Wargamer vs AI – Single player experience

[NikiAlex]

Hi fellow gamers,

I had most fun in video wargaming when I have to make multiple smart decisions against adequate computer component. I’m confident that most of you experienced the same in strategy games.

So during the process of extensive AI prototyping we identified several major areas where player should have the opportunity to decide how to outplay the computer. All of these are from the perspective of modern Air/Land war.

That’s why I wanted to share with you our key challenge circle. Could you as experienced gamer share your expectation of AI capabilities in full-blown warfare strategy game?

From the list we have, what would you miss and what you consider not so fun to play with?

Air Operations:
Air superiority
Escort
Ground attack
SEAD
CAS
Air Lift/Supply
Rebase

Ground operations:
Force Stances: Advance to objectives, Defend own strategic assets, call for reinforcements, retreat
Force formations – grouping by complementary combat and support functions
Combat recon specific behavior
Logistic support map allocation

Warfare economy:
Weapons’ adaptive purchasing patterns
Unit training ratios and composition

I will be happy to get your valuable input and further discuss in-depth any of the options.

[Mobeer]

Any of these could be good or bad things to have; which of these to include depend upon how your game approaches its subject.

To me an enjoyable AI opponent does a good job of managing whichever options are available, the list of options available does not determine if the AI is good.

[NikiAlex]

Agree, it's all about player experience.
It's hard to judge how it works before you face it in-game. We are working on this.
Ambition is to put hard, but fair challenge even to most experienced war-gamers.

[aaatoysandmore]

I would miss ground attacks and air attacks. Wouldn't care if they left out naval and all that other stuff.

I'd like someone to make a new WWI air warfare game like "Eagles" by SSI only updated graphics and plays more like that tabletop miniature game Aces of the Sky I think it were.

The only naval game I would play is like Carrier Wars with planes. Iron tubs always struck me as sitting ducks in war vs planes.

[wodin]

I'm useless at nearly all games which is a blessing as only very few AI's have been so bad it ruined the game for me.

[major.pain]

A good place to place to start would be to have a good look at the AI in the Close Combat games.
Then disregard everything it does and you be well on your way

[altipueri]

Me too Wodin. In fact I generally find I want a weaker AI not a stronger one. This is one of the benefits of having about a hundred games, none of which I specialise in.

It seems there are a number of games in which some people become super specialists and then blame the AI for being dumb. Frankly, AI is for casual play, PBEM or head to head or whatever form of human vs human is for specialists.

PS - this is actually the first time I seem to have been able to get the funny Matrix login code thing right for ages on this PC. I'm so dim it usually thinks I'm an AI robot. Whoever thought old people could get their "Shift" or "Caps Lock" timing right clearly wasn't an old person.

Alcohol is sometimes involved.

[Max 86]

The problem with AI is that it can never formulate a comprehensive plan that includes all that you mention, coordinating land, sea and air assets. Also, AI has trouble with reacting or adapting to changing situations, identifying new priorities, etc.. It just sticks to the script in most cases.

I will say that you may want to take a look at Flashpoint Campaigns, that AI always seems to know what I am doing! I can never get more than a minor victory at best and others end in a draw or a minor loss. Best AI I have played against. [&o]

[Kuokkanen]

ORIGINAL: Max 86

I will say that you may want to take a look at Flashpoint Campaigns, that AI always seems to know what I am doing! I can never get more than a minor victory at best and others end in a draw or a minor loss. Best AI I have played against. [&o]

Sounds like fog of war does not apply to AI = AI cheats

[warspite1]

ORIGINAL: wodin

I'm useless at nearly all games which is a blessing as only very few AI's have been so bad it ruined the game for me.

warspite1

Remove the word nearly and that sentence could have been written by me! [:)]

[operating]

Sid Mier's games used to or still do give the player the option of selecting the AI's commander, plus a number of other AI attributes the AI could perform. My point here is: Many of the games marketed today only have selections such as 1) hard 2) average (normal) and 3) easy. What these selections lacked was the ability to shore up the AI on it's weaker points, like some would say: "Oh, the AI's naval ability sucked!". My feeling is: might there be an in game selection to have the AI favor a stronger role in the naval aspect of the game, plus other selections to improve the AI's ability in certain ways, instead of complaining to the dev's to make a change in the patch about AI deficiencies .

[NikiAlex]

Quite interesting replays out there!

I'd like someone to make a new WWI air warfare game like "Eagles"

aaatoysandmore, what you like most about this WWI air war settings? It’s essentially the beginning of effective use of military aviation.

wodin, altipueri Not surprising many of us will enjoy to loosing occasionally, but game been too hard to beat is certainly not fun

The problem with AI is that it can never formulate a comprehensive plan that includes all that you mention, coordinating land, sea and air assets. Also, AI has trouble with reacting or adapting to changing situations, identifying new priorities, etc.. It just sticks to the script in most cases.

I will say that you may want to take a look at Flashpoint Campaigns, that AI always seems to know what I am doing! I can never get more than a minor victory at best and others end in a draw or a minor loss. Best AI I have played against.

You are damn right! In the end it’s not true AI, but of bunch of scripts regardless not many patters and behaviors it covers it’s not going to adapt and plan in a more advanced way.
Having an AI which knows everything about the player is something we will avoid. In most of the cases it’s just too obvious and considered by many to degrade player experience. For example we will not allow AI to calculate the precise strength of your units. AI behaviors based decisions will be calculated with imprecise stats/ration data

Sid Mier's games used to or still do give the player the option of selecting the AI's commander, plus a number of other AI attributes the AI could perform. My point here is: Many of the games marketed today only have selections such as 1) hard 2) average (normal) and 3) easy. What these selections lacked was the ability to shore up the AI on it's weaker points, like some would say: "Oh, the AI's naval ability sucked!". My feeling is: might there be an in game selection to have the AI favor a stronger role in the naval aspect of the game, plus other selections to improve the AI's ability in certain ways, instead of complaining to the dev's to make a change in the patch about AI deficiencies .

I guess you are referring to variety of patterns for the AI behavior and the ability this to be controlled by the player.
Unpredictable patterns for AI could bring a lot of variety into the game. For example on the start of the war campaign AI could with a focus on air power, later to switch to infantry and armor patterns. This will bring more surprises to the player who needs to adapt his approach to AIs general strategy. Whether player should be able to set AI to specific behavior is an open topic.
There could be even more interesting discussion about difficulty settings.
IMHO they should influence more on resources available to AI thus increasing the challenge and to lesser degree trigger some drastic changes in the behavior.

[operating]

NikiAlex.

You came close to hitting the nail on the head, none-the-less a good assessment of my comment..[;)]

Bob

[Max 86]

ORIGINAL: Matti Kuokkanen

ORIGINAL: Max 86

I will say that you may want to take a look at Flashpoint Campaigns, that AI always seems to know what I am doing! I can never get more than a minor victory at best and others end in a draw or a minor loss. Best AI I have played against. [&o]

Sounds like fog of war does not apply to AI = AI cheats

That's what I thought too but the makers swear that's not the case. Either way, when every scenario ends in a draw, the game loses its appeal after awhile.

[aaatoysandmore]

I liked the AI in Age of Wonders I as you could give it several personalities instead of them all playing the same in their later versions. The scorch n burn ai is one of the best in most games that have them. An AI that razes one of your long term cities has got to be good. Sure makes you feel bad when a city you have spent the entire game building gets burned to the ground. Those are good ai's to me.

[CGGrognard]

Just jumping on this thread concerning AI in games with a thought. I'm not certain how feasible it would be, but consider a developer collecting moves of players during an online session to build a more robust AI. Or, perhaps in single player, the game could connect to a server to stream strategies (from saved files) from multiplayer games.

[operating]

ORIGINAL: CGGrognard

Just jumping on this thread concerning AI in games with a thought. I'm not certain how feasible it would be, but consider a developer collecting moves of players during an online session to build a more robust AI. Or, perhaps in single player, the game could connect to a server to stream strategies (from saved files) from multiplayer games.

Wouldn't this AI information "sponge" also be receiving bad strategies and tactics and not knowing the difference between bad and good and the other side of good to one player might be bad to another player, it certainly would be interesting how this sponge would interpret everything being thrown at it, it could up in an asylum. [&:]

[aaatoysandmore]

But, there also are many game developers who don't know strategy and tactics as well as some also.

I feel like something in the program that records the certain moves and units purchased of each individual player vs the ai is a better idea. It records and plays strategy and tactics against the player it is playing not everybody else.

If you have a tendency to move all your troups from the sides of the map the AI will record that and play for it the next time you play.

Of course an exploiter human player just can't be beat by an ai unless their are parameters set against him from the start like you can't change certain units and settings for one more game afterwards of beating the AI. Thus the AI has a chance of coming back and beating you because you won't be allowed to exploit it the next game. It will also seed the die rolls so you can't crash the game and then come back and change your strategy. Die rolls will be "random" when they take place in the game so the human player will never know when the die rolls are set so he can try to crash the game. They may be randomly set on turn 1 for turn 3 and such like that. So, the human player can't get in a pattern of crashing the game at the first or end of a turn because the random number die rolls have already been set 3 turns ago. [:'(]

I've always thought this a better way to make an ai. Record the player (the one playing against the ai) and use those recordings against him in the present game and in the future game 2 and/or 3. The player is not allowed to change his setup in game 2 nor the units that he purchased. Seed the die rolls randomly with them being rolled in 2 or 3 turns or even more back from the present turn. Like seed roll the die for turn 5 in turn 1. That'll stop some of the exploits.

[operating]

Sid Mier's games are RTS, so there may be quite a difference between the AI in a RTS game versus the AI in a turn based game...

[NikiAlex]

Just jumping on this thread concerning AI in games with a thought. I'm not certain how feasible it would be, but consider a developer collecting moves of players during an online session to build a more robust AI. Or, perhaps in single player, the game could connect to a server to stream strategies (from saved files) from multiplayer games.

I think it’s a huge effort this to be implemented properly. It’s not only about collecting/recording data of player moves. This data need to be analyzed and mapped to the vast number of different situations in the game. It will be like an evolution of a chess playing program, that need to select the move based on big number of conditions and in most of the cases this will mean a lot of processing time. With the current hardware IMHO it might be applicable to some degree only in turn based game. In RTS it’s a game over for the CPU.

Sid Mier's games are RTS, so there may be quite a difference between the AI in a RTS game versus the AI in a turn based game...

I don’t know if I’m the only one that instantly associates him with turn-based masterpiece like Civilization.
Maybe the biggest advantage AI has in turn base game is the luxury of the processing time between the turns while player’s objects are static.

What are advantages/disadvantages of TB vs RTS model… I’m sure a lot of discussions could arise here