AI adjustments

[Mad Russian]

Originally posted by McIvan:

Well, the question is more like this:

Does CEB mean the AI implements smart plans instead of stupid plans, or does it just mean that the troops are more capable of implementing exactly the same stupid plans as before.

Does that equate to the same thing as does it play better because it's actually smarter or does it play better because the units are better?

Isn't that just another way of simply asking does it play better period?

Well, I think I've made my position on how I feel about it fairly clear.

What do you think?

MR

[JasonC]

MR - I see. What you actually mean is that the AI is somewhat harder to defeat on high + settings. Duh. It is also harder to defeat the AI if you start the human with one green infantry platoon and give the AI 40 tanks, or a 50% fanatic crack battalion. But no, this does not mean that is "better".

Will you see more orders executed on the high + or on high quality settings? Sure. In short scenarios time-wise, in particular, since the move delays go down. But what you pay for this is, none of the usual firepower and robustness match ups will work as they do in a real game (meaning human play).

In vehicle combat, doesn't much matter - the higher qualities will just move out faster, engage with less cowering, and hit more often. In combat in which the infantry matters at all, it does matter. MGs don't deny open ground the same way against elites, odds match ups don't mean the same things, ammo likewise. Tactics you learn for +2 or +3 games simply won't apply in a real game.

What actually works to make the dumb AI challenging is long odds, carefully arranged set ups and flag placements, and shoestring-ish human forces that force the player to get something out of each team. But those will work without quality falsification.

[GreenAsJade]

JC, the claim that MR clearly made is that the AI does smarter moves when you give it a CEB. He said:

It will do such things as flank you, rush the flags, stalk your tanks, do combined arms attacks. etc

He also said that manually changing the quality of the troops does not deliver the same results.

These are the sorts of observations that I recall reading in the past.

Are you saying that these are _not_ the case?

GaJ

(I feel an experiment coming on!)

[Mad Russian]

Originally posted by JasonC:

MR - I see. What you actually mean is that the AI is somewhat harder to defeat on high + settings. Duh. It is also harder to defeat the AI if you start the human with one green infantry platoon and give the AI 40 tanks, or a 50% fanatic crack battalion. But no, this does not mean that is "better".

Will you see more orders executed on the high + or on high quality settings? Sure. In short scenarios time-wise, in particular, since the move delays go down. But what you pay for this is, none of the usual firepower and robustness match ups will work as they do in a real game (meaning human play).

In vehicle combat, doesn't much matter - the higher qualities will just move out faster, engage with less cowering, and hit more often. In combat in which the infantry matters at all, it does matter. MGs don't deny open ground the same way against elites, odds match ups don't mean the same things, ammo likewise. Tactics you learn for +2 or +3 games simply won't apply in a real game.

What actually works to make the dumb AI challenging is long odds, carefully arranged set ups and flag placements, and shoestring-ish human forces that force the player to get something out of each team. But those will work without quality falsification.

That's exactly what I mean. The AI is harder to beat with a higher CEB. That was the question. You even agree with the statement even though you don't use it.

I have no idea how you would ever tell if the computer would or will execute more orders per game. What I do know is that the computer does other things I consider "smart". Such as flag rushes, flanking moves, using combined arms attacks, less dance of death maneuvers by tanks...etc...that I can see happen. Those are easily identifiable.

As you say...DUH....to me that is better. If you want to break down the way the computer does each order and if that sequence is now in a more favourable light I can't help you. I can't prove that. I would be surprised if you can do any disproving of that either though. What I see from you is your usual assumptions. You have no facts, figures, or firm evidence of anything. Just that you don't use the CEB and therefore it's not worth the effort for anyone else to either.

Again, if you don't want to use it then don't. I recommend that those that play my scenarios use it. They seem to get very good results using the CEB. I get no complaints.

ROFL!!!!

I love this statement of yours....

In vehicle combat, doesn't much matter - the higher qualities will just move out faster, engage with less cowering, and hit more often.

It doesn't matter much? Higher experience levels just makes them move quicker, cower less and hit more often? That's all? IMO, that's plenty in armored combat.

What else is there for vehicles?

To be honest though, there's quite a bit more than that. The higher level CEB makes them stalk your tanks, use keyhole positions, make combined arms attacks, in addition to just being better at the basics.

Since you don't use a CEB how would you know what it does or doesn't do? This level of discussion is strange. You profess to not use it, yet continue to tell me, and all involved, all there is to possibly know about the subject.

If you don't want to use it that's fine with me. More power to you. As for me and my scenarios I will continue to recommend that those who play them use it. Not because I "think" it works but because I've tested it and seen it work. I've seen the results of the reviews on the scenarios as well.

Can an experienced designer make a better scenario vs the AI? No question about it. Is one of the tools that make that scenario play better against the AI the use of the CEB? IMO, no question about that either.

MR

[Mad Russian]

What actually works to make the dumb AI challenging is long odds, carefully arranged set ups and flag placements, and shoestring-ish human forces that force the player to get something out of each team. But those will work without quality falsification.

The vs the AI scenario is capable of certain things a H2H scenario can't do.

As you correctly identified, there are certain situations that the AI excels in. High unit density actions on one side that a human player would crush his opponent with can be shown quite well. The issue of stupid comes into play as well.

You can actually have the AI be as dumb as a box of rocks if that's what's called for. Or you can smarten it up with the CEB. The question was does the AI play better with a higher CEB? Does it do better? The answer, even by your own words, is yes.

How do you explain the difference in gameplay from an OOB full of elite troops with 0 CEB to the same OOB and a + CEB? According to you there should be no diffence. That's not what I've found in the past when I tested the CEB.

MR

[JasonC]

GreenAsJade - right, I am trying to get him to claim something concrete enough that it has testable empirical content, instead of vague "I thinks it has eaten its wheaties".

He doesn't seem to notice at all the point about false weapon match ups and tactics. Perhaps because he is thinking almost exclusively of vehicles, I suppose.

He also can't seem to grok "harder isn't better", or that harder is easy to get without it offering an iota of better. If you manage to re-create Space Invaders inside CMBB, the closeness of a given round of it won't make it a better sim or test of tactics or recreation of tactical match ups or the achievements of those involved.

The accuracy of the game is essential to it being worth anything, and if ubermenschen charge through gales of machinegun fire like a Sgt Rock comic book because they can't be harmed, any "harder" that gives is utterly worthless, and not at all "better".

I now see scattered approaches to actual claims - keyholing e.g., or flanking maneuvers. I am prepared to utterly deny such claims.

As for what I know of it, of course I have tried + computer settings - and abandoned them. I also concluded long ago that real tactical interest in CM occurs between humans, because the AI is very dumb.

I've written at length on the situations in which the AI can give you a fight despite it being dumb - e.g. it has superior armor and your AT forces are shoestring, or it has 3 to 1 and up odds, so you can't afford any of your units to fail at their jobs, etc. But it is still dumb. Demonstrably dumber than any competent human. On any setting.

My challenge for anyone who thinks otherwise is to make the game you think the AI can win against a human (say, me) on +2, that a human, (say me) can't win on 0, when given the AI's force and start instead of the human's.

MR thinks the + experience setting is like a chess computer difficulty level, and that the AI is crunching numbers longer or better inside to make more brilliant decisions. And I deny it because I see it make quite similar stupid AI trick mistakes regardless. Can a scenario designer mitigate some of those, with odds and starting positions and the way the flags wave in front and beckon the AI in? Sure. But it is mitigation - no human with the same resources and half a brain, will "underplay" it, on any setting.

[JasonC]

How do I explain the difference between elite and + CEB? I deny there is one. Simple. If you play several times you might see a dispersion of results, and you might be tricked by randomness that way - or by seeing what you expect to see. But it isn't there. I deny your experience in the matter has any cognitive status, any more than your insights gleaned from your horoscope.

[Mad Russian]

Originally posted by JasonC:

How do I explain the difference between elite and + CEB? I deny there is one. Simple. If you play several times you might see a dispersion of results, and you might be tricked by randomness that way - or by seeing what you expect to see. But it isn't there. I deny your experience in the matter has any cognitive status, any more than your insights gleaned from your horoscope.

At least you're consistent, I'll give you that.

You have nothing and at least this time you're not even trying to make things up.

Not sure what you think you have that makes you all seeing all knowing. Nobody ever said the AI was as good as a human player. That wasn't the question now was it?

But you know what, you win....

You're absolutely right. I've not seen a difference in the way the CEB adds to games. I've not had reviews that back up that the CEB makes the scenarios play better if you use it. I've not put them on here with the scenario names and location so that what I put on here can be checked.

No, what really needs to be listened to is JasonC and his vast amount of knowledge on all subjects great and small. Where you quote no testing, have no scenarios that you have put out for the CM community to play and rate for you. Where you have admitted that you don't even use the CEB and don't play vs the AI anyway.

It's not at all unusual that you wouldn't agree with me. You've called me a Troll, a Nazi Fanboy, told me that I'm full of horsefeathers and poppycock...so why would I not expect the same out of you now?

I'm still waiting for you to write that all knowing and ever popular best seller...

Jason C's Version of WWII: The Grand Strategic, Operational and Tacital Facts of WWII. The bible on everything anyone would ever want to know about WWII, and this case CM. Maybe you could just include a chapter on CM as a kind of reward for the rest of us.

That could take the place of all the missing resources that you never quote but demand of others. It would certainly make it less of a mess when it comes time to back up any of the "stuff" you put out there.

You are VERY GOOD at discussing "The Drill" and what the TO&E's are.

Not so good if you have to produce any kind of references for your assumptions.

Works for me.

As I told you before. If you don't want to use the CEB...don't. I'll continue to put the disclaimer in my scenarios that it should be used because it makes a difference in the how the game plays. That's my opinion. I at least have some reason for thinking that way that I have some evidence of actually trying to find out the actions taken in the game. Not just my opinion that I'm selling as the, "Be All - End All" of the snake oil market.

Where you say I can't prove what I say you can't disprove it. Over the course of bunches of vs the AI games you find whatever will make a game play better. You have already told us numbers of times how humans play better. We got it. This is a discussion about if the CEB helps the AI play better. I say it does you say it doesn't.

That's good enough for me.

Have a good day.

MR

[ December 13, 2007, 07:47 PM: Message edited by: Mad Russian ]

[Kineas]

I also think the CEB only modifies the TacAI, makes veterans from your regular troops, nothing more.

Of course the resulting gameplay will change a lot, but it's the same change if you had manually modified your units in the editor. The 'Strategical AI' will be the very same.

I think this because the StratAI development is costly and a major selling point, why would BFC hide a lot of AI tricks behind a rarely used combo box?

We won't know the truth unless the devs tell it or someone can create a test scenario, where the AI can show new tricks on the higher CEB levels.

[Kingfish]

Originally posted by Mad Russian:

Let's just take that comment right there at face value. If you have no doubt the AI gives you a better game with the CEB increased isn't that what the original question was?

No, the original question was:

Originally posted by rocketman:

In what sense does the adjustment to AI (+1,+2,+3) change the game? How will you tell that the AI got "smarter"? What aspect of AI decisions will improve most significantly?

Does it change the game? Yes.

Does it make it smarter? No.

Consider that it is called Computer Experience Bonus, not Computer Intelligence Bonus.

[Baron Ritzner Von Jung]

Just play drunk , you'll find the AI amazing and fun too !

[JasonC]

The Baron has the right spirit, that is how to play the AI. Don't use its weaknesses too much, and don't expect finesse of any kind.

Everyone now having made their opinions in the matter clear, and such claims as they intend to make, I will now substantiate my claim that the AI remains as dumb as a bag of rocks on +2 CEB setting.

My chosen setting is a game specifically designed to reveal the underlying nature of the strategic AI - it is at bottom a path finding algorithm, and doesn't know tactics from its anatomy.

The Germans get 600 points of infantry force type defenders, veteran quality. The time is May 1942, since I wish to show the AI as pre tactical sophistication Russians. The Germans take -

2 sIGs

2 HMG-34 teams

1 Jager platoon

1 105mm line FO

1 TRP

2 Trenches

3 AP minefields

7 wire

The Russians get 1200 points of green infantry, which the +2 CEB will bump to veteran morale - whatever else it does for the AI commander. The Russians get -

2 infantry companies (B type, each platoon 4 11 man squads)

2 82mm mortars

4 50mm mortars

8 Maxim MG

8 ATR

1 76mm line FO

2 regular sharpshooters

4 extra platoon HQs (stripped of squads) to command heavy weapons sections.

The terrain is rural open gentle slopes, aka steppe like, with the following specifically tailored configuration, again meant to show the specific nature (and weakness) of the AI, regardless of CEB setting.

Here is a turn 1 view -

The Russian set up zone is a sizable block of scattered trees, sufficient to cover them all.

There are two thin routes forward - one a single tile wide line of scattered trees bisecting the map, leading forward several hundred yards - with occasional slightly thicker sections - to a "bulb" of scattered trees 100m or so across, at the end of the "stem". A single small flag is smack in the middle of the bulb.

The other route is off on the right side, and consists of a depressed "balka" below the level of the surrounding field and lined with brush. It leads to a short distance away from a rocky hill on the Russian right.

There is a similar rocky hill off on the Russian left, without any route to it.

Dead ahead of the "bulb" is a large patch of rough on a dominant hill, which has a single large flag. There are lines of scattered tree off to either side of it, the left side being on the forward slope and the right on the reverse. Slightly higher "arms" extend from the rough, center hill to the two flanking ones.

To any sentient being, this screams kill-sack in the "bulb". The right side balka is a better route, since it will allow taking first the right side rocky hill, then using that as cover, the scattered trees on the far side of it being dead ground to the rest of the defensive terrain, allowing the center hill to be flanked from the right.

To a clueless "try everything" dumb major, you could (1) send a company along either route and find out which one works better empirically or (2) ignore the routes and step out on a broad front, through open steppe, to avoid bunching up too much, or (3) have some of the heavy weapons and a platoon from each company do (1) while the rest of the infantry does (2). At the least, you could (4) set up the heavy weapons in the starting wood line, and overwatch the advance, any way it is conducted.

Does the AI do any of these things on +2 CEB?

On turn 1, the AI fired off its FO at an aim point a little shy of the small flag. It lands for 4 minutes. Nobody is there.

Here is a view on turn 8, during the approach march.

Notice that the AI is bunching up along the scattered tree "highway". Not a single unit has stepped into the open more than a few meters on either side of that line of trees. No one ever takes the brush balka route on the right. The AI will in fact send every squad and every weapons team, right up that center pipe, 20 meters wide, straight to the kill sack.

There are two "cuts" made across this highway, in the form of single tiles of AP minefield that cover the entire scattered tree passage. The AI sends unit after unit right through those mines, detonating some repeatedly. A few units skirt them a few meters into the open on the right, and a German MG34 harasses them as they do so.

This creates the first visible "pile up", and one of the sIGs tosses in a few 150mm shells, between the first AP mine "cut" and the second.

By turn 12, the AI has pressed ahead into the "bulb", and boiled out beyond it. 60 meters beyond it is a line of barbed wire, with a single hole at the center. An AP minefield plugs that hole. The AI boils out into the mines. Following units go through the wire just to the right of the mines.

The bulb is of course right under a TRP, and the German 105s fire on it once there are enough stars accumulated in the bulb. Both sIGs are firing. The German HMGs hit anyone who steps into the open on either side of the trees, but precious few actually do so, until clear through the wire arc. Nobody goes around the wire arc.

The worst fire is taken by the German sIG on their left, the first to open fire. As many as 3 Maxim MGs fire at it from ranges of 250 to 300 meters, and push it as high as "cautious" a few times. No German will actually be hit in the entire scenario, and not a single Russian mortar round will be fired.

At turn 19, the bulb is a ruin of scattered body parts. So is the minefield "plug" in the wire barrier. Scattered routed units have spread a bit to either side of the bulb and kill zone, but no Russian unit has made it within 100 meters of any German, the entire time. The Jagers have used less than half of their ammo, firing only at the wire barrier of men already through it.

The Russians continue to die in droves until the remainder flat surrender at turn 26. Here is what their gloriously commanded special +2 CEB supersmart "formation" looks like by then -

The Russians lost 335 men, plus 64 surrendered at the end. 7 ran off the map alive. The kill totals for the German weapons were -

sIG left side - 134 men and 6 mortars

sIG right side - 76 men

105mm FO - 47 men

MG34s - 23+8 = 31 men

Squads - 14+11+7+2 = 34 men

Mines (deduced, remainder) = 13 men

German losses - not a hangnail.

2:1 point odds and +2 CEB.

The AI did not flank through the balka on the right.

The AI did not overwatch with its mortars, FO, MGs, ATRs, or sharpshooters. Several of the mortars were found at the end, KOed by one 150mm round as they tried to push through the first AP minefield - long since detected of course - straight up the middle. Ergo, even on +2 CEB the AI has no idea what a mortar is or what it is for.

The AI did not spread its men across the frontage. Instead, with a single cover-created "optimum" in its path finding routine, it shoved every man it had right along that single-solution path. The AI doesn't even have the typical collision-avoidance spreading bias seen in typical (cartoonish) real-time strategy games. Even on +2 CEB.

The AI did not use its FO or mortars to "counterbattery fire" the German on map guns. On no plus setting, the AI might not have even used the FO at all - the best that might conceivably be said for the setting changing behavior, is that instead of marching the FO into a minefield to die (maybe), it fired it all off on turn 1 at the nearest (empty) flag.

Now, sure game designers can arrange things to the AI works better than this. They can avoid primrose path flag strings, and channeled terrain. They can provide the AI with multiple starting "angles" to the human force, so even path finding this deterministic will produce several routes of approach instead of one. Above all, they can give the AI some armor, which being less concerned with cover calculations in its pathfinding, isn't quite so horribly weak and predictable. And can make things harder for the human by giving it more tanks or thicker fronted ones, etc.

But all of that is arranged mitigation to cover for an AI that is at bottom, not a descendent of a long line of mystical warriors, but the descendent of a toaster oven, to quote pop culture. At bottom the AI is A*, the pathfinding algorithm, and not a tactician. And it remains a dumb as rocks pathfinding algorithm, even on +2 CEB.

[JasonC]

Extending remarks on the specific weaknesses of the AI I just highlighted, and ways designers can avoid stressing them...

The AI does not know to not bunch up its infantry. Open avoidance and most covered route seeking are so essential to what it is and does, that you cannot expect it to use superior infantry numbers correctly, unless you elaborately tailor the situation for it.

Wide blankets of equivalent cover help it manuever. Second best is heavy cover situations with no one route strongly favored and only modest open gaps from one cover block to the next. Even in the latter case, though, it can readily get stuck "homing" on flags, and not paying attention to obstacles, kill sacks, or other opposition in those open ground patches.

If you want the AI to hit from two directions you have to start it in two places, such that even determinstic path finding will create the desired multiprong attack. You cannot expect the AI to conduct *any* internal tasking or articulation of subunits. You have to provide all of it, by distributed set up or staged reinforcement arrival from various points.

The AI does not know what a support weapon is. The only way to get it to use guns and mortars etc effectively, is to start them right where they should spend the battle, and then to put a flag in the immediate vicinity.

The AI will advance heavy machineguns far too much, but once checked they will at least set up and fire back. Infantry will frequently fail to do so if stopped too far from the enemy. MGs have the ammo and the fp carrying to range, to provide at least some return fire, once set up. But you cannot expect them to be placed sensibly. They will just drag up in the wake of the squads, and sit down to fire only if checked be enemy action.

A team rated "slow" and expected to move during the scenario, commanded by the AI, will be quite ineffective. Give it medium speed teams if possible. Remember that they will move with the infantry they start or arrive with, with no articulation or planning whatever.

Because it does not know how to use mortars or FOs - at all really - and guns will generally only have LOS in defense situations, the AI has extreme difficulty attacking gun based defenses. The only thing it can do against them is push armor forward and hope it wins the resulting duels. If the enemy has guns, the AI needs numerous AFVs to have a chance - and its chances are best if at least some of those are invulnerable to said guns through the front aspect.

The AI is better at defending than attacking. The tac AI makes reasonable decisions about what to shoot at, and the strategic AI is least stressed when it does not have to move a mass of units toward flags, but instead already owns them. Even so, the attacker *capturing* any of the flags will frequently trigger a very poorly coordinated counterattack. This makes attention to what the AI needs to attack, important on both sides of the table.

Since the AI knows nothing of tactics, it attacks simply by trying to move its fighting forces onto the flag locations. If this is to present any danger whatever to the human player, the AI needs forces that are dangerous when they simply attempt to move forward into a defense. That means, in practice, reasonably thick tanks and reasonably high quality infantry. And the last are much more effective in continuous or wide block of cover terrain.

The infantry movement routines of the AI seek ground cover (not LOS cover, which it can't see) and do not know anything about avoiding bunching up. That is sensible only at low unit densities and against infantry type enemy fire. Then staying in the covered tile is a reasonable tactic. Against large caliber HE or at high unit densities, and especially when both are true, bunching up is deadly, far more deadly than straying into open ground.

So the AI will fight better at lower unit densities. Spread it over several axes of fighting at set up or using reinforcement arrivals. Keep the force it must coordinate along a given axis at a company or less. Really, along a single route even a company should arrive in stages, to come in waves along that route.

And the AI will be more effective against a human force that has mostly infantry fp weapons, or with the number and caliber of HE weapons strictly limited (a few on map mortars, one leIG, for example).

Note that the pathfinding is weak enough that a very small number of obstacles, placed with full knowledge of how A* works, can be sufficient to gum up the AI horribly. Be very sparing of obstacles you give to the human player in a game meant to be against the AI, where the AI is expected to attack. Consider padlocking their placement, and test how the AI will navigate near them.

In the end, though, you can nearly always find a stupid AI trick to fool its path finding. AI games are fast, and that will remain a reason to use it. But real CM games with real tactics are games played by 2 humans.

[JasonC]

OK Mad Russian, now tell me I don't know anything about the AI...

[GreenAsJade]

JC: I think the example you gave is an interesting one, but not conclusive.

You showed an example where the AI does a poor job.

This can be easily countered by an example where it does an acceptable job. That example will prove that it is not as dumb as a bag of rocks, it just has limited situations where it can be competitive.

The other thing that the example you gave fails to prove is whether +2CEB makes any difference in a 'reasonable' situation.

I would rate a 'reasonable' situation as one where the AI has a chance in the first place: one where it's not set up to fail in a situation of known weakness.

If there was a situation like that, and the AI played as dumb on +2 as on normal, then I'd be convinced.

The sort of test that would be really interesting is one where there's an opportunity to flank ...

GaJ

[JasonC]

And you don't think a brush covered sunken highway to a vulnerable right is an opportunity to flank, right.

The AI is A* with little else to help, that was the point of the example. And it is still just A* with little else to help on plus settings, which makes no difference at all in its basic operation, its inability to avoid bunching up, its handling of support weapons, its reading of obstacles, etc.

So here is the challenge. What exactly is it that you think the AI is ever capable of doing, that you think it can do better on plus anything? I mean, the ludicrous claim that that brain dead thing can calculate keyholing and flanking etc - it flat can't. A designer can set things up so its deterministic behavior leads it an intelligent way - but the intelligence is the designers, and not the AIs, and does not turn on the CEB setting.

[GreenAsJade]

So here's the thing.

Why is there a CEB.

I'll tell you why I think there is one.

It's so that a given scenario can be played by playeres of different ability. All games have a feature like this, and so the CM designers needed to provide one. They chose this way.

Now suppose that the only thing it does is increase troop experience level.

Great - that achieves what I describe above _and_ the effects that MR has described to us. The computer is harder to beat with higher CEB, which is the bottom line of MR's posts, and answers the OP's question.

The fact that the AI is not as clever as a human and needs lots of good designer tricks to produce good play (which is a paraphrase of JC's posts) is granted by all as well. FCOL, this is what makes good designers for vs AI games "good" ... the ability to use those tricks to produce fun scenarios.

And what MR is telling us is that the AI needs to be given CEB +2 to make his designs work. That seems indisputable too!

So to answer the OPs question: it does make a difference. There's a debate about how and why it makes a difference, but it does. So much so that designers tell you what CEB they designed for.

It may be that this boils down to "the AI needs crack troops to cope with its otherwise not-so-good thinking" but hey ... so what?

(I will note that the CMAK manual tells us at least what the designers were willing to say about what CEB does).

Moral of the story: if you're playing scenarios against the AI, check what CEB the designer recommends.

Cheers,

GaJ

[JasonC]

Nice - you point out that the manual says something on the subject - in fact something transparently clear - but then don't say what that something is, and even insinuate that the designers are hiding something (which they have no reason to do).

Here is what it says, verbatum -

"For an added challenge the Experience level of the computer player's troops can be increased up to 3 levels above the scenario designer's settings. For example, if the computer player has one Green and two Regular platoons a +1 setting would change that to one Regular and two Veteran platoons."

Clear as the sun.

Note further that the designer sets the level to start with. This leaves only a couple reasons to reach for CEB instead of just boosting skill levels -

(1) so the skill level will be different when 2 humans play, than when a human plays the AI (only matters for scenarios meant to be played either way)

(2) to access elite quality level in QBs (normally that is restricted, and troops top out at Crack)

(3) to decouple the point balance used from the quality levels i.e. if you set it up to have 1.5 to 1 attacker point odds, but then boost the defender several levels, you are effectively playing perhaps a 5:4 point odds battle.

Anybody who likes can pretend there is more going on or that the AI gets smarter too, but they are just pretending...

[MeatEtr]

In spirit of the Baron, this is what I prefer to use for the AI modifier.

[DavidFields]

I agree with JasonC. I believe the CEB bonuses make the AI troops more resiliant, not smarter. Given that I almost always play against the AI, I enjoy his analysis of how good scenarios are made.

Two more little design quirks that can be mined: Extreme FOW, and lack of advance familiarity with the scenario.

With Extreme FOW, a scenario designer can take advantage of the fact that a human player can be made ignorant of the silly moves the AI is making--thus preventing early countering, and making the AI look smart when units "suddenly" pop up somewhere.

As to lack of familiarity with a scenario, I usually have a special pleasure playing a scenario for the first time. My attacks are usually much more timid, and probing. A scenario designer can take advantage of this by having some surprise reinforcements, or unit placements (as long as, in my opinion, the initial briefing is not so deceptive as to be insulting)

The one point I can see in Mad Russian's favor: I could imagine scenarios where a preset flanking attack could be made such that it would easily be beaten back without bonuses, but would become much more effective with CEB bonuses. Thus one would be "adding a flank attack" with the CEB bonus.

But the point of that would be to have two levels of a scenario, novice and advanced.

Otherwise, in general, giving CEB bonuses to the AI upsets the unit balance, in my opinion. Always fighting high-level early war Russian troops, for example, would lose the flavor of the era.

[Vergeltungswaffe]

If you want to play against the AI, make sure its on the defense, and play by Franko's True Combat rules. It will actually be challenging in most scenarios.