Welcome to the LnLP Forums and Resource Area

We have updated our forums to the latest version. If you had an account you should be able to log in and use it as before. If not please create an account and we look forward to having you as a member.

Comments, observations on the AI for coming East Front modules..

kipanderson

Member
Joined
Oct 23, 2014
Messages
114
Points
18
Age
66
Location
Surrey. UK.
Hi,
Most here know Command Ops better than I do although I am a very long stand fan. However with that qualification out of way I will just note a possible problem I have found when playing those stunning Kursk users produced scenarios.
I have found that the defending AI may struggle a bit with “holding a line...”. Of huge importance when playing most, near all Eastern Front scenarios... as it was in the real world. It tends to cluster around given points to an extent that leaves itself very open to being outflanked by not very distant breakthroughs further down the front line.
I took the largest of the wonderful Kursk maps produced by CapHillRat and duplicated some of the Soviet infantry and artillery forces. Then formed a classic frontline of SS battalions at round 60% strength, each covering about 2km of front. All dung in but not fortified. Mobile units behind the line.
Then classic Soviet attack.
I discovered that even when I had placed an AI “defend..” maker at 1000 points under each SS frontline company position there was still some tendency to ignore flanks and cluster, even counter attack into potential pockets resulting from Soviet breakthroughs.
I of course do not know how to training up the AI as well as many here but just thought it worth mentioning that trying to hold a line, bend with the wind but somehow keep a frontline will be very important when we get to the Eastern Front. Not always, but normally.
The fact that I found this possible problem did not all together surprise me because I had noticed that all the previous setting for Com Ops had been quite friendly or forgiving of units clustering around a given point, or driving up certain roads.
But say in Normandy, and certainly on the Eastern Front some attempt to maintain a line, even very weakly held, was important. Of course this often failed for short periods and then there was chaos ;)... but the line then did reform to restore matters.
With more expert handling of the AI defend objectives this may not be a problem... but just thinking ahead to all those great Eastern front games we have coming.. ;). Nor do I wish to exaggerate the problem.
Great looking forward to version 2.0,
All the best,
Kip.
 

Iconoclast

Member
Joined
Oct 26, 2014
Messages
130
Points
18
Location
germany
Good points.

related: I have AI rarely seen forming and comitting reserves on Bn level (the level I usually tend to issue orders). Is that an observation others would share? Most of the times, AT-Plts are moved into reserve, but I have never seen a line unit acting as reserve.

Would be interesting for the east fron games, and others too, how the ai handles small troubles with own reserves.

A
 

john connor

Member
Joined
Oct 22, 2014
Messages
2,488
Points
63
Age
60
Location
Brussels
Spot on, Kipanderson. I was going to write to Dave about it in the light of putting together 3 Kursk scenarios (not the one you played, because I took my Axis AI scenarios down) and then struggling for a long time to balance them for play against an Axis AI. In the end I gave up and have now only put up two 'play as Axis' versions that I was happy enough with, because I couldn't get - EXACTLY AS YOU HAVE SAID - the AI to just bloody stay put and defend. There needs to be an option that allows that, I think (for the scenario designer, I mean - an option to defend in-situ, as a preference, that you can apply to a unit). Or a task to defend a line. But in the maps I put together I could carefully and historically place the Axis AI forces where they would have best effect, in dug-in/fortified positions, and it was all to no avail. No matter what I did (including stringing together many, many objectives to form a line) the first thing the AI does is get up and move out of fortified positions and re-org according ot its own logic. It was maddening.

In a way that it defends on 'points' instead of lines wasn't, I think, the worst of it. The worst of it was just not staying put on the 'points' I wanted it to hold (as scenario designer). I think this is why we have seen - in other scenarios - 'static' defender units (south end of Arnhem rail bridge in the Arnhem scenarios, for example) - to absolutely make sure that they remain in place!! But that's a poor solution, because eventually, it would be good if they moved.

Like I said, I've now only put up Axis v AI versions of my Kursk maps for this reason. Glad you saw it too, Kip.

I also had problems with the Axis putting together very large scale (soviet tank) attacks, in fact, but could get better results with this than with it defending. The way the AI puts together a tank attack of 3 brigades strength, say, looks VERY different to how it looks when a human gives an attack order to a force that size and the AI then executes it.

By the way, Kip, Caphillrat did one Kursk map and scenario (the one you just played, I assume) and the Prokhorovka scenarios I did (2 of which are available in the resources section here) were on my own maps - traces from soviet-era maps - not copied from his.
 
Last edited:

kipanderson

Member
Joined
Oct 23, 2014
Messages
114
Points
18
Age
66
Location
Surrey. UK.
John, hi,
Apologies for crediting the scenarios wrongly.. have of course seen your name around a lot and knew you were “one of the inner circle..” ;). In the best of ways BTW..
Here is a shocking fact... I was a beta tester way back.. :).. May well be eight of more years ago.. time flies.. but did not have time to give it the attention it deserved.
But have learnt that it does help and is welcome to the developers if you politely bring to their attention something you think is a possible problem.
Version 2.0 is going to be huge leap forward an already great game..
All the best,
Kip.
 

john connor

Member
Joined
Oct 22, 2014
Messages
2,488
Points
63
Age
60
Location
Brussels
Beta tester? That puts you in the central, privileged, private, core inner circle, Kip, even if it was a while ago - once you're in there's no getting out........ ;) I'm afraid I've never got near such delights.

But if you were saying you liked my maps (as well as Caphillrat's), then thanks! I really enjoyed making them.

I hope this year is going to be a great year for Dave and Command Ops!
 

Dave 'Arjuna' O'Connor

Panther Games Designer
Joined
Jul 31, 2014
Messages
3,415
Points
113
Location
Canberra, Australia
Website
www.panthergames.com
I acknowledge that the defend AI could benefit from better modelling and in particular the need to model fronts and flanks better. It all boils down to priorities doesn't it. No AI is ever perfect. There's always room for improvement. I know that the next two highest priority items are, namely sequential tasking and mounted ops. These are both big undertakings. Modelling fronts and flanks with area of operation, boundaries, refusing flanks etc is also going to be a big undertaking and hence why I have not touched it so far.

Perhaps as a stop gap would be the option to pre-assign orders in the ScenMaker that are carried into the game. Eg In the SM you can issue defend in situ orders to the forces you want and set a time that these will remain in force for. When the game loads it will hive off all forces with pre-assigned orders and only use the remainder to allocate to new tasks. It's not a perfect solution at all but it may go a way to achieving the same effect you desire.
 

john connor

Member
Joined
Oct 22, 2014
Messages
2,488
Points
63
Age
60
Location
Brussels
That would be a great interim solution, Dave. Certainly for shorter scenarios it would be almost an entire solution. It would be good, in fact, to be able to 'script' an attack by the AI opponent also, when putting together scenarios in scenmaker. Mainly for those scenarios where you want to really start them off as they started historically.
Thanks!

It's still not only the best game AI out there, but virtually the only game to take AI seriously.
 

Dave 'Arjuna' O'Connor

Panther Games Designer
Joined
Jul 31, 2014
Messages
3,415
Points
113
Location
Canberra, Australia
Website
www.panthergames.com
The trouble is with this option is that to implement it we would have to add lot of functionality to the ScenMaker for giving orders that we don't currently have. It would also entail changes to the data structure of a scenario bringing it more towards that used for a saved game. One further thing that comes to mind is that if we did align the data structure then maybe we could create a scenario from a save. Interesting. In any event it will require a fair amount of work.
 

john connor

Member
Joined
Oct 22, 2014
Messages
2,488
Points
63
Age
60
Location
Brussels
Creating from a save would be very interesting, true.

In the Prokhorovka scenario - just as an example - the Soviet side has a very significant superiority in armour, men and force (according to the scenario stats). If I put more or less realistic stats in for all units, both sides (which would mean giving the Axis a significant edge in training and experience, in aggression, stubbornness and leadership qualities, generally) then play as Axis, with objectives set up to tempt the Soviet AI to attack, then the AI will generally - I've found - put together attacks, more or less as the soviets would do historically. I mean with that kind of level of success. An achievement already!

But if I then play the Soviets against the Axis as AI then I can completely and utterly destroy the Axis (I mean almost destroy every single unit on the map!) with the most basic of orders. In that scenario it can be done with 5 simple attack orders which instruct the entire Soviet force to get to the main objectives. The AI then puts together 5 huge beautiful attacks and bob's your uncle. Basic play, for the human.

When the AI constructs these attacks it seems like it uses much less strength and the attacks seem to grind to a halt quicker in the face of opposition. And a HUGE part of this, I think, is that when I defend as Axis I do not do what the AI does. I stay put and place careful bombardments to stall obvious attacks. But the AI moves all the forces around and takes up - I think (guessing) - an 'all round' posture on certain objectives which leads to less firepower in a line up front and, in addition, loses all the carefully set modifiers from being dug in or fortified etc.

But I think the 'issue' isn't so very important, since I've only encountered this kind of difficulty in either very large scenarios (Maas-Rijn, for example, and Spearhead v Reich, where the AI can end up - after initial efforts - not going for any objective, or smaller set of objectives, in a concentrated way, (because, I think, it has too much to worry about....!!) so it just sort of sits back instead in a defensive posture and puts in occasional smaller attacks without gathering any decisive concentration (which I think would in any event be more like what would happen in real life). Ditto in scenarios where one side is very much stronger than the other, which is rare.

I do think - as I've said before - that the AI handles Regimental size scenarios better. Just my little opinion. I look forward to seeing how it fares in KOAD, and seeing how the designers have put together those scenarios (which I've heard are very large) to make it work well. I'll learn from that, no doubt!

So if it's time consuming, I wouldn't worry, is what I mean, because you don't encounter it so often in typical games. And I could make that Kursk scenario work better for the Axis as AI, I assume, if I could work out how to mess with the estabs and create some 'static' units to solve this 'issue'.

On the other hand, I guess Kipanderson started this thread because he feared that in the East Front future there might be more scenarios like this - large with a large disparity in strengths between the sides.
 
Last edited:
Top