up, and thus more cover to set up in, etc.

I think the AI could be vastly improved if some way can be found to give it a list of "don'ts". Mechanically applied, a few rules of thumb would make it a much stronger opponent.

1. don't put infantry squads out of command range. Trail company HQs to pick up stragglers. Use group moves by platoon to maintain "station".

2. don't send vehicles close to already dead friendly vehicles. Try another place 100 meters to either side.

3. don't fire an entire artillery module at the same location. Use 1/2 to 1/3 of it, then switch targets. Call fire missions soon when an enemy is spotted by anyone, whether the FO has LOS to the target or not.

4. don't move AFVs worth more than 100 points ahead of the farthest friendly infantry unit. Light armor is OK, but the big guys need protection from AT teams.

5. don't move slow dismounted teams across 100+ meters of open ground (slow means anything less than "fast"). "Hop" or "stop" in covered locations.

6. don't move a slow weapons team that has LOS to an enemy unit. If possible fire instead. Only exception is flamethrowers.

7. don't lead with HQs. Pause them 1-2 times to trail their units. When enemy aren't spotted, lead with a single squad one "pause" ahead of the rest (i.e. 1 squad without a pause), then the main body (with 1 pause), then the HQs (with 2). When enemy units are spotted, pause only the HQs and only once. Plot 2-3 minute moves.

8. don't open fire immediately with towed guns. Hide and use ambush markers instead.

9. don't bunch up on objectives. Target spotted enemies, or support from 100 meters on either side with about half the force available.

10. don't run under, or sit under artillery barrages, unless in stone buildings. Run to the rear instead, and return after 2-3 minutes.

I don't know if the AI can be "taught" to follow such rules, but if it can, its play would improve enourmously if it simply avoided such obvious and predictable "unforced errors".