|
Von Rom -> (7/19/2003 5:07:58 AM)
|
[QUOTE]Originally posted by Mr.Frag
[B]No, it adds even MORE problems as now you need to pass through the evaluation as now you need to add a layer of evaluation to test whether each and every possible option interferes with the specific event and/or negates the event.
Think of the logic for a sec ... using Midway as an example.
Japan mounts an attack that is perceived by the event engine as a Midway event. The AI steals all available CV's from whatever missions they may have already been engaged on and reroutes them to Midway to try and save the day.
The Japan player, being a sneaky evil person like me turns around knowing that he triggered a Midway event and actually goes ahead with his real plan (as Midway was just a trick to sucker the AI into withdrawing the CV's from Noumea) which has a huge invasion fleet hiding nearby.
By abusing the event engine, I gain the ability to trick the AI into doing things that are completely illogical. There would need to be even more levels of evaluation now added to validate that the event is a real event, not a trick event, and it would also have to calculate the impact of reacting to the event (assuming it was real) vs the impact to changing it's existing plans that may already be well underway. You would literally be forcing the AI to play in a very non-strategic manner, always reacting at a tactical level to ever changing situations.
Now you have to add another layer of evaluation to validate that the costs of each operation that the AI mounts are worthy of the investment in time and materials no matter what the rest of the theater happens to be doing at the time. You also have to weigh these plans properly so the AI sees altering the plans after the fact as costly and can draw the logically conclusion of messing with operational plans after the plan is running is ill advised.
Now you run into another mess. Lets say the Midway role is reversed and the AI is the one triggering it. Do I, being the human want my forces ripped from my control in the middle of what I was doing to react to the event trigger? Certainly not!
I would suspect he has had to hide all sorts of hidden weights all over the place to get the AI to actually do anything other then sit back and defend. This is what causes UV to work well in short scenarios and poorly in long scenarios. These weights (what drives the AI as such) need to be constantly adjusted as things change, but not adjusted so much to confuse the AI. This would probably be fine if we played 2 month turns, but being as we are not locked into a plan and can constantly mess with it, the computer AI is at a major disadvantage here.
If Gary can manage to get AI into this game that can actually play the whole campaign from start to finish in a sensible and logically manner with actual objectives that make strategic sense as part of the overall war plan, I'll be completely amazed as it will probably be the most complex AI ever done on a PC based game. I suspect the best he'll be able to manage is sections of the area up to a year in duration as PC's just don't have the power yet to be able to operate on this level of AI.
The sheer hareware requirements to have the AI even get remotely close to being able to look 40 turns into the game in semi real time is staggering. To be able to actually look that deep and choose to mount an operation that makes sense would require Gary to pretty much compose millions of possible battle plans and have the computer AI evaluate each round to see if any of them happen to apply to the current situation and merit use. Without this kind of predicate look ahead, the best the AI will really be able to do is sail along through some pre-coded master plan, with minor deviations in order of execution mixed in to keep us guessing for the first dozen or so games.
If you are expecting to play the AI beyond that, I doubt you'll be happy. WitP is at heart going to be a two player game, with the computer managing the dice rolls for us. [/B][/QUOTE]
Hi :)
I think there is some misunderstanding as to how events work.
Events normally are written to help the AI. So if an event fires when the AI plays Japan, the same event WILL NOT FIRE (or can be made not to fire), when the human plays Japan.
It is actually quite simple and elegant with the right event tools. Currently, modders have created over 300 NEW events for HoI. The reason why Paradox has not incorporated them into the default game, is a question best left to them. But these events help SOLVE most of the game's problems.
Having an event editor relieves the game's programmers of a lot of work.
Most of these events can be added after the game has been released. When thousands of people are playing the game against the AI, many similar situations will arise that can be fixed with an event, that will only fire when it applies to the AI.
WiTP will be a fairly open-ended grand strategy game. This is the perfect game in which events work the best, especially since there are only two protagonists.
The fact that any game (especially a wargame) is being made today without scenario or event editors, does a disservce both to the game and to the gamers themselves.
Here is an example:
In UV when the player is playing Japan vs the AI, and he invades Australia, it is basically game over - as the game now stands. However, if UV had had an event editor, I, or someone else, could then create an event such that IF any city in Australia is occupied by the Japanese, THEN Australia will mobilize the "Home Guard", which would mean adding additional, but less experienced, troops to other Australian cities. This event could also trigger additional B-17 bombers being sent to Brisbane, as well as other ships, destined for other theatres, being sent to help stem the Japanese advance.
There is a suggestion that the AI can be tricked through events. But the event can be written in such a way as to over-come these situations.
A good event editor can make a good game - great.
IMO, to suggest that a game can only be played by two humans, does an injustice to what can be done with the AI through programming, events, placement and some imagination.
Cheers!
|
|
|
|
