Jump to content

STILL too easy...


BDW

Recommended Posts

ME vs AI is now 7-1. Thought you'd be happy to hear that, Fionn. smile.gif

It was a bizarre scenario:

1) I was the Germans defending the town, and my flak gun, the one way up on the hill on the "bottom" of the map, the one I was relying on to make quick work of the enemy tanks, was taken out by machine-gun fire from the other end of the map - the FAR end of the map! Steve, is that POSSIBLE? Or is it a bug of some sort? The ENTIRE gun crew got wiped out - no survivors! It was freaky. There was no explosion or anything, just incoming machine-gun fire from "infantry?" from 1000 yards away. It was especially weird when you consider the pounding the other flak gun took (see #3 below). Can someone explain to me what the heck happened to that gun crew? Did they have a group mind-snap and commit suicide?

2) I emptied all the ammo from one mortar onto the top of an enemy tank that was clearly in my LOS. I watched at least 5 rounds land directly on the tank. No damage whatsoever.

3) my second flak gun I left in the default position pointing down the road coming into town. It got off the first shot against a tank coming down the road, missed, then had a three-turn shoot out with the tank. My gun crew actually was taking cover for the second of those three turns. I thought they were gonners for sure. The tank shelled them for an entire turn! Then, lo and behold, on the third turn they remanned the gun and began returning fire! I cheered! It was a great, dramatic CM moment. Unfortunately, the last two rounds exchanged knocked out both the tank and my flak gun at the same time. frown.gif

No problem, I thought. I'll nail those tanks with my panzerfausts. Unfortunately, the AI was, um, smarter than I expected, and kept those two tanks way back up next to the trees. I tried sneaking my panzerfaust guys up the sides, but they got discovered and killed. Those tanks just sat there pummeling everything in site. When I retreated my troops for a final stand in the core of town, the tanks slowly moved in for better firing positions.

Basically, the battle was lost in the first five turns for me - when both flak guns were history. I guess if I was playing a campaign game, the thing to do would have been to get the hell out of there at that point and save some lives.

Ah, what a relief to finally lose to this AI. I gained new respect for it - even though it did get LUCKY...

[This message has been edited by BDW (edited 11-16-99).]

Link to comment
Share on other sites

BDW,

Mortars shouldn't take out tanks in real life. Tanks had a lot of armour on their top (usually) specifically to help them survive mortar hits .

As for the AI getting lucky to take out the 88s wink.gif. Yup, in exactly the same way as human players need a bit of luck to take out the 88s too wink.gif

One interesting point is that the AI was able to beat you on the EIGHT playing. If it was scripted etc then it wouldn almost certainly not be able to do this since you'd know the script by now.

------------------

___________

Fionn Kelly

Manager of Historical Research,

The Gamers Net - Gaming for Gamers

Link to comment
Share on other sites

Just a comment on my experience with the AI in Last Defense the first time I played that scenario. I was the U.S., and on the turn the TDs came in I left them sitting on top of the hill. Dumb move. The Tiger was at the top of the hill at the opposite site of the battlefield, and the Tiger flamed two TDs with its 88, and the crew of the third TD panicked and abandoned it. Then I remembered what I had read about the 88 being effective at long ranges! Needless to say the AI creamed me in that scenario. So I have a lot of respect for the AI in CM!

Link to comment
Share on other sites

As an experienced ASL player and an average one at that, I have to agree with gauchi. The AI is mediocre at best, but it's worlds better than the crap that most other so called wargames call AI. I was pleasantly surprised to see the computer try to adjust his defenses before I took an objective. But, the first time I played every scenario, I won a total victory taking about a platoons worth of casualties versus 90% losses on the other side.

The AI does not appear to implement MET-T or the older US Army Tactical Principles very well. Although the AI puts up a stiff fight, I think it could be better. I noticed the AI makes the following mistakes all the time.

On Defense:

1. Fails to identify or adequately defend critical terrain if the computer is allowed to setup.

2. Main Line of Resistance(MLR) is to far back.

3. Does not establish sufficent defensive depth.

On Offense:

1. Approach marches are generally too slow allowing defender time to identify the main line of attack and adjust.

2. Approach marches do not make adequate use of terrain to cover movements.

3. Smoke is not used to screen critical offensive movements in open terrain.

4. Poor coordination between supporting arms, attacks piecemeal.

5. Fails to establish bases of fire to support offensive moves. Germans do not use "Nail" tactics.

6. Tries to attack on multiple axes when it has inadequate resources do so diluting its combat power.

7. Suffers from the horizon effect. When enemy units go out of the LOS, the computer seems to forget about them.

8. Fritters away units needlessly. For example, Tanks duelling AT guns at long range and infantry getting into extended firefights.

I think if BTS were able to close some of these holes, the AI would be better tool to practice on. But, I really doubt there is enough time left before release.

BTW inspite of this rather harsh and pompous post, I am pleased by CM so far wink.gif and I'm looking forward to having my 'fangs' pulled by the people on this BBS.

Jeff P.

[This message has been edited by Jeff Pattison (edited 11-16-99).]

[This message has been edited by Jeff Pattison (edited 11-16-99).]

Link to comment
Share on other sites

We should all pay some attention to Scoops commments re: leaving is TD's on the Hill. There have been so many commments about the TD being too good, ect. Here's three gone to the bone yard. Why? AI tactics held the tiger in reserve at the top of their hill. Not bad! The game plays differntly much more often than any game I've encountered. To make assertions with two scen played over and over is just not a good idea. I keep thinking of the quote "contempt prior to investigation" I know I need to relax and wait for the game

Link to comment
Share on other sites

The only time, I have seen the Tiger go to the top the hill opposite like Scoops says was when I was running a concealment test on the AI. That seemed to happen only when the GI's were not hidden. The Tiger appeared to be going to a position to get a good angle.

When the GI's were hidden, the Tiger and the rest of the armor blundered down the road into an ambush ahead of their infantry support practically everytime.

But regardless of where the Tiger is, as long as the TD's keep moving the best I have seen the Tiger and rest the armor do is kill one TD. On the other hand, if the TD's stand there and duke it out, they lose most of the time. The AI usually seems to stop the TD's on top of the hill.

But, what does that really say about the AI?

Smart, Dumb, or just unlucky?

Jeff Pattison

[This message has been edited by Jeff Pattison (edited 11-16-99).]

[This message has been edited by Jeff Pattison (edited 11-16-99).]

Link to comment
Share on other sites

Hi Jeff: Just MHO: the AI is responsive to the sits rep and responds. When The Us is not Hidden but open up w/ everything its got (not necessarily a bad thing) the tiger will stay up where is weapon can be laid on more targets. Hide the US Including MG and mortar and sooner or later the germans will come with its armor up fron... a sound strategy in an unknown situation. Seems clear to me that the AI is thinking rather than just heading for the victory Hex! Not bad, says I

Link to comment
Share on other sites

I think Jeff's post is very good.

I might quibble a little about whether the AI is really wrong to move slowly on the approach march since I feel that allows it to have fresh troops who can sprint when in combat instead of weary guys who can only walk and a few other bits and pieces but mainly I think he's hitting some good points.

I don't think that the AI in CM can be best evaluated solely on the basis of these demo scenarios (although that's all you've got I know). Every tester has been beaten by the AI now. I've lost two battles to it so far. Sure I've won a lot more but I am VERY rarely beaten by any AI in a wargame unless it has like 5 times the units or something stupid.

The key to discussing CM's AI is to realise that it plays like an average human player. It will use support weapons correctly etc in the main BUT it isn't going to have flashes of tactical brilliance all that often.

The best way to put it is that it is going to be a solid opponent which will virtually never be a walkover but probably isn't going to beat you consistently.

On the other hand if I played most other wargames and gave the AI equal forces it'd be a massacre almost every time with the human player winning easily.

CM's AI isn't perfect or even as good as a good human player BUT it's good enough to beat any player who gets careless and that's more than can be said for most.

I think Jeff brings up the need to compare it to other AIs well. Not perfect but better than most (maybe all?) non-scripted AIs.

I call most AIs "Assinine Intelligences" due to their complete incompetence usually wink.gif

------------------

___________

Fionn Kelly

Manager of Historical Research,

The Gamers Net - Gaming for Gamers

Link to comment
Share on other sites

Fionn,

I agree with most everything you say. But, I think that the faster approach march I was avocating requires more explaination. I would

think that it would be counter productive to have your troops at the point where you make initial attack exhausted. I tested that, they

got massacred.

What I have been doing is testing my ability to move infantry along the same path that computer moves. Across open ground, I'd run to minimize exposure to long range enemy fire(mortars, AT guns, and MG's, etc.). Through concealment terrain, I'd sneak hopefully minimizing my chances of being ambushed and rest my troops. I found that I could cover the same terrain and launch a fully coordinated company sized infantry assault with relatively fresh troops in about 50% to 66% of the time of the AI took to launch its first uncoordinated assault.

In any scenario, launching a "good" attack sooner than later is better, because you have a chance of getting better odds by either beating the reinforcements or forcing the defender to make a quicker and maybe wrong threat assessment.

What I'm asking BTS is that the AI at least match that speed in an infantry assault. What I noticed is that CM seems to goof off in open terrain for a couple turns where I can mortar him before he commits to an axis of attack.

Hi Mark,

I don't know what tactical manual advocates leading armor through possible armor ambush kill zones, such as turns in the road with woods on either side that can be full of bazookas or panzerfausts. Doctrine states that infantry should dismount and clear the ambush points as rapidly as possible. And then, the tanks should move on.

Since AFV's are such valuable units and maintaining armor superiority can often be the basis for winning or losing, I'd hope you 'd want to keep them out of harms way until the critical moment arrives. I'd like to find out what threats to my tanks exist up ahead with my infantry before committing my tanks to an advance. But, you can be too timid, and sometimes you have to break a few eggs to make an omlet. wink.gif

Jeff P.

[This message has been edited by Jeff Pattison (edited 11-16-99).]

Link to comment
Share on other sites

Well I see I've neglectated to mention the infantry dismounting from APC and moving carefully to flank the roads, get a view of the little down turn dog-leg turn. Keep in mind that the milling of forces takes what 2 min--3 min. Doesn't seem to be a terrible waste of time. Frankley if you are thinking about an ambush around the next bend what better mode to use than a careful combined arms approach from two or more postions (such as the road and the woods. That's what I've seen the AI do...again not bad, me thinks. To put a point on this... The AI does a darn good job of acting like a military organization. So many games I've played (and enjoyed for that matter) can offer only an AI triggered by victory hexes, attacking in odd clumps of hardware (fastes and nearest come first) This IS NOT what happens in CM.

Link to comment
Share on other sites

Jon,

I know regarding the weariness :-(. In a PBEM game at the moment my guys became weary JUST as they were within 200 metres of the village GULP!.

That's when the mortars and MGs opened up and they could only WALK anywhere. Shudder. lost 8 men due to that frown.gif

I subscribe to the hit them hard and fast and accept a few casualties to maintain momentum.

------------------

___________

Fionn Kelly

Manager of Historical Research,

The Gamers Net - Gaming for Gamers

Link to comment
Share on other sites

Guest Big Time Software

Hi Jeff, good post. I want to assure you that we do intend on tweaking at least one more thing before we ship:

Keeping squads together better on the attack.

This is the root problem to nearly everything on your attack problem list. We have made improvements, but AI work is time consuming and very often you break stuff that used to work in the process, which requires even more time.

The AI *used* to keep its squads together VERY well. Little bastard would overrun my positions like a steamroller. But the problem with aggressive AI is that it is VERY open to instant destruction if it does the wrong thing at either the wrong time or wrong place. So we changed the way the AI attacks to be more conservative, and generally this has increased its chances of winning (or at least not being slaughtered). However, there was a side effect in that platoons are hesitant to all charge at once. This is what we need to fix.

Unfortunately, the level of intelligence that would be needed to have a grasp of things as simple as "MRL" are beyond reach for now. AI is exponentially CPU intensive when it comes to stuff like this, so no go for now. There is a decent degree of asset coordination, but it is still largely coordination through circumstance when looking at the big picture.

The good news is that CM's AI is better than others out there, especially for a game as complex as this one. Over time we will improve the AI and as CPUs get faster we can have it do things that today aren't practical. Thankfully Charles *likes* AI coding, even if it is a frustrating and sometimes thankless job.

Steve

Link to comment
Share on other sites

This is somewhat addressed in BTS's last posting, but I have seen a number of cases where teams will follow one another into kill zones blindly (Incidentally, I've only seen this when playing vs. the AI - not against people - I think this has something to do with how people coordinate assaults vs. the computers method - dunno). Admittedly if the computer kept multiple squads together better it would improve the situation (i.e. the assault would be coordinated with three or four teams rather than one at a time) but I was wondering if there were a way to have a 'local' morale effect for a team that has just seen someone in front of them slaughtered. I.e. say you had four teams assaulting across a crossfire zone. Even IF they did not perform a coordinated assault, even IF say the first team goes in, gets slaughtered, second team goes in, gets slaughtered, could there not be some way of having a local effect on the units that are waiting to attack across that space - something to the effect of the squad leader observing that the last two squads got their butts handed to them, and maybe it would be better to wait for orders (say, at the 60 second mark...) than to try the futile assault. Just seems to me the third and fourth team would have less interest in charging blindly into what was obviously an ambush.

I don't know if this is the best answer, but it seems to me there should be SOME effect on units attempting assaults if they see that the last 12 guys who walked across there got popped in the first 20 feet.

I hope this is clear. I'm kinda tired right now so I'm not at my most coherent....

[This message has been edited by A Arabian (edited 11-23-99).]

Link to comment
Share on other sites

×
×
  • Create New...