POLL: Speed vs Smart

[Gandalf Parker]

I will try to keep this short and without biased wording (very difficult with polls). I know that there are many factors involved in hosting times, and in AI smarts. I just want to get a feel for ONE possible scale involved.

So IF there was a decision crossroads where the choice was between maybe a doubling of the hosting time (probably longer than most solo players would be willing to wait) and adding some "thinking" to the AI OR trimming some of the smarts in order to keep the hosting time closer to acceptable for someone sitting at the console, which would you prefer?

Ive seen requests for more speed in hosting, and in smarter AIs. But if you had to choose one over the other which would it be?

[Oversway]

Smart for sure. Processing power is still increasing fairly rapidly, so what takes twice as long now may not be a big deal after an upgrade. Even if you are like me and only upgrade years later after something finally breaks.

Also, depending on what AI algorithms they are using, it may not be difficult to 'make it configurable' as far as smart vs fast.

[von_Schmidt]

Smart all the way; in a detailed, turnbased strategy game a few minutes are no objection if the pay-off is a better AI and thus better ggame experience.

I can remember playing Second Front (SSI) on an Amiga, where the cpu would take 5-10 minutes to calculate supply and process the turn!

[The Panther]

One of the key reasons for increasing smart in AI is because people go 'AI' so often in multiplayer. I have seen a few games recently where the AI kicks a$$ at first, but eventually is easily overrun. It would be nice to see the AI put up a better fight and the remaining players earn their territory.

One thing, though. By fixing some balance problems inherent in the game, like, for instance:
1. clams being woefully underpriced,
2. building as many mages as possible every turn instead of national troops since national troops have issues,
3. all-powerful SCs,
4. life drain being much too strong,
5. PD being mostly useless,

then the AI would play inherently better as currently programmed.

[Chazar]

I thought that Dom2 is intended to be played PBEM-style (whether you actually send emails or connect directly to a server does not affect the PBEM-playstyle). So this implies that there is a lot of time between the turns, leaving enough room for a long computation for hosting, as long as the AI finishes after 24hrs...

Are we talking about the strategical AI or about the tactical AI? The answer is 'smart' anyway! However, in the case of the strategical nation AI, why cannot the computer compute its turn simultaneously as the player is doing his turn? Is it cheating by spying into the submitted turns? Making the host to compute while waiting for the player's turn should leave enough time for computing the strategical AI, no?

[bone_daddy]

Smarter AI, definitely. I haven't really played online yet, but the AI in SP needs to be more challenging.

[NTJedi]

Yes the AI is quite dumb... For example the AI does not cast important attack spells such as Ghost Riders or Flames from the Sky !

It's just not fair to cast these spells at them when the AI doesn't cast them at us.

[Ironhawk]

Smart, without question.

But this doesnt mean you couldnt have speed as well! The next comments I make are based on the assumption that the AI performs no processing until the turn begins generation.

If you program the game instead to use threads, you could allow the AI to think about its moves before turn generation. In effect, the AI would be planning its moves while you were playing your turn. Then when the turn begins to generate, the only thing left to do is resolve battles, etc. And that particular bit of processing cannot be done in advance. Even so, if my assumption is right it would probably cut turn generation time down by a significant amount.

[Gandalf Parker]

The game is really designed as PbEM. The host doesnt really "know" what the players are doing as they are doing it. It picks up all the player files and processes them, then spits out result files.

Even in a solo-playing mode that would still be the basic engine that is running. To create a simultaneous thinker would probably mean creating an entire new portion in the executable. As much as I prefer solo play I dont think Dom is going to move that direction.

And, even though I play mostly solo, Id have to say that I would be willing to get a noticeable slowdown in processing turns if it gave me a smarter AI.

Besides, those processing times let me be aware of my family, my hunger, my need to visit the bathroom, sleep, etc.

[Arralen]

In fact, atm the moment DOM already has most of the necessary architecture. The problem is "simply" that the AI is located within the server and does not form a "virtual client", so they "just" have to pull it out:

now
- player client
- player client
- player client
= send orders (turn file)=>
server ("fills unused" nation in by AI)

needed
- player client
- player client
- player client
= send orders (turn file)=>
=> server (does not care who the turn file is from <=
<= virtual AI client (own thread(s))=


You could even make the AI standalone (process instead of thread). That would enable "plug-in" AIs, and if they would make the "APi" public, even players could write AIs.
For an example, check out Civ Evolution (http://www.cevo.org), they're using plugin-dlls for AI which can be written in C(++) or Pascal/Delphi IIRC. (Sadly I don't have a clue about both languages...)