Large Number of Objectives= Trouble with AI

[benpark]

CC2 had small maps. Coupled with fewer objectives, which were closer together resulted n a more manageable task for the AI player. With the massive newer maps (which have benefits i other ways), the AI struggles to present a reliable plan for defense and is consistently terrible in the attack vs. even forces.

I would really encourage the developers to look at having no more than 3 VL on any given map- not spread all over the map, but close enough together that the AI doesn't spread itself all over the map in the attack.

I realize that the large number of VL appeals to MP players. Make 2 scenario versions, if that is a main concern- one version to be played vs. AI and one MP. I'll never play a game vs. a human in CC again (no time at all for that anymore), so some more attention to the AI problems is a must for me.

[Conrad mana]

Hi benpark,


Your mentioned suggestions apply to our approach when we first started developing the maps. For example VL's had to be placed carefully depending on the amount of points, the largest VL has a value of 200 which attracts and directs the AI. Some maps are larger than others which depends on the battlefield type.


Although I would encourage you to play H2H since that has always been a better experience for me, even with Close Combat 2.



Perhaps we can play a game after the release?

[octyss]

I don't agree that a smaller map means a better AI, maybe only to a certain way.
If you want to decrease the size of the maps, be sure to improve the way the units remain hidden, because they remain hidden after firing according to the distance from which can be seen, especially AT units.

I finished two large Pitf campaigns against AI, I can say AI is more advanced compared to previous games, of course is much room for improvement.
I saw the AI knowing to attack aggressively when it is usually surrounded in the directions that brings him out of the encirclement and can restore the supply line or wants to put you in encirclement.
In a campaign, the key to the game is the enemy encirclement, a location has no value unless it blocks or supervises the way to an exit.
How do you want to make a map with 3 VL when it has 4 or more exits without at least one center VL and at least one VL per exit ?

[Conrad mana]

ORIGINAL: octyss

I don't agree that a smaller map means a better AI, maybe only to a certain way.
If you want to decrease the size of the maps, be sure to improve the way the units remain hidden, because they remain hidden after firing according to the distance from which can be seen, especially AT units.

I finished two large Pitf campaigns against AI, I can say AI is more advanced compared to previous games, of course is much room for improvement.
I saw the AI knowing to attack aggressively when it is usually surrounded in the directions that brings him out of the encirclement and can restore the supply line or wants to put you in encirclement.
In a campaign, the key to the game is the enemy encirclement, a location has no value unless it blocks or supervises the way to an exit.
How do you want to make a map with 3 VL when it has 4 or more exits without at least one center VL and at least one VL per exit ?

It depends on the type of battlefield, as I said some maps are larger than others.

Regarding the entry/exit VL's these carry less value than a 200 point/100 point VL, the AI wil be directed and deployed according to those values.

For example the map 'Les Nouillons' (Belleval estate) is a central orientated map with 1 high value VL and 2 medium value VL's placed relatively close together at the Belleval Château. In total there are 9 VL's.
The outcome was that we've seen some very intense battles for the Belleval Château against the AI, just as happened during Operation Epsom.
So it depends in the type of battlefield and how the maps are designed, plus how the VL's are arranged.

[Javolenus]

Hmm, this is interesting. I own all the CC remakes and while I love the overall CC design I must say that I find the AI to be a major disappointment.

This week I played scenarios in several CC games (Modern Tactics, COI, LSA, PITF, Longest Day). Each scenario played out exactly the same -- the AI shuffled its units about aimlessly before launching sustained suicidal advances into kill zones. In all scenarios the AI was annihilated while I suffered, at worst, a few guys wounded. I played all scenarios on maximum difficulty.

Another interesting thing to consider -- I also found (this week) a game called Eric Young's Squad Assault Western Front (on the website Gamersgate). This game is an old 3D version of CC and plays very much like CC. It also has an easy-to-use editor. The AI in this game is no worse than the CC remakes. And yet it costs only £5.

I actually really love & admire the overall design of CC, its psychological modelling, it's mission editor. I even prefer the 2D or "topdown" graphics. Hell, it's actually a beautiful game. Except for the AI.

OK, I know that I should play online against humans. But I'm just not into it. I always play single-player. But -- and I think this is important -- every new remake of CC is marketed as a "single player" game with "improved AI" and, like the trusting/optimistic/hopeful guy I am, I'm always sucked in. Only to be disappointed by the AI.

And so I won't buy the next CC games unless I read lots of positive stuff about AI improvements post-release. There -- I said it!

[mooxe]

On one hand you have great looking large maps with tactical features you can fight over. Looks good from a marketing point of view. Large maps with many VLs make the AI very unchallenging. So the AI leaves 99% of (single player) players disappointed in the game.

On the other hand these large maps are said to be great for multiplayer. This is highly subjective. As the game has not improved on its multiplayer functions, ie; still 1v1, still hard to connect, lack of players, all the multiplayer guys are left disappointed.

So in order to satisfy the most amount of players, the AI has to be improved because most people play vs the AI. The only way we can see the AI being improved based on the history of Close Combat is to make smaller maps to make it appear like the AI is better.

[Javolenus]

Yeah, I see what you mean here. Seems reasonable to me.

I just played a custom scenario in CC COI. It was a small map on which I placed 3 objectives. I gave the enemy AI a big numbers advantage and I played on (for me) 100% "realism" on a 15-min. timer. Before the time limit the AI had fled the field. I had no casualties at all. As in all other single-player CC games I ever played, the AI could not organize an effective attack or defence. The AI did not use smoke or terrain cover/concealment. Neither did it remain in the good deployment zones I gave it (good cover+LOS). The AI in all CC games seems to have one plan: crawl towards the objectives, even if it is suicidal.

It's a big shame. I really love the CC design & features. I think it's a great concept & I really like how it tracks individual soldiers and simulates combat psychology. But the game does not deliver the "single player" and "improved AI" so often quoted in the marketing. Having now spent some £150 on the series to date, I feel I cannot justify spending more unless the new game(s) really delivers. The price is just too much for nice graphics & cool explosions.

There's another long-in-the-tooth squad-based tactical game that's worth looking at for inspiration. It's called John Tiller's Squad Battles. The graphics are pretty bad and there's just too much mouse-clicking for my taste. But here are some handy features from the Squad Battles series that could help CC single-players create challenging scenarios:

1. Customizable visibility (entered as a value in the scenario settings).
2. Customizable experience, quality, morale, equipment for soldiers/units (can be changed in the scenario editor).
3. Weapon/equipment performance degrades after excessive use in combat.
4. Units can be toggled as "fixed" in the editor -- meaning they won't move.
5. Reinforcements can be added to a scenario -- deployment-zones and chance-of-arrival can also be pre-set or even randomized.
6. Objectives can be set in the editor and values ("victory points") assigned.
7. Limits for smoke usage can be set in the editor.
8. "Asymmetric" scoring can be set in the editor.
9. Locations can be given negative values (in "victory points") if damaged -- simulating punishment for collateral damage.
10. All data is moddable. Customized OOBs can be created. New weapons can be added to the ToE.
11. Command & control is simulated by giving leaders special abilities (e.g. "Rally", calling for offmap support etc.).
12. Small sub-maps can be created from the large maps that ship with the game, enabling players to fight limitless custom scenarios at varying scales.
13. Ingame music can be customized/modded -- there are 3 options here -- music for victory, stalemate, defeat.
14. Map labels can be added in the editor.
15. Objectives can be edited -- capture, preserve, exit etc.
16. Units suffer fatigue and this can be edited (an option in the editor) so that units start the battle fatigued.
17. Offmap fire missions are not instant or 100% accurate. The min./max. times for fire support arrival can be set in the editor. Accuracy of fire support depends on quality of spotting by ingame units.
18. Obstacles, trenches, foxholes, sewers, caves, pill boxes, tunnels, bunkers, wire, mines, booby traps can be added via the editor.
19. Units have a "stance" -- "up" or "down", which effects visibilty/vulnerability and also movement rate.
20. The parameters for "major defeat", "minor defeat", "draw", "minor victory", "major victory" can be set in the editor.

[Txema]

ORIGINAL: Javolenus

And so I won't buy the next CC games unless I read lots of positive stuff about AI improvements post-release. There -- I said it!

I am thinking the same. Please, give us a decent AI in the next games.


Txema

[Renato]

I'm really longing for a better AI, but I know that I'll buy it all the same. [:)]

[TIK]

ORIGINAL: mooxe

On one hand you have great looking large maps with tactical features you can fight over. Looks good from a marketing point of view. Large maps with many VLs make the AI very unchallenging. So the AI leaves 99% of (single player) players disappointed in the game.

On the other hand these large maps are said to be great for multiplayer. This is highly subjective. As the game has not improved on its multiplayer functions, ie; still 1v1, still hard to connect, lack of players, all the multiplayer guys are left disappointed.

So in order to satisfy the most amount of players, the AI has to be improved because most people play vs the AI. The only way we can see the AI being improved based on the history of Close Combat is to make smaller maps to make it appear like the AI is better.

Agreed. Smaller maps are better overall. Better AI, better single player, and generally quicker and more intense games. It's a shame these facts were never realized. Luckily, the new AI in TBF will hopefully make these issues obsolete, but I stand by the opinion that BIGGER doesn't necessarily mean better.

As far as multiplayer is concerned, map size is still an issue. When you only have limited teams on the field, large maps are pointless. They take longer to fight over, fatigue your infantry, take multiple battles to win or lose the same map, and you have to play with an artificial time limit so as not to get fatigued yourself (how realistic is that?). Some would argue that it gives you more room for maneuver. This might be true, but you don't need a HUGE map to maneuver. I can maneuver just fine on a CC5 map

1v1 is hurting multiplayer. I would love 2v2 or 4v4, unless there's uncontrollable lag. Lag's an issue now, so I can imagine it being worse in a team game, and is probably the reason 4v4 isn't in the game now.

[Gen_Jack]

I also agree. Small maps are much better for both play against AI and multiplayer (1v1). Large maps look great to scroll around, though even that can be a pain as you bounce around, but with only a maximum of 21 squads (15 for LSA and prior) it is virtually impossible to cover the map when defending, and when attacking you have to take a huge risk that opponent will not counter attack where you are weak in order to amass enough units to attack in one area of the map. Essentially battles become multi-turn sneak around and run a grab fests.

Once true multiplayer is available (2v2 or more) large maps will be more feasible. Otherwise, allow players to have more than 21 teams total. Maybe up to max of 21 teams that the player directly manages in game (as at present) plus up to another 10 (or maybe 15) teams that the player places but remain static, other than maybe to retreat towards player's home side of the map or entry VL. Unfortunately this still would not solve the current poor AI issues on large maps.

[Kanov]

A better solution would be Total War series implementation of reserves:

In that game the player controls 20 units in any given battle, but when fighting a battle involving several allied armies, the player can choose between to allow the AI to control them or not for each and every single allied army.

Having the AI control the reserves has the advantage that they enter right away into the battle. For cases when you want to overwhelm the enemy. Of course the AI allied commander is as dumb as the enemy AI, not very good idea to handle your elites.

Having the reserves be controlled by the player allows piece meal like entering into battle. If you're already using 20 units no other reserve can enter until one of your active units is killed or fled from the battlefield. If you're using less than 20, reserves come in until the 20 slots are filled (that is if there's enough reserves to fill them).

[benpark]

What CC modding program can be used to remove VLs from the map? I'd like to see what having 2-3 VLs will do for the AI.

[Misconduct]

I've been waiting for Cross of Iron to have a better AI, the idea to buy units and place them in a campaign always appealed to me. Problem has always been a lack of AI or proper multiplayer, from what I gather here the AI still hasn't improved much.

As for some people say - do the multiplayer, well I did tournaments for Close Combat 3 for years, Cross of Irons even, while its pretty good the biggest problem is doing a campaign, short battles are not that interesting, in the long run a decent AI would be interesting, i.e if they spot a tank, try to bring tanks over or anti tank guns.

Rather the AI places its anti tank guns in the open field for no reason, or shove 5 tigers in a forest when it gets bogged down and you end up in a stale mate for 15 minutes.

[Platoon_Michael]

ORIGINAL: benpark

What CC modding program can be used to remove VLs from the map? I'd like to see what having 2-3 VLs will do for the AI.

I think Mafi's bed.9 will do it.
Easy tool to use too
http://www.closecombatseries.net/CCS/mo ... t&lid=4189

[benpark]

Thanks for the link. I'll try it out.

[Javolenus]

I agree that Cross of Iron gives the best single-player experience (but that's not saying much). I like the CoI design better because maps are smaller and, crucially, the player can place his own objectives and deployment zones in the Editor.

The later remakes look better than CoI in terms of fx & graphics but they are useless games for single-player. A typical single-player game for me (in PitF, for example) consists of sending units zooming around the big map to capture VLs while waiting for the AI to show up. This is not fun. And it's not realistic.

In CoI you have some control over scenario design. This is a good thing. If the AI could actually fight then it would be a really fantastic game. But even when I set my "realism" level on "Grognard" and the AI's on "Comic Book" I still win hands down. Every time. Without fail.

People say that I should just accept that single-player is a no-go, and that I should go for multiplayer, but the games are all marketed as being "single-player" and with "improved AI", so I don't think I can be blamed!

Anyway, CoI is the only CC game I still (occasionally) play. The rest -- LD, LSA, PitF, Modern Tactics were, sadly, wasted money.

The series needs a good AI and a (very) flexible Editor for those who want single-player, and a solid, reliable, accessible multiplayer for those who prefer that mode. Or am I wrong?

[Misconduct]

Javolenus and I agree, I just started a campaign for COI, sadly I just ruined all my memories of conquest in 15 minutes when on the third map - I remembered the AI loves sticking tanks out in the open, so I placed an 8.8cm in some tree lines and simply waited for the AI to get bored.

5 minutes later a pair of T-40s and a KV-1e were burning while I took only 3 infantry losses. Sadly I couldn't continue on the 4th map because it was corrupted (doh).