Difficulty Level as a Slider Bar?

Hey everyone!

I have never played GalCiv, but what I have seen looks very intriguing. Just one question though ... the difficulty level you pick seems to have a number associated with it 30 as Fool, 60 as Bright etc, which if I understood one of the interviews correctly shows the percentage chance that AI abilities will be used.

Well if this is the case (and only at high levels does the AI get any other bonuses such as extra revenue etc), why is the choice of difficulty levels broken up into discrete units? Why not make it a slider bar, so that someone who finishes a great game says "Wow! That was fun, but I think I am going to try it at a difficulty just a wee-itsy-bitsy bit higher" Kind of like a physical fitness program for the mind ...

Just my 2 cents ...

Dano
340 views 4 replies
Reply #1 Top

This is RustyBlade here. Just another user. Below I pasted verbatum the lead developers responce from a very long previous thread. These guys really care about getting this right. I'm betting they will.

Just to work through the math here so that people understand exactly what we mean by colony resource handling:

Let's say a colony normally gets 100 gold from a colony.

Before it reports how much gold actually gets put into that player's economy it goes through this function:

MONEY = SubTotal * (40 + Intelligence)/100;

So in our example, subtotal = 100.

IF the AI is set to "normal" intelligence, then it would be:

MONEY = 100 * (40 + 42) / 100;
MONEY = 82.

So at normal AI intelligence, the AI player would only get 82.

Realistically, the numbers in GalCiv are much more like this:

Subtotal = 24
MONEY = 24 * (40 + 42) / 100
MONEY = 19

Which in the bigger scheme of things is not a big difference, the AI is penalized slightly.

Now, that said, in some future update we could do something like this:

Intelligence is broken up into two parts.
(a) What % of its income it gets to keep between 1% and 200%
(b) How "intelligent" the AI is in terms of going through the algorithms.

But for those who haven't played the game, it is important that you play before worrying too much. The object of the AI is to provide youw ith a fun challenge.

Our goal has been to make a game where the computer players play like humans. Our system is based on literally a decade of game AI work. We're trying to make human-like players.

A REAL human is not going to be totally efficient at handling their economy. Players of Civilization, for instance, know the trick of decreasing their research spending near a new discovery so that they get it in the same # of turns but without having to invest so much on research. A real human is not going to mess with that every turn though.

The algorithms in GalCiv are designed to provide a human-like behavior. That's why we PENALIZE the AI at most of the difficulty levels -- a human player penalizes themselves because humans are not infinitely anal-retentive. Computers can be. We simulate this in taking away some of their resources and not having them act on information they have available.

But we did also want to ensure that someone who has truly TRULY mastered the game could still get a challenge which is why we created the "Incredible" difficulty rating option (might as well be called "Impossible" because I predict you'll have contests on how many turns the player survived in a impossible game).

I hope this clears things up.




by Avatar Frogboy - 3/6/2003 6:28:48 PM
Reply #2 Top
Dino, the intelligence selection is done thru a thumbwheel, that goes from fool to incredible. There are 8 levels of intelligence, with increasing difficulty.

Here is a post from FrogBoy that outlines how it is implemented.

"One of the things we did do during the release candidates is make the game much easier to win on the lower levels. We found that experienced strategy gamers would get creamed even when the AI was set to "fool".

But at the same time, at the medium to higher levels, the AI is much more intelligent than it was in the betas. In fact, the beta GalCiv AI was, well pretty crummy compared to the gold version in terms of intelligence (not saying the gold is "harder" to beat, but that the AI is more intelligent in how it behaves).

In the betas, each AI level simply increased the intelligence rating of the player by say 15 points. Here's the problem we had with that:

You'd play on "normal" and get a challenge and then you'd up it a level and get creamed. So a lot of the extra month of development was spend play testing and adding more difficulty levels.

So now each computer player can indvidually be set to the following difficulty levels:

[IntLevel0] Fool
[IntLevel1] Beginner
[IntLevel2] Sub-Normal
[IntLevel3] Normal
[IntLevel4] Bright
[IntLevel5] Intelligent
[IntLevel6] Genius
[IntLevel7] Incredible

With the corresponding intelligence being 4, 14, 36, 42, 50, 60, 65, 100.

We may tweak these or add more after released based on what you guys say.

So now, you have sub-normal, normal, bright, intelligent and genius quite close together. I have not personally won with all the AI's set to genius on a medium sized map or higher. And incredible is just, well the AI gets lots of extra stuff.

Speaking of extra stuff, how do the resources get handled?

The intelligence ratings do two things: 1) They determine how much money they get from colonies.

So in pColony->CalcTaxesCollected() at the end it gets:
ulTaxes = ulTaxes * (40 + ulIntelligence);

So at intelligence = 60 (intelligent) it's getting the same from a colony as you. At Genius it's getting a bit more. And at incredible, all bets are off.

The other way intelligence matters is that it determines whether the AI is going to a) Do a thorough analysis on what players are doing b) Which AI scripts are available to use, and c) Determines whether the AI is going to do anything about it.

Example:
An AI with intelligence 60 will be able to check out its sensor areas and see if you are building up for an attack, fortifying starbases, trying to culturally dominate it, etc. It will then have to roll to see if it is going to do anything about it. I.e. "Do I recognize this as a threat or am I too dumb to see this?". If it rolls that it is going to recognize the threat then it has to roll whether it's goign to actually do anything about it. And if it is going to do something about it then it rolls to see if it will tell the human player about it (if it's a human player that is the threat).

IF the AI FAILS the roll, the human player gets a message from the AI letting it know it's on to what it's doing. Otherwise it quietly lowers relations and prepares for a war.

If you play on say Genius, and build up star bases and ships in a sector you're planning to attack you'll notice ships from other sectors quietly coming in to fortify. Fighters shadowing your transports. Etc.

But on lower intelligence levels, none of this happens. Which is one of the changes. So when you get the game, on Fool and Beginner you should be able to play the game without getting creamed now. This is different than it was in the beta when fool was = Int 15 and the next one up was 30."

by Avatar Frogboy - 3/5/2003 4:02:14 PM

~SDC~
Reply #3 Top
Thanks for the info!

But doesn't this show that the ideal control for the difficulty level (or at least the intelligence level) would be a slider bar that ranges from 1 to 100?

Dano
Reply #4 Top
Dano,
you are correct in that a slider bar would work, but the vast majority of players would have no idea of how a number from 1-100 would relate to difficult in play. I'd rather trust Stardock to set 'levels' with names indicating the challenge (bright, genius, etc) than have to guess what settings do what.

During the beta they tweaked these values a lot. Initial games had the intelligence levels 15 points apart (linear spread) and this didn't work well because the a linear change to intelligence does not give a linear change to difficulty. The current values work well.

Paul.

~SDC~