Frogboy Frogboy

The AI Meat Grinder

The AI Meat Grinder

image

This is a little map I made called “AI meat grinder”. Nothing fancy but it is a good testing ground for the AI in terms of building an economy, managing resources, and finding different ways to play the game.

There are 4 islands. How should the AI get to them? Should it build ships? If it gets the Earth shard, should it raise a land bridge to the others? What about the precious few resources in the world. What if there are no good resources near by? How will it decide who to attack or whether it should be friendly with someone? How should it deal with NPCs? How much should they pay for them?

These are all questions that the AI has to answer and that has been the focus of a lot of my work for the past several weeks. Slowly but surely these things are getting in.  I read in the comments area where people wonder if this or that will get in, usually features that are less than 6 engineering hours to implement.

The long, hard, scary time consuming feature is the AI.  It’s the one feature that could delay the game beyond August 24th. If the game gets delayed, the alternative release is second week of February.  Everything else in the game is relatively straight forward (for instance, winner take all vs. retreating, regardless of what we decide, the solution is less than 5 engineering hours).  By contrast, having the AI intelligently colonize an island is a matter of 60 to 80 engineering hours.  As soon as we decided to have more than 1 land mass in the game, things get a lot more interesting, especially since we don’t have pre-made cities or maps where we can script in what the AI does (ah that would be a nice easy “cheaty” way – have the map editor let me put in AI priority points on maps).

We are thinking of including the meat grinder with the game so that people can actually see the AI do its stuff.

146,327 views 89 replies
Reply #76 Top

Quoting Bill_Door, reply 73



Quoting Mike850,
reply 72



Quoting SavageBananaMan34,
reply 1
Good luck! Is September too close to the holidays for you to delay to then if necessary? 


 

September is the projected release date of Civ V.  I doubt that Elemental will want to compete with it.



You call a planned August 24 release date not wanting to compete?  I think Elemental will want as much direct competition with Civ V as possible, I honestly think Elemental will wipe the floor with Civ V, if given the chance.

Of couse, some people will buy civ V because it is a better known franchise.  But I think Elemental will be a far superior game.

What is this "Civ V" you speak of???? lol j/k

Agreed though. My money is on Elemental...ALL THE WAY BABY, YEAH!!!

Reply #77 Top

I have been waiting for CIV V for years. I played the first CIV , I have played the last. That said, I will play Elemental first rather than CIV V.

Reply #78 Top

ok I registered an account especially for this msg....

 

PLEASE PLEASE PLEASE include an option for cheating AI....

also it would be great if we can set what and how much the AI should cheat :)

 

it will not take too many hours to do :) - i am a game developer so i have some idea how much it can take

+ it will be great bonus to the game

Thank you!

 

P.S. I am tired of [pretty graphics - shallow gameplay] games....  so i am waiting anxiously for this one

 

Reply #79 Top

The AI will cheat (get additional resources) on levels higher than "challenging".

Reply #80 Top

Quoting Bill_Door, reply 73

September is the projected release date of Civ V.  I doubt that Elemental will want to compete with it.

You call a planned August 24 release date not wanting to compete?  I think Elemental will want as much direct competition with Civ V as possible, I honestly think Elemental will wipe the floor with Civ V, if given the chance.

Of couse, some people will buy civ V because it is a better known franchise.  But I think Elemental will be a far superior game.

Well, for example, i'll play both but so far i prefer Civ 5 MP mechanics (HEX combat like in Wargames!) to Elemental so if they'll be released at the same time i'll play Civ 5 first. Personally, i will prefer if i'll not be forced to choose between two. Besides, Elemental will be much better in, say, half a year after release with patches. Civ 5 will be better with an expansion, there will be no big difference with patches. So, it makes sense to play Civ 5 then improved Elemental then Civ 5 expansion etc. :)

Reply #81 Top

This is addressed to pretty much everyone who is thinking along this line:

How much "beta testing" were you part of for Spore? Or The Sims? Or Civilization IV? Or what have you?

Don't assume because we have public betas that somehow those public betas are somehow representative of the full project.

Every public beta (every. single. one) that I've done -- going back to 1993 with the OS/2 version of Galactic Civilizations has had this same thing go on.

As other game developers can tell you, the actual "game" is maybe 10% of the total project in terms of engineering hours.

Back in 1994, I made Star Emperor over a weekend - by myself (more here). That 2 day period included modifying the AI for the different game mechanics, a new user interface, new game mechanics, new screens, etc. Two days. The game went on to be a big hit. (Star Emperor was basically Warlords in space).

The point is, what takes actual TIME in game development isn't the game. It's the assets and engine. Spending 3 months on "polishing" or "balance" would be, literally, insane.

What typically delays a game are stability and missing features. Stability is something I'm pretty confident in nailing long before we go gold and other than the AI, the features are in. Internally, we're in polish and balance mode already and awaiting beta 2 and 3 for more feedback.

Thanks for the reply.  I'm aware that the engine is by far the most time consuming but gameplay is by far the most important (IMO anyway).  Don't companies like Blizzard spend huge amounts of time on balance and polish?  That's the impression I get but I have never actually been in one of their betas.

I just find it hard to comprehend how all these moving parts can get balanced that quickly, not to mention that many aspects aren't even there to get balanced and could have unforseen effects on other areas of the game.  I guess your internal version is much further along than what we get so I could very well have a different opinion if I saw what you saw.  Personally I would like to see everything now, whether complete or not but I guess that's not feasible.

Anyway, since you have done this many times before I'm sure you are right.  My inexperienced self just has trouble seeing it at this time.

 

Reply #82 Top

Quoting edpfister, reply 81

Thanks for the reply.  I'm aware that the engine is by far the most time consuming but gameplay is by far the most important (IMO anyway).  Don't companies like Blizzard spend huge amounts of time on balance and polish?  That's the impression I get but I have never actually been in one of their betas.

And then they balance it with patches. Yet, Starcraft become what it is now (the most balanced RTS) only after the expansion.

Quoting edpfister, reply 81

I just find it hard to comprehend how all these moving parts can get balanced that quickly, not to mention that many aspects aren't even there to get balanced and could have unforseen effects on other areas of the game.  I guess your internal version is much further along than what we get so I could very well have a different opinion if I saw what you saw.  Personally I would like to see everything now, whether complete or not but I guess that's not feasible.

Anyway, since you have done this many times before I'm sure you are right.  My inexperienced self just has trouble seeing it at this time.

That's how almost everyone develops games. Even a "last-ditch effort before gold master" kind of balance is good enough for a singleplayer game, most players don't notice. Blizzard is making competitive multiplayer games where balance is hundreds of times more important. AI doesn't complain when someone kills it 100 times in a row on a highest difficulties using 1001 exploit in game balance and AI stupidity. Players in an imbalanced multiplayer game WILL complain and stop playing.

Say, Civilization 4 wasn't balanced for multiplayer either, Frogboy is right in that. Actually, Firaxis never even tried to do it, and they broken some popular multiplayer formats with the new rules in expansions.

Reply #83 Top

 

Please allow the AI to intelligently use navel units as well as navel transports. In AOW:SM this is the one problem that I have with it. It does not use Naval transports very well at all. So making an Island hopping map is only good for multiplay because the AI will stay on its island and never send settlers or troops to other islands unless it has flying units then they are good to go. Now the AI does use aggressive naval and water units well.

 

Oh and please have Independent AI units in the game. And it would really be cool if there were different factions of AI.

 

Reply #84 Top

I say wait until the Feb release so that you guys have more time to work on the game.

Reply #85 Top

Quoting Zaisha, reply 61



Quoting blk_sheep2004,
reply 60



Quoting Saije,
reply 56
I hate the way business works... I want to live out in the jungle.

It seems ridiculous to have to pay a retail store to sell and make profit on your stuff.

That would be like tipping your tax collector. I prefer living in a small world, by myself... Without annoying fools around that I have the urge to kill.


 

So what do you actually think is the point of retailing? Because of all the kindness in their hearts, the retailers allow producers to sell their stuff on their shelves? Ofc not. Everyone needs to make a buck. the producers need to spread their produce and the retailers give that option, for a fee, obviously. It's all about interdependency, just like in all aspects of social life.



Ehm, there's this thing called a "markup". Retailers make a profit by charging customers more for products than the retailers buy them for. I'm sure charging your supplier for buying their stuff is nice for the retailer, but you have to admit it's a little weird. (I doubt a dairy company has to pay a supermarket to sell their milk and butter, for instance.)

But it's industry standard, so publishers just have to live with it. Might (market share) makes right.

Actully the Dairy company does pay the Supermarket to sell thier Milk and Butter this is how retail works.

 

Reply #86 Top

Depending on how modable the AI is you might actually create a whole subgame.  (see OTTD AI developmentfor what can result)

Reply #87 Top

Out of curiosity: do you guys use some form of Q-learning (a type of reinforcement learning) to train/rebalance the AI ?

Can't wait to see the game!!! Cheers :)

 

Reply #88 Top

Quoting Hawawaa, reply 75
(Good memories of Rise of Nation Islands battles against ai and people...)

 

Early is good but then again... longer for more features...

 

Generally, when software is in its beta phase, it's "feature complete." Probably more likely to lose "features" as time goes on; users may find issue with certain mechanics.

Reply #89 Top

Quoting SPARTANVI, reply 88

Generally, when software is in its beta phase, it's "feature complete." Probably more likely to lose "features" as time goes on; users may find issue with certain mechanics.

This game is not anywhere near feature complete.