A constructive critique from fisrt impressions

[HansBolter]

ORIGINAL: iberian

ORIGINAL: HansBolter

You are correct in your assumption that I do not oppose the AO concept. I also do not play against human opponents. What is important to me is to know if the AI exploits it's knowledge of the AO limitations. I am not arguing for the mere sake of argument. I really do like the concept of the AOs. I just want to make sure all considerations have been made and accounted for. Since the answers I get tend to address only one side or aspect of the situation I do feel compelled to point out the other side that is being omitted for whatever reason.

SSG has explained many times in the past for their previous games that their AI doesn't cheat exploiting knowledge not availabe to the the human player. I don't think they have changed their politic now.

Also, SSG has a known past of being very carefull to what changes and modifications go in the system, so it's preatty safe to assume that they have taken into account all that is reasonble, and every consideration possible.

Having been a war gamer for over 35 years and participated in wargame design and development I never make those kinds of assumptions. I ask.

While I didn't offer my reason for asking I do have a valid one. I started my second game last weekend. In my first game I attempted to build a new line in front of Kharkov immediately behind the original line, falling back as little as possible. The line was too brittle and teh Russians stormed right through it much the same as in the AAR recently posted on the SSG site and linked here.

In my second game I fell back to the river line that runs east/west near the boundry of the 29th Corps and then turns south to link up to Kharkov. The line there is solid and has completely stymied the AI. However, at the far eastern end of the line where it attaches to and hinges on the original fortified line the AI has completely denuded it's front line for a two hex distance while it appeared to have pushed everything forward. Since units behind teh lines are hidden to me I have no idea if there are reserves covering this frontage, but it has all the appearances of the AI choosing to leave unthreatened hexes empty as the units on my side of the line are prevented by their corps boundry from attacking further southward into the unocuppied front line hexes. While it may have a simple explanation, it has the look of a gamey tactic.

So, before you take some one to task and lecture them on how all encompassing the design and playtesting of a game by a professional game company can be assumed to be you might one to consider that they may have valid reasons for asking the questions they do, but have simply not yet made you aware of them.

[iberian]

ORIGINAL: JudgeDredd
At the same time, surely Hans has a point?

He had point in his first post. Then the game designer answered him, gave a reasonable explanation , and, under my point of view, he stopped having one.

ORIGINAL: JudgeDredd
I don't think you can poo poo Hans discussion. He's not keeping anything going "...for the sake of it".

Sorry, but no. Bringing to the table the need of having historical AO is discussing for the sake of it. Mistery Variants are the answer against gamey tactics by human opponents. Historical AOs don't have anything to do with it.

[iberian]

ORIGINAL: HansBolter
So, before you take some one to task and lecture them on how all encompassing the design and playtesting of a game by a professional game company can be assumed to be you might one to consider that they may have valid reasons for asking the questions they do, but have simply not yet made you aware of them.

Oh, I'm sure you have valid reasons. No doubt about it.

The problem is that AI cheating doesn't have anything to do with AO design and its purpose in the game mechanics. It's totally another issue that can only be addressed by the game designers.

Still, the AI can be designed to cheat no matter if the AO concept is present or not. But programming an AI to use AO's information, in case the user has it turned on, and programming it also not to use it, in case the user has it turned off, looks too much trouble to be a reasonable possibility.

[Gregor_SSG]

No system is perfect and all systems struggle to cope with ingenuity of gamers. The combination of Historical and Mystery Variant AOs is our way doing so, and I think it works very well.

The AOs are designed to simulate a whole range of factors that had a great impact on the battle. These are factors that we hadn't seriously addressed in our previous games.

Firstly though, can I just say that in the 'Whatsnew.PDF' doc that is installed in the Kharkov game directory, I state that the AI will always use the AOs. So it will always face the same restrictions as the human player using the AOs.

One of the issues that the AOs try to address is the problem of hindsight and perfect knowledge. We as gamers already know what happened in the historical battle, or if we don't its only a mouse click away on Wikipedia. Not only that, but we can boot up the game and play each side, thus knowing the exact OB and deployments of each side, down to the last samovar.

The AOs can't prevent such knowledge but they do restrict your ability to exploit it, putting you much closer to the position of the historical generals. The mystery variants strongly discourage trying to take advantage of any knowledge of the historical AOs, as you just don't know what advantage (and the variants always confer an advantage) the other side has, and therefore have to be much more sensible in your deployments.

Gregor

[Fred98]

ORIGINAL: HansBolter
What is important to me is to know if the AI exploits it's knowledge of the AO limitations.

You know that the enemy regiment "X" cannot move into the adjacent AO and therefore cannot attack you're left flank - and you can exploit this. Swap sides and the AI doesn't know this and will presume it's left flank is vulnerable.

-

[jmlima]

ORIGINAL: Joe 98

ORIGINAL: HansBolter
What is important to me is to know if the AI exploits it's knowledge of the AO limitations.

You know that the enemy regiment "X" cannot move into the adjacent AO and therefore cannot attack you're left flank - and you can exploit this. Swap sides and the AI doesn't know this and will presume it's left flank is vulnerable.

-

That's intriguing. To say the least. How is it possible that the AI does not factor that onto the calculations, if it is working within the same limitations?

That is, the AI knows it cannot move it's own regiment X onto a different AO, but does not factor that the opponent (human) regiment Y also cannot move? Certainly if it is something to increase realism, the AI should operate within the same constraints and knowldge as a human player, it should not have more knowledge (cheating) , but it certainly should not have less knowledge or compute less facts.

[Gregor_SSG]

ORIGINAL: jmlima

ORIGINAL: Joe 98

ORIGINAL: HansBolter
What is important to me is to know if the AI exploits it's knowledge of the AO limitations.

You know that the enemy regiment "X" cannot move into the adjacent AO and therefore cannot attack you're left flank - and you can exploit this. Swap sides and the AI doesn't know this and will presume it's left flank is vulnerable.

-

That's intriguing. To say the least. How is it possible that the AI does not factor that onto the calculations, if it is working within the same limitations?

That is, the AI knows it cannot move it's own regiment X onto a different AO, but does not factor that the opponent (human) regiment Y also cannot move? Certainly if it is something to increase realism, the AI should operate within the same constraints and knowldge as a human player, it should not have more knowledge (cheating) , but it certainly should not have less knowledge or compute less facts.

Its encouraging to see people treating the AI as if it has knowledge about the game in the same way that a human player does, as this means that its doing a reasonable job. In truth, like the Wizard of Oz, it's all done by smoke and mirrors. The AI doesn't 'know' anything about your forces, and certainly doesn't know that it has a left flank - that's a very human concept and the AI is nowhere near smart enough to deal with such a fluid concept.

What the AI has is a bunch of rules and hints/commands from the scenario designer and each formation applies those rules without ever trying to consider the bigger picture. The bigger picture is supplied mostly by the change of ownership of geographical objectives. For example, the Russian 6th Army and AG Bobkin will continue their offensive towards Krasnogrd and Kharkov until Izyum falls, whereupon they get a new plan.

The big advantage for the AI with the new AO system is that the area that it operates in and the forces that it has to operate with are precisely known. This makes it much easier for the scenario designer to give precise instructions.

In the example above, the question of whether the AI worries about its left flank is primarily answered by the scenario designer. If he wants the flank guarded he can specify some defensive hexes that the AI will try to protect. However, even if he doesn't do that, the AI does have underlying routines that will cause it to seek out enemy units and attack if possible so AI units could end up on the left flank anyway. It's hard to be more precise because the scenario designer is mostly giving high level commands/suggestions and the underlying routines that are finally in charge of moving and attacking are necessarily autonomous.

So to summarise, the AI never sees the whole game like a human player does. The only possible advantage that the AI could have is its ability to calculate all possible combats and attack at the most advantageous odds. This is removed by the fact that we give the human player access to the same ability through the Combat Advisor and that the only basis the AI has for choosing between competing high odds combats is the higher level hints from the scenario designer, which never be as acute or relevant as those generated by a human player able to look at all levels of the game in a single glance.

Gregor

[Fred98]

ORIGINAL: jmlima

How is it possible that the AI does not factor that onto the calculations, if it is working within the same limitations?

the AI should operate within the same constraints and knowldge as a human player,

In this case the AI has less knowledge than the player. What the AI cannot do is is practice by playing from one side or the other and learn how to exploit the limitations of the Areas of Operation. Example: You know the opponent has exactly 10 armoured units. If you can destroy them all you know he has none left. An AI would never know the opponent has none left.
-

[Fred98]

ORIGINAL: Gregor_SSG

It's hard to be more precise because the scenario designer .......

You could be more precise as currently there is only 1 scenario. Someone pointed out that this style of wording as used in the manual makes the manual a tad weak.
-

[Gregor_SSG]

ORIGINAL: HansBolter

Having been a war gamer for over 35 years and participated in wargame design and development I never make those kinds of assumptions. I ask.

While I didn't offer my reason for asking I do have a valid one. I started my second game last weekend. In my first game I attempted to build a new line in front of Kharkov immediately behind the original line, falling back as little as possible. The line was too brittle and teh Russians stormed right through it much the same as in the AAR recently posted on the SSG site and linked here.

In my second game I fell back to the river line that runs east/west near the boundry of the 29th Corps and then turns south to link up to Kharkov. The line there is solid and has completely stymied the AI. However, at the far eastern end of the line where it attaches to and hinges on the original fortified line the AI has completely denuded it's front line for a two hex distance while it appeared to have pushed everything forward. Since units behind teh lines are hidden to me I have no idea if there are reserves covering this frontage, but it has all the appearances of the AI choosing to leave unthreatened hexes empty as the units on my side of the line are prevented by their corps boundry from attacking further southward into the unocuppied front line hexes. While it may have a simple explanation, it has the look of a gamey tactic.

So, before you take some one to task and lecture them on how all encompassing the design and playtesting of a game by a professional game company can be assumed to be you might one to consider that they may have valid reasons for asking the questions they do, but have simply not yet made you aware of them.

If I understand the issue here, the question is what is the AI Russian player doing in the Chuguyev salient in the historical scenario? The answer is that the Russian player is doing pretty much what they did historically. Initially, the 38th Army is tasked with supporting the southern flank of the drive on Kharkov. It includes Chuguyev as a possible objective, but is mostly focused on Kharkov and the hexes south and east of Kharkov. It will attack German strongpoints and units in a direct line between its start line and Kharkov but will ignore the strongpoints on the eastern face of the salient.

This can all change if the Russian has Variant 3. In this variant, the 38th Army makes a serious assault on Chuguyev and the entire salient. The variant is intended to keep the German player honest and prevent him from stripping too many troops out of his salient defences.

Gregor

[Redmarkus5]

I think Hans is specifically asking whether the AI is pushing assets forwards towards Kharkov that, militarily, should have been left in place to secure the 'hinge' on the right flank of their advance; that point where the Soviet salient pointing west towards Kharkov joins the original line of fortified positions.  He feels that if this is the case, then the AO concept is preventing him from exploiting a gap in the Soviet defence that a historical Axis commander would probably have taken advantage of.
 
It is a precise question, but precisely the kind of question that led to the Axis victory in this operation historically

[Gregor_SSG]

ORIGINAL: redmarkus4

I think Hans is specifically asking whether the AI is pushing assets forwards towards Kharkov that, militarily, should have been left in place to secure the 'hinge' on the right flank of their advance; that point where the Soviet salient pointing west towards Kharkov joins the original line of fortified positions. He feels that if this is the case, then the AO concept is preventing him from exploiting a gap in the Soviet defence that a historical Axis commander would probably have taken advantage of.

It is a precise question, but precisely the kind of question that led to the Axis victory in this operation historically

Its a precise question that doesn't have a precise answer. The AI will use the bulk of it forces to push towards Kharkov. There is nothing to stop them using all of its units to do so but the AI is actually made up of two parts. One is the high level instructions given by the scenario designer. The other is low level AI that actually moves each individual unit.

Questions about the high level AI are easier to answer, since they correspond to plans that a human player might have.

The behaviour of the low level AI depends in large part on what the enemy units are doing and so is a lot less predictable.

However, I do believe I can answer the question philosophically. I have no problem if the AI uses all its 38th Army units to push on Kharkov, leaving the north/south face of the Chuguyev salient defended only by the strongpoint line as its likely that a human player would do exactly the same. The AOs only restrict units to operating within a certain area, they make no attempt to regulate behaviour within those areas.

The question then becomes, what about a German counter stroke that a historical German commander could have exploited. Well, the historical German commander did no such thing. He had a huge mass of Russians (not just 38th Army) bearing down on Kharkov and he had to deal with that first. Nobody would have responded to this with an offensive east from Chuguyev unless they had been in receipt of some very detailed information from the future which gave them exact info on the Soviet dispositions and they were happy to risk the loss of Kharkov anyway.

Instead, our historical German took his best units, the two Panzer divisions and dealt with the immediate threat first. When the situation permitted, they were then used to form the northern flank of the pocket, a choice which makes much more sense anyway. Remember, the map edge is a necessary evil in our games. A drive east could have run into Russian reserve formations without creating any serious encirclement, and also would not have been directly supporting the 1st Panzer Army's drive from the south.

There's absolutely no question that AOs restrict your movement, but they do so for proper military reasons in order to give you a much more realistic experience of the battle. The Mystery Variants add more choices and more replayability to the historical situation but you can always just turn them off if you want to.

Gregor

[Toby42]

I guess that a case for AO's would be the closing of the Falaise Pocket!!

[Fred98]

The use of AOs makes this the best AI, I have ever seen.
 
So far, the AI is unbeatable!
 
 
 
 
 
 

[jhdeerslayer]

ORIGINAL: Joe 98

The use of AOs makes this the best AI, I have ever seen.

So far, the AI is unbeatable!

Agreed!

[Alan Sharif]

I would agree too.

[HansBolter]

I didn't intend for this to create a "great debate". While I am enjoying the game I merely wanted to provide some feedback that I thought the guys at SSG could use. Here is a screenshot depicting the situation where the Russians are ignoring front line hexes. The fact is that once the Russians break out there is no real definable front line and there are many locations in this particular game where both sides are not manning front line hexes(in the south outside teh area of this screenshot). The particular situation in the north, where I managed to prevent a breakout simply has the "look" of the AI exploiting the AO boundery as I used the extreme limits of the 29th Corps boundery to set the front line so the 29th Coprs units could be used to man that line (does that mean I'm exploiting the AO limitations?). All I was asking is if this represents the AI "gaming" the boundry.



p.s. I don't agree with those claiming the AI is unstoppable.

[e_barkmann]

Hans,

I would observe that:

AI is unstoppable

is not the same as

So far, the AI is unbeatable!

cheers Chris.

[Gregor_SSG]

In answer to the question as to whether the AI is 'gaming the system' I would say that both sides are 'guilty' of that, but 'guilty' is not really the right word. The German player has stripped his front line in order to bolster his northern flank defences and the AI player has stripped his front line to throw everything into the push on Kharkov.

This always happens in wargames and I'm not concerned by this instance. The AO system greatly restricts the scope for this sort of behaviour and provides us with tools to change things if necessary.

Remember though that the AI isn't 'seeing' the map in the same way as the human player. The AI is just doing what it's been told to do by the scenario designer. The AO system helps the AI by breaking down an impossible problem 'think like a human' into a series of much more discrete situations 'operate in this area, with these forces, until some condition changes'.

Gregor

[iberian]

In my opinion, that screen shot does not show the AI exploiting AO boundaries. In fact, I find it to be a good example of how Areas of Operations restrict the human player to do something gamey.

If this is the Historical variant, the 21st Soviet Army is given an AO with an axis directed to Liptsy. The computer player has two different AI levels: the Operational and the Tactical one. The operational one is directed/scripted by the scenario designer. The tactical is managed by the AI in the field, and depends on enemy movements/units and local intelligence.

For the 21st Soviet Army, the primary operational objective in this scenario is the capture of Liptsy with a full assault. Only if it loses a percentage of its strength (I recall somewhere around 20%), it will revert to a defensive operational stance. So what you are seeing is the AI using all the 21st Army units pushing toward Liptsy.

In doing so, the tactical AI looks for the path of less resistance, being restricted and funneled by its AO. Now, the important thing to understand is that the AI is stripping its frontline not because it "knows" that the German AO does not allow to break the front to the north, but because the scenario designer has given the 21st Army a full assault operation to Liptsy, and the AI does what is told to do. Now, given that screenshot, what I would find really gamey by a human player is trying to encircle the AI with the 29th Corp, but fortunately the AO given to the German HQ restricts such action.

So, the behaviour of the AI is mainly directed by the scenario designer. With the editor, one could provide a "more historical" behaviour for the 21st Army, through several means:

- Change the AO of the 21st Army, giving them a shallower and broader area so they aren't funneled to Liptsy as much as in the picture.

- Keep the actual AO, but break the 21st Army in several sub-units to give them different Operations, so that its northern division keep manning the line, while the rest pushes against Liptsy.

- Create a number of small local objectives and a broader AO just in front of the 21st Army so that it tries to push the line in order, instead of a major spearhead/breakthrough like its doing now.

And probably many more. In the end, imagination and skills with the editor are the limits.