Does the Distant Worlds AI Cheat?

[LarryVE]

I am playing a game with a medium sized galaxy, 3 races (including me) and a 3 year time limit. Toward the middle of the 2nd year I encountered another race. This race has 4 times my military strength, 4 times my strategic value, 2 times the tax revenue, 4 times the populations, and 4 fifths my population. How can this level of success be so broad across all areas unless the computer opponents are cheating?

Larry

[Spacecadet]

Quite a few ways really.

If you're playing RotS and have the Shakturi enabled then odds are you've met the Guardians, and yes, they will be kick!@# right off the start - part of the story line.

The Empire you met could have found/taken over one or more Independent Colonies.

Your game starting Setup parameters may have "Random" for Age, Starting size, Starting Tech, etc.
If you just left the "Auto Generate 11 Empires" (or something like that) checked, then yeah, you're going to have quite an unpredictable mix of the above Starting parameters.


I really don't think the AI cheats, in fact, I think it probably needs a bit more tweaking to boost it's capabilities, but overall it's not too bad right now.

[Sabin Stargem]

Some of the AI empires expand way more than the other ones. I don't know why, though.

[Spacecadet]

ORIGINAL: Sabin Stargem

Some of the AI empires expand way more than the other ones. I don't know why, though.

I think a lot of this depends on the luck of the draw.

I've seen certain Races just explode in one game, and the next they're a one Colony wonder.

I think it really depends on Galaxy setup/location/resources.

[sbach2o]

ORIGINAL: Spacecadet

ORIGINAL: Sabin Stargem

Some of the AI empires expand way more than the other ones. I don't know why, though.

I think a lot of this depends on the luck of the draw.

I've seen certain Races just explode in one game, and the next they're a one Colony wonder.

I think it really depends on Galaxy setup/location/resources.

Racial parameters also play a role. The most important is propulation growth. The difference between <10% (Quameno, Dhayut) and ~25% (Gizurean, Securan, Atuuk) is just enormous, although it would take a little bit more than three years to really show. The high growth races consistently sweep the low growth races away *almost* every time, the only exception being when they hit on some very bad luck.

The low growth races under AI control, on the other hand, need a lot of luck in the form independents and cross-empire migration to have a chance to keep up.

[cmdrnarrain]

If it does it needs to cheat even more.&nbsp; All AI's cheat and nobody in their right mind should be moan it, unless it is done really poorly, but that is poor programming.

I think that is part of DW's AI problem.&nbsp; It doesn't cheat much, if at all because it is so easy to blow it socks off in terms of worlds, tech, and military.&nbsp;

[unclean]

ORIGINAL: cmdrnarrain
All AI's cheat and nobody in their right mind should be moan it, unless it is done really poorly, but that is poor programming.

I don't know, seems like cheating AI's are one of the pet peeves of the devs since only things that can really be called difficulty sliders in this game actually penalize you instead of help the AI.

[WoodMan]

One of my pet peeves too. I would rather they focus on improving its behaviour, by examining the strategies and tactics of Human players and imitating them. Obviously it would never exceed the human player, because even if it acts like them, it won't be able to adapt like they do to changing circumstances etc.

However, I think if over time it is slowly improved by imitating us (I don't mean program it to copy us or anything complicated like that, I mean tell Eliott how we play, so he can try and get it to do more of what we do) and using the sliders to give it a headstart should be okay.

Even without giving it a headstart with the sliders, new or casual players will still find it difficult if its programmed to basically imitate the proplayers who post their strategies.

[cmdrnarrain]

Everyone plays different and there are multiple paths to win.&nbsp; It would be difficult if not impossible to design an AI which could deal with all of them.

[Data]

If you could, that AI would play against Kasparov or Anand [:)]

[WoodMan]

Making the AI cheat is the easy way out and I'd much rather see a more intelligent AI than one with stupid bonuses.  This makes the game harder, but not more enjoyable.  Making the AI more intelligent makes it more enjoyable and harder.

For example, in another game I play (Blood Bowl) the AI was made harder by giving them better starting teams.  The result was it was slightly harder, but still idiotic, and once you overcome that gap and make your team equal its p**s easy again.  Same would happen in DW, give it an economy bonus, or make its ships toughter, but if its not more intelligent your economy will soon outstrip it along with your military.  Your still playing against an idiot which is boring.  Btw, I don't find DW boring, I love it, but you get what I mean.  [:D] Cheating AI will make the game longer, not harder, your inevitable victory will be delayed as you overcome the starting obstacle. If you lose however, the only reason you lost was because the cheat setting was set too high, you weren't actually challenged in anyway, the odds were just stacked too far against you in the first place.

Of course, multiplayer is always the other easy way out, but I'd probably in the case of DW prefer a better AI.  Both would be good though

[LoBaron]

ORIGINAL: Data

If you could, that AI would play against Kasparov or Anand [:)]

You are understating, Data. [;)]

If you could, that AI would beat the hell out of any high-end chess AIs like DeepFritz as long as the playground does not
consist of a standart 64 field board.

Chess is simple in comparision. I don´t want to belittle chess, I love the game myself. But compared to a game with loads
of ever changing variables, where only a complete asessment of the overall situation plus a individually weighted asessment of a specific
situation can result in a workable decision (and this initiated only by a predefined set of possible intentions), chess is not even a challenge.

The AI of every current came this complex is way below the requirements needed to beat a human on eyelevel. This is not
because its not possible to create an AI which gives orders, but from what process the orders come from.

Considering these obstacles DW AI does quite well.

[Data]

Yap, I think so too. Have read somewhere that the AI on games like this is actually made more stupid as it would beat the crap out of us if let at full potential; I tend to not believe this.

[Bingeling]

It is probably possible to have an AI that could steamroll very early player controlled empires quite reliably. Given the right resource luck and starting races. However, making an AI that don't have a weakness would be very hard. So you can script behavior, but having a generally good, improvising AI is a different beast. The approach of chess robots (search all) won't work, but I don't believe any of us play Distant Worlds at a chess grand master level, anyways

[Data]

yeah....and it also has a 10x10 sector grid anyway [:)]

[LoBaron]

ORIGINAL: Data

Yap, I think so too. Have read somewhere that the AI on games like this is actually made more stupid as it would beat the crap out of us if let at full potential; I tend to not believe this.

We humans are accustomed to laying traps, time concentration of forces based on the circumstances, react to percieved threats
before anything actually has happened, weight potential gain against the cost of protecting this gain in the future, estimate the impact
of prioritizing one task over another, and so on, and so on.

To accomplish this this we use loads of available information which are not neccesarily directly linked to hard numbers, compare with actions
from the past (learning process is a major factor here), pull reqiured data at will from a non standardized pool which we can expand or limit
at will, and access new data in case information there is recongnized as useful, and this is probably not even half of what we are able to do.

A AI comparable to those human traits for a game like DW would have to be to the nth number as complex as the game itself and require processing power not
even remotely available on current desktop PCs.
Also it would require the nth number of developement time compared to the time required for develop the game itself.
Although I seriousely doubt that it is possible to accomplish at all, at least in the more intuitive aspects of human intelligence.

[Data]

Yap, in perfect agreement on this.