Welcome to Close Combat Series
  Login or Register Home  ·  Downloads  ·  Forums  ·  Combat Camera  ·  Help  

  Survey
TRIVIA: Are soldiers credited for killing surrendered troops?

Yes
No
Unsure



Results
Polls

Votes 2943
Comments: 2

  Shout Box!!

Only registered users can shout. Please login or create an account.

  Main Menu
Articles & News  
    Help
    Player`s News
    Site News
    Multiplayer
    Terrain Challenge
    Boot Camp
Community  
    Forums
    Downloads
    Combat Camera
    MOOXE @ Youtube
    Statistics
Members  
    Private Messages
    Your Account
    Logout

  Donations
ZAPPI4 - $20.00
10/10/2020

ZAPPI4 - $20.00
10/10/2020

__Creeper__ - $10.00
07/08/2020

Bazookajoe - $25.00
06/02/2020

Pete - $25.00
04/11/2020

Anonymous - $10.00
03/08/2020

m2carbine - $20.00
02/14/2020

Dima - $25.00
01/26/2020

mikwarleo - $20.00
01/25/2019

0202243 - $10.00
12/04/2018

Find our site useful? Make a small donation to show your support.



Search for at
Close Combat Series Advanced Search


 Author
Message
 
Jagre3-3

Rep: 2.1


PostPosted: Tue Feb 12, 2019 8:16 pm Post subject: I Figured Out The AI. Here's What I Discovered Reply with quote

TIK's Small Map Mod presented an interesting mystery to me. Why does the AI play better on smaller maps? The obvious observation is that it sucks, hard, on the large maps of the new releases. So, I'm not even going to bother with an analysis of it on the big maps. There's a video of that on YouTube that does it just fine. However, TIK's mod inspired me to not only make my own, but to find a way to optimize it for better gameplay versus the AI. So, here you go. Each spoiler contains each test I did in a series of four tests after making a small version of Bemmel in Last Stand Arnhem, ala TIK's method he put up on YouTube, with some differences of my own. See the fifth spoiler for my final opinion on the whole "Small Map Debate".

TEST ONE

Hidden: 
////////////
//TEST ONE//
////////////

Small Map Bemmel
9 Victory Locations (5 Map Connections, 4 Regular)

Equal Forces, Same Setup

Line Difficulty, Both Sides

"Always See The Enemy" turned on for me in order to watch the AI

No Data Mods

Meeting Engagement

//RESULTS

AI was static for a little bit. Then it ordered it troops south to take a victory location. The fight was on quickly (credit to the small map) Kept two squads in reserve with its mortar, and two squad with a leader on the "to Elst" VL.

When it ran into my defesnive lines in the buildings, it stopped moving to throw down suppression and return fire. Used mortars on the building closest to the VL, also was closest to its troops.

By sheer firepower alone I was forced to abandon my position at the buildings and withdraw to the town of Bemmel to the east. It pounded my retreating units with mortars and pushed forward to secure my old position near the "to the Orchard" VL. As I occupied the town, it pushed up its reserve and shifted mortar fire to the town, mainly where it was receiving fire from.

It stopped pushing forward its reserve and stayed put on its orginal starting position returning fire with my troops in the town. Both sides sustained about 50% casualties.

//CONCLUSION

The stock AI, with no data changes to BGroups or anything else, is definitely more active. The fight's on in a hurry, and the AI participates until it feels it has the ability to. The Ai had aggressively pushed until it took too many casualties to be combat effective. A better fight, no doubt.


TEST TWO

Hidden: 
////////////
//TEST TWO//
////////////

Small Map Bemmel
9 Victory Locations

Equal Forces, Same Setup

Line Difficulty, Both Sides

"Always See The Enemy" turned on for me in order to watch the AI

BGroups modded for battlegroup AI, tactics, commander traits, etc.

Meeting Engagement

//RESULTS

It's the same setup as before, but with the modded BGroups added in.

It started out the same, but the AI was more aggressive, with fewer reserves. It had more units to bear on my positions near the Orchard VL. It did an early assault on one of my squads that ended very badly for the assaulters. The battle progressed that way for some time, the AI feeding more units into the Orchard VL to take it. It took more casualties in this battle due to the aggressiveness, and a setup in the beginning that left more units in the open, I can't confirm if the game did that to the AI or not, so I can't count that against it, but I can say that the amount of units in the open with the previous test was less.

//CONCLUSION

The AI was really focused on the "to the Orchard VL", and it seemed like it wasn't going to go anywhere until it took it. It was the same way in the previous test, just that fewer units were thrown at it. My next test will be with fewer Victory Locations to see what happens.


TEST THREE

Hidden: 
////////////
//TEST THREE//
////////////

Small Map Bemmel
7 Victory Locations (5 Map Connections, 2 Regular)

Equal Forces, Same Setup

Line Difficulty, Both Sides

"Always See The Enemy" turned on for me in order to watch the AI

No Data Mods

Meeting Engagement

//OBSERVATIONS

An interesting observation right off the bat. The deployment area available to either side usually centers around the VLs that are worth the most points, or places the starting positions at map connection VLs. Let's see how this goes with fewer Victory Locations. My hypothesis is that the AI focuses on Victory Locations that are the closest to it, so fewer Victory Locations will bring the battle into hotspots of Victory Locations. Fewer Victory Locations has also been reported to cause better AI performance, so let's see how this goes.

//RESULTS

In this battle, my starting deployment area was in the same corner as the other two tests. There was more activity than before, though more cautios without the BGroups changes. The battle was more interesting, the AI attacking in toward the town to take Bemmel and The Church. There was movement southward toward the map connection VL, but still pushes toward the town at the same time. It kept reserves and asked for a ceasefire when it lost too many troops. Once I got into the town and set up a perimeter, it was pretty much over for the AI, because it had to attack across open ground and the axis of attack left the Germans silhouetted as it came up the rises to the bridge.

//CONCLUSION

The AI is definitely victory location driven. It has the ability to react to the player, but primarily chases victory locations. With mortars and MGs it targets units that threaten to suppress its forces. Its mortar fire is well placed. Overall, it's a better experience. Now for the BGroups changes test.


TEST FOUR

Hidden: 
/////////////
//TEST FOUR//
/////////////

Small Map Bemmel
7 Victory Locations (5 Map Connections, 2 Regular)

Equal Forces, Same Setup

Line Difficulty, Both Sides

"Always See The Enemy" turned on for me in order to watch the AI

BGroups modded for battlegroup AI, tactics, commander traits, etc.

Meeting Engagement

//RESULTS

In this battle,the tactical situation is different. Though the battle was set up as a meeting engagement, my deployment zone was in the town, giving me Bemmel and The Church. This makes the battle a little bit of a defense. However, the AI's starting zone is directly above, almost neck and nack with my own zone. It's very much a close fight. And something was done right, because there was no way in hell I was going to get fire superiority. Long story short, It was an intense fight that started almost immediately.

//CONCLUSION

The AI does much better with the BGroups changes. The fewer amount of Vcitory Locations makes contact with the AI inevitable, and the AI seems more coherent. During the last test it just kind of held its ground when it figured out it wasn't going to take The Church by assaulting across the street. It tried that a couple times. Then it just kinda let me come at it. Eventually, I was going to lose that fight on casualties alone. No doubt.


FINAL OPINION

Hidden: 
////////////////////
//MY FINAL OPINION//
////////////////////

In my mind, the debate has been settled. Small maps are the way to go. While satisfying player demand for larger maps, the developers never re-coded, or even adjusted, the AI to suit the larger maps. Then, they loaded the maps with too many victory locations. So, you have a victory location driven AI with the ability to react to human players, where the AI's core design is based on MUCH smaller maps. All in all, the AI, while not genius, is capable enough to handle the player, even beat them, but you have to give the AI its element back. In essence, the game has be adjusted to suit solitaire play against the AI. By all means, play the large maps on multiplayer if you like, but the only way to have fun against the AI is to take it back to its roots.

I do agree that the AI leaves much to be desired, but in my tests I saw intelligent behavior under the hood that I hadn't noticed before. The AI does have the ability to be a decent opponent (Keyword "decent". It's never going to match the intelligence of a person on multiplayer.) I could see how Stwa, yes Stwa, could have favorable experiences in Modern Tactics. Its scenario editor is robust, and because there aren't any map connections for can place a single victory location if that's what you wanted to do, then lay out the deployment zones accordingly. So, yes, the AI is good, not great, but it delivers an experience that's fun to slug it out against. And, frankly, that's all I really ask for from this. At least I can have fun in Close Combat again, and that was the goal to begin with in these tests!


Now, from my end the issue is decided. I've already started work on making a small map mod for Last Stand Arnhem, then I'm going to move on to the other titles, eventually doing Gateway to Caen (Sorry, TIK. I'm gonna have to compete with you, now. There's a silver lining, though! I think that mod is one of the single greatest contributions to Gateway to Caen since its release).

Be on the look out for the mod, guys! These experiments have given me something to work with, and I haven't even started on other data changes, like morale in AlsTeams and AxsTeams (which, by the way TIK, is why Gateway to Caen's infantry are next to useless when combined with the weapons data), more support in the campaign. Basically, some of the same methods used in vetmods. Either way, it's going to be a cool experience.



SmallBemmel.png
 Description:
Small Bemmel
 Filesize:  499.42 KB
 Viewed:  3025 Time(s)

SmallBemmel.png


Back to top
View user's profile Send private message
 
Zep

Rep: 41.5


PostPosted: Wed Feb 13, 2019 8:54 pm Post subject: Re: I Figured Out The AI. Here's What I Discovered Reply with quote

Thank you, That would be Awesome Very Happy
Back to top
View user's profile Send private message
 
Pete

Rep: 102.2
votes: 11


PostPosted: Thu Feb 14, 2019 8:06 pm Post subject: Re: I Figured Out The AI. Here's What I Discovered Reply with quote

Thanks for your time investigating this aspect of the game. But now I want to know which map sizes you would recommend. Did you do testing with the sizes and proportions?


Dulce Bellum Inexpertis
(War is delightful to those who have no experience of it.)
Our facebook page: https://www.facebook.com/buckandpete/
Back to top
View user's profile Send private message
 
Sapa

Rep: 64.1
votes: 8


PostPosted: Sat Feb 16, 2019 4:36 am Post subject: Re: I Figured Out The AI. Here's What I Discovered Reply with quote

Great! Smile  All thats now needs improvement is to change the girlie soldiers to real men Very Happy
Back to top
View user's profile Send private message
 
dj

Rep: 118
votes: 7


PostPosted: Sun Feb 17, 2019 7:17 pm Post subject: Re: I Figured Out The AI. Here's What I Discovered Reply with quote

Great job.  Confirms what many of us have suspected over the years.  Ironic that CC2 or even CC3 has vastly superior AI performance vs "rewrites" which shifted to massive maps rendering the AI completely useless.  No point in even bothering to play as single player for the "rewrites".  Yet developers kept heads in sand and continued to insist they keep making the huge maps. And then claiming they "fixed" the AI.  Not so much...
Back to top
View user's profile Send private message
 
pvt_Grunt

Rep: 83.8
votes: 5


PostPosted: Tue Feb 19, 2019 6:41 pm Post subject: Re: I Figured Out The AI. Here's What I Discovered Reply with quote

Great write up!
Back to top
View user's profile Send private message
 
Pzt_Crackwise

Rep: 63
votes: 1


PostPosted: Wed Feb 20, 2019 7:43 pm Post subject: Re: I Figured Out The AI. Here's What I Discovered Reply with quote

I wish AI could also be tweaked to deploy their AT Guns and AT Teams properly Sad
Back to top
View user's profile Send private message
 
Uberdave

Rep: 26.9


PostPosted: Mon Apr 29, 2019 7:56 am Post subject: Re: I Figured Out The AI. Here's What I Discovered Reply with quote

Smaller maps = more aggressive AI, as VLs are closer to reach.

Did I just save you a couple of paragraphs?

Laughing
Back to top
View user's profile Send private message
 
Jagre3-3

Rep: 2.1


PostPosted: Sun Jul 28, 2019 8:32 pm Post subject: Re: I Figured Out The AI. Here's What I Discovered Reply with quote

Sorry, guys! Been very busy. I haven't had much time to devote to this, but I'm starting work on it again.

Pete wrote (View Post):

Thanks for your time investigating this aspect of the game. But now I want to know which map sizes you would recommend. Did you do testing with the sizes and proportions?


I based my sizing on CC2's map sizes. If the original AI was centered on that image size, and my tests proving it hadn't changed much since then (aside from pathfinding and such), I didn't want to venture outside the AI's core coding and designed play space. CC2's maximum map size was 2280x2280 pixels, so I set that as my map limits, or at least as close as I could to it. Some are planned to be smaller than that, depending on the contents of the downsized map, and where I place the combat area for it (usually where the roads junction to keep the map exit VL's logical. Otherwise it's the part of the map with the bext chance of fun gameplay).

dj wrote (View Post):

Great job. Confirms what many of us have suspected over the years. Ironic that CC2 or even CC3 has vastly superior AI performance vs "rewrites" which shifted to massive maps rendering the AI completely useless. No point in even bothering to play as single player for the "rewrites". Yet developers kept heads in sand and continued to insist they keep making the huge maps. And then claiming they "fixed" the AI. Not so much...


This is a symptom with a lot of games these days. The focus has gone to multiplayer, leaving the single player elements of games malnourished, like it was a chore the developers had to do to sell the game. Close Combat is no different. The AI was never redone (which they had the capacity to do, because they had access to the source codes, which you'd need to bring it to modern systems, anyway. In short, the developers got lazy).

Sapa wrote (View Post):

Great!   All thats now needs improvement is to change the girlie soldiers to real men


I've been thinking about that. I always modify the BGroups in every single CC I play to have better morale, tactics, strategic stats for the battlegroups. Now, I think about it, I haven't really noticed too much of the girlie soldiers phenomenon. I wonder if that's because the "morale" column in the BGroups file has an effect on that. I'll look into it, because I haven't seen the difflevel setting in the registry entries for the re-releases. I wonder if they changed how girlie soldiers is implemented in the game.

It's important to mention, too, that this is a project to develop my attention span and dedication to a project. It's not only to make Close Combat better for single player, but also to actually finish what I start. Being a self-taught independent game developer, I'm training myself to stay on mission and get sh*t done. I've got a lot of cool projects planned for the future, and I'm getting to the point of having the skill level required to undertake them. It's gonna be a lot of work, but this map project will instill the patience I need to see it through. Time to end the hiatus and get you guys sets of maps that will give you the experience the veterans of the franchise deserve. I'm a lifelong fan myself, so trust me I understand the frustrations with the AI.
Back to top
View user's profile Send private message
 
Jatke

Rep: 10.1
votes: 2


PostPosted: Mon Jul 29, 2019 5:27 pm Post subject: Re: I Figured Out The AI. Here's What I Discovered Reply with quote

I dont believe BG chacteristics have had any effect since CC5. Those traits seem to have been randomized per battle. You can delete those values entirely from BGroups and the game wont even hiccup.

Adjust the max CC2 map size for the revised scale when you resize the maps in PITF/GTC. 2280x2280 will become 3648x3648.

As always, the solution to Sapa's girlie soldiers complaint is to check 'Always Obey Orders'.
Back to top
View user's profile Send private message
 
Hicks

Rep: 31.6
votes: 2


PostPosted: Fri Aug 02, 2019 6:07 am Post subject: Re: I Figured Out The AI. Here's What I Discovered Reply with quote

Jatke wrote (View Post):
As always, the solution to Sapa's girlie soldiers complaint is to check 'Always Obey Orders'.


That, or play the pre Cross of Iron games. Wink
Back to top
View user's profile Send private message
 
 
Post new topicReply to topic printer-friendly view Close Combat Series Forum Index -> Modding Workshop


 
   
 


Jump to:  
You cannot post new topics in this forum
You cannot reply to topics in this forum
You cannot edit your posts in this forum
You cannot delete your posts in this forum
You cannot vote in polls in this forum
You cannot attach files in this forum
You can download files in this forum




Forums ©





In August of 2004, Zappi, Homba, Bambam887, RedScorpion and MOOXE all pitched
in to create this Close Combat site. I would to thank all the people who have visited and
found this site to thier liking. I hope you had time to check out some of the great Close Combat
mods and our forums. I'd also like to thank all the members of our volunteer staff that have
helped over the years, and all our users that contributed to this site!