Soar_Slitherine
Posts: 426
Joined: 6/7/2020
Status: offline
|
quote:
ORIGINAL: DingBat
I don't have any opinion on whether or not SE would be better of as a WEGO game, but can you explain your reasoning on how it would be better for AI processing? ... I also don't understand how a WEGO game would make better use of multi-core CPUs.
In an IGOUGO game, the possibility for an AI to perform analysis on the game state during another player's turn is limited because the game state changes every time the other player makes a move, potentially rendering any existing calculations out-of-date. Making heavy use of multithreading (i.e. multiple CPU cores) for processing a single AI is not necessarily easy or practical either - you can only compartmentalize the calculations so much, especially if information can become out-of-date because of things like a random element to combat resolution or fog of war being uncovered by moves.
In a WEGO system where all the players plan their moves first and they are then executed simultaneously, it is possible for each AI in play to perform their analysis of the game state in parallel utilizing different cores of a multicore CPU because they are all working from a game state that is fixed until their next turn. Since that means you can make the AI players play in the background during a player's turn, you can also have them take much longer without causing turn processing times to become annoyingly long, provided the player has enough to do each turn.
Obviously, this is more relevant for games with a lot of moving parts that require processing power to account for and potentially lots of different AI sides like Shadow Empire, compared to games with only two players or a limited scale (and of course the game would need to be designed for it from the ground up). However, I do also like WEGO for faster turn times (and less exploiting stuff like reloading a turn to scout different areas or obtain favorable combat results) in PBEM games with lots of players.
quote:
In an IGO game, the AI exactly knows the state of the world and can make the best moves possible. In a WEGO game, you probably have to devote some thought to predicting what the players moves will be, or to "cheating" and examining the players moves and predicting their outcomes. This would absolutely be an additional burden on the AI.
An IGO AI should ideally make some predictions about what its opponent can or might do on its next turn, to avoid putting itself in a vulnerable position. I don't see AI under a WEGO system as radically different in that regard, but being able to fully exploit multicore CPUs should create quite a lot of room to absorb any overhead. But hey, I am not an actual AI expert, I only dabble in programming, so maybe I am missing some big implementation difficulty. Cheating by looking at other players' moves is not compatible with running multiple AIs in parallel.
quote:
I mean, if you just want to say that the AI turns in SE take too long, then just say that. It's far more likely that's a result of poor optimization than of the fundamental game mechanic.
I recall reading Vic estimate before his summer holiday that he could reduce Shadow Empire's AI turn times by 50% if he focused on AI optimization. However, turn processing times on large maps (which I personally don't play in singleplayer because of this issue) in Shadow Empire can get into double digits of minutes.
EDIT: Clarification.
< Message edited by Soar_Slitherine -- 11/30/2021 10:48:42 PM >
|
Printable Version
RE: The State of Digital Wargames Nov. 2021 - 
New Messages
No New Messages
Hot Topic w/ New Messages
Hot Topic w/o New Messages
Locked w/ New Messages
Locked w/o New Messages
Post New Thread