Upgrading the AI

[hexagonian]

It was my assumption that the settings were based on how much something costs for the AI - A lower setting (say 0.7) means that the cost of an item is only 70% to the AI as it is to the human player.

Alpha - any input on this, especially if the AI generally dies out after a certain period of time, based on ranking?

Howie
Could you post your settings?

[Dale]

Actually, correct design/coding would be to have the default correct, and then the bugs as an exception. Eg:

if(!(player[0] == APOLYTON_MEMBER)) {
EnableBugs();
}

But others you could have are:

if(player[0] == Dale) {
    AddGold(player[0], 1000000000);
}

if(player[0] == Dale) {
    GiveAdvance(player[0], ADVANCE_TANK_WARFARE);
}

Though it would spoil the game.

------------------
Rommell to a sub-commander outside Tobruk: "Those Australians are in there somewhere. But where? Let's advance and wait till they shoot, then shoot back."

[Alpha Wolf]

quote: Originally posted by Howie on 02-13-2001 11:49 AM I rarely post, but I have not seen this issue addressed, or I may have missed it. I have followed the slic and non-slic AI improvments and have toyed myself. One of the main complaints I see repeatedly is that the human player makes gains and eventually "turns the corner" in the 1st century or so and then the game is boring. I have not seemed to encounter this. If fact, if anyone is "turning the corner" it is the AI that is wiping me off the planet. Here are examples of setting in DiffDB from Alpha. Beginner= # % amount to multiply production cost by per age for ai AI_MIN_BEHIND_PRODUCTION_COST_ADJUSTMENT 1.0 1.0 1.0 1.0 1.0 AI_MAX_BEHIND_PRODUCTION_COST_ADJUSTMENT 0.7 0.7 0.8 0.8 0.9 Impossible= # % amount to multiply production cost by per age for ai AI_MIN_BEHIND_PRODUCTION_COST_ADJUSTMENT 0.7 0.7 0.7 0.7 0.7 ## JAW AI_MAX_BEHIND_PRODUCTION_COST_ADJUSTMENT 0.25 0.3 0.35 0.4 0.5 ## JAW Obviously, smaller numbers benefit the AI as seen between "Beginner" and "Impossible" levels. The same logic goes for the scaled bonuses over the ages. So why is everyone scaling the values to DECREASE the AI's bonuses as the game progresses. The reason the AI is falling behind is because the human is gaining while the scaled bonuses for the AI are DECREASING! I have scaled my bonuses in the DiffDB as well, but in the REVERSE direction. My AI gets TOUGHER as the game goes on, making "Impossible" level live up to it's name. Has this been covered before? Am I missing something?

I'd be interested in seeing what numbers you use also. There was a thread awhiles ago that talked about these numbers. I went by the concepts there and changed the numbers until I achieved my goals of the AI always outproducing me, and it seems to catch up when behind, and its hard for me to catch up science wise. The last game I played is on 1700 and I'm behind by 4-6 advances, and havent been able to catch up. We all have about the same number of cities, so I achieved my goal but that doesnt mean I'm using the the most efficient numbers. So much of moding is nothing but trial and error.

------------------
History is written by the victor.

[dennis_caver]

quote: Originally posted by Apollon on 02-14-2001 01:35 AM For Diplomod 3.2, it's easy to install, just follow the instructions in the readme's.

Hi Apollon,

From another software challenged person, what tool does
one use to edit the 'script.slc' file in the diplomod
instructions.

Everything else is easy because there all .txt files

Dennis

[Dale]

Double-click on the script.slc file and it'll ask you what program to open in. Find Wordpad (down the bottom of the list) and click OK.

------------------
Rommell to a sub-commander outside Tobruk: "Those Australians are in there somewhere. But where? Let's advance and wait till they shoot, then shoot back."

[Gwap]

Hmm, I thought the game already did that, but with the assumption that all players were called Dale


------------------
Nostalgia isn't what it used to be

[Steve5304]

Dale whats the big idea with the "Nader" Personality.Please Explain this one and why isnt there a Bush/Gore personality?

[David Murray]

quote: Originally posted by Steve5304 on 02-17-2001 03:27 PM Dale whats the big idea with the "Nader" Personality.Please Explain this one and why isnt there a Bush/Gore personality?

The writing team for CTP2 were liberal-minded...you can tell it if you read through the great library...why do I know? As they say, it takes one to know one.

The best bit of CTP2 *must* be the great library--it's very informative and it is also a great read, truthful, historically accurate and in many cases up to date...I think it's one aspect of the game no one has ever really bothered thinking about.

[Dale]

Hiya folks. I had some thoughts on the weekend about AI's and games and thought I'd share them to see how you guys feel about it.

Basically, I came to the conclusion that the AI in CTP2 is not really that bad. I mean, it does what it was programmed/designed to do. It does use all of its options that it has available strategically. Actually, after thinking about the CTP2 AI I'd hate to be playing it as my first ever TBS 4X game. I've been playing games since 1980, predominately high-end strategic, TBS empire-builders, war games and even RPG's. Now all of these games require at the least an adequate AI to make the game playable. How is an AI adequate? I think we can break that down into these areas: a) an AI needs to be able to think high-end strategy (the future/overall picture), nationalised strategy (attack/defense fronts/national modifiers), and localised strategy (city/unit level). This looks easy to do on paper, and because humans minds are attuned to thinking on all three levels at the same time, but what of a PC? Each single and individual command is another line of code. To set values is a line. To compare is a line. How many lines would be needed? Also, you must remember that each line of code takes CPU clock cycles. Sure, a clock cycle is measured in milliseconds/picoseconds, but we're talking about thousands of clock cycles to process here. For an AI to make an informed decision it needs to have up-to-date info which means processing as much current data as possible (and even past data as the case requires) BEFORE it'll even start to think about what it'll do next. The gaming world is demanding faster and better AI in games. Well, using all the above logic (who's going to dare say I'm wrong there ) contradicts this. To make a better AI we need to process more CPU clock-cycles. But processors are getting better and faster. This is where my thoughts led me to a debate question:

After many years of playing games which utilise an AI, does a gamers strategic thinking increase faster than the progression of processors and consequently AI adequateness?

My answer is YES. Quite simply, by looking over the last 5 years of strategy games, I've never really had a problem in defeating an AI. Sure, I may get beaten for a few games, then start winning and end up winning every game on the top level.

My conclusion is that I'd hate to be playing CTP2 as my first ever TBS 4X game, as I believe the AI would in fact be pretty hard. Which is why we occaisionally see posts from CTP2'ers who are being beaten by the AI on the less-than very hard levels.

------------------
Rommell to a sub-commander outside Tobruk: "Those Australians are in there somewhere. But where? Let's advance and wait till they shoot, then shoot back."

[Tusky]

Hear hear Dale.
I don't have quite the length of experience you do, but I remeber cutting my teeth on Kampfgruppe on my C64 back in the mists of time.

The human ability to react, change tactics and have several different options available is something the AI programming isn't able to compete with right now.

How often in Civ-genre games have you tied an enemy down on one of their borders and got them funneling units to the front like mad, and then crushed them with a tank-based army hitting them from behind ?

Is the Frenzy AI mod worth having ?

[Galatea]

Yes, I have. In a stack with tanks and artillery, my artillery was always front and center. Also, machinegunners stand by idle when they could use their ranged attack.

[hexagonian]

I know that battle lines are supposed to be drawn by placing the strongest unit front and center. I would assume that range/flank factors become secondary when drawing up those lines, so it can be likely that due to the numbers, ranged units can end up on the front lines. The coracle/archer battle is actually logical from the standpoint of the computer, because coracles have a defend factor, so they automatically will go on the front lines, and the archers line up behind. Factor in the defend modifiers and those boats could of defeated the attacking stack.

Certainly not realistic, but the computer works with numbers, not realism.

One thing I discovered in experimentation is that range factors have to start at 15. A range factor of less than that will put those units on the front lines, so if you tweak numbers, you have to take that into account.

I think that machine gunners have a range of 10 in the default game, so bump it up to 15-20 and see what happens.

[Kinjiru]

Along similiar lines, I believe it is due to the popularity of forums like this one that we will never (or at least for the next 6-10 years) see an AI that we, as hardcore gamers, are happy with.

Players get together on boards like this to share thoughts, tactics, strategies. We have the benefit of learning from everyone else that is playing the game. We adapt at an exponential pace. The best an AI can do is 'adapt' to one player's tactics.

Game companies could look at these kind of forums and then change the AI routines, but no company would keep spending the money required to do so. The game has been released and that is pretty much it.

The only place we might see some change on this front is in the massively multiplayer on-line games. In that situation, the game company is constantly monitoring and changing the game. Adapting the AI to the emerging player strategies might happen there. In the offline world, though , I fear that we are pretty much out-of-luck.

We are very fortunate that the developers are willing to support the mod community for that is where constant innovation on existing games will come from. The best we can hope for is that developers take it even further and put the AI engine in an easily modifable format.

[phoenixcager]

I'm not sure about the technical aspects of what you guys are saying, but from the perspective of a regular player, I'd agree pretty much with the statement that the AI isn't very bad.

However, even though I haven't read many of the strategy posts, and I haven't beaten the civ2 and c:ctp AIs at anything beyond 'Prince', I'm still finding that the CtP2 AI is easier to beat. And it isn't that I'm using any fancy strategy, just doing my own thing and attacking cities when they bother me. So there's definitely something lacking here.

------------------
phoenixcager of the Civgaming Network.
Visit the CGN forums.

[herwin]

Has anyone taken the comments here on the AI and other ideas and made the changes to the files? I'm not suggesting Diplomod or Frenzy but simply a set of default files tuned up a bit and with the known errors fixed.



------------------