Matrix Games Forums

Forums  Register  Login  Photo Gallery  Member List  Search  Calendars  FAQ 

My Profile  Inbox  Address Book  My Subscription  My Forums  Log Out

Historical AI?

 
View related threads: (in this forum | in all forums)

Logged in as: Guest
Users viewing this topic: none
  Printable Version
All Forums >> [New Releases from Matrix Games] >> World in Flames >> Historical AI? Page: [1]
Login
Message << Older Topic   Newer Topic >>
Historical AI? - 11/14/2006 11:37:45 PM   
hakon

 

Posts: 298
Joined: 4/15/2005
Status: offline
A challange one is faced when creating an AI for a historical game like wif, where a lot of the players will already be quite familiar with the actual history of the simulated events, is that the game mechanics of the game in many cases will favor non-historical play.

Examples:
- A German invasion of Norway does not happen very often in WiF.
- It is not very common for Germany to go both for a med campaign simultaniously with a 41 barbarossa.
- Japan often focuses on a land war in wif, attacking china and/or the USSR in much greater strength than historical
- The USSR rarely defend at the border (except when attempting to keep garrison)
- When playing a med strategy, many german players will usually take Malta and/or Gibraltar, even if it means attacking Spain.

Creating and AI that (usually) plays historically, will be most appealing to many players, and i suspect this will increase the sales to less-hardcore-grognards. On the other hand, it will take away from the challenge from veteran wif players.

Creating an AI that actually uses an algorith (aside for a statistical one) for deciding which main strategies to shoot for, i expect a game of this type to be based on some kind of decision tree, where each fork in the tree will have a hard-coded probability. For instance:

German 41 main strategy:
40% All out barbarossa 41, pulling all axis forces out of africa
20% All out med strategy, trying to seal the med, prepare for 42 barb or sitz
20% All out med strategy, attempt seelöwe, prepare sitz vs the USSR
20% Historical strategy of a medium effort middle east strat combined with a barbarossa 41

My suggestion, would be to, for most of these kinds of decisions, to designate one choice as "historical". It would then be possible (indeed quite easy) to make the AI favor the historical option by supressing alternative choices by a certain percentage. Imagine having an AI option, saying "historical AI", that supresses non-historical AI strategy selections by 75%. The above distribution would then transform to:

German 41 main strategy (historical AI):
10% All out barbarossa 41
5% All out med strategy, trying to seal the med, prepare for 42 barb or sitz
5% All out med strategy, attempt seelöwe, prepare sitz vs the USSR
80% Historical strategy of a medium effort middle east strat combined with a barbarossa 41

(The historicalness of the AI could even be made a slider, where non-historical strategies could be supressed by 0-100%)

So why do I want this option. Well, first of all I want to be able to turn it off :) But I realize that having the AI play more historically will be an advantage when facing critical reviews by non-wiffer-wargamers, as well as history-buff wargamers.

Second, I expect that an AI that plays more according to history will be more attractive if the player chooses to only control one of the smaller major powers, like china or france, or even Japan or the Commonwealth.

How much would this require in terms of work? Not much, I would imagine. The supression algorithm is trivial, so the problem is more or less reduced to creating an additional UI element at startup, as well as going through the main AI code, and adding a historical-flag to any historical decision. Of course, if the main strategy AI is not decision-tree based, it could be a lot harder.

Now what do you think?

(My applogies if this has already been discussed.)
Post #: 1
RE: Historical AI? - 11/15/2006 4:01:38 AM   
cockney

 

Posts: 83
Joined: 11/15/2006
From: London
Status: offline
As a player I would also like to see an AI that you could turn on 'Historical Mode' so that if for example you wanted to play a solitare game, you could judge for yourself wether your strat would have been better the the one implemented.

Eg playing Germany you would like the Russian AI to be 'surprised' at babarossa.

Or later if you havn't placed 2 panzer in Arnham the paras take it.

(in reply to hakon)
Post #: 2
RE: Historical AI? - 11/15/2006 4:59:14 AM   
Shannon V. OKeets

 

Posts: 22095
Joined: 5/19/2005
From: Honolulu, Hawaii
Status: offline
The player's control of the AI opponent is a low priority for me. If something can be done which involves little or no effort, then of course I will enable that capability. But on this feature I promise nothing.

_____________________________

Steve

Perfection is an elusive goal.

(in reply to cockney)
Post #: 3
RE: Historical AI? - 11/15/2006 6:31:56 AM   
Neilster


Posts: 2890
Joined: 10/27/2003
From: Hobart, Tasmania, Australia
Status: offline

quote:

ORIGINAL: hakon

A challange one is faced when creating an AI for a historical game like wif, where a lot of the players will already be quite familiar with the actual history of the simulated events, is that the game mechanics of the game in many cases will favor non-historical play.

Examples:
- A German invasion of Norway does not happen very often in WiF.
- It is not very common for Germany to go both for a med campaign simultaniously with a 41 barbarossa.
- Japan often focuses on a land war in wif, attacking china and/or the USSR in much greater strength than historical
- The USSR rarely defend at the border (except when attempting to keep garrison)
- When playing a med strategy, many german players will usually take Malta and/or Gibraltar, even if it means attacking Spain.

Creating and AI that (usually) plays historically, will be most appealing to many players, and i suspect this will increase the sales to less-hardcore-grognards. On the other hand, it will take away from the challenge from veteran wif players.

Creating an AI that actually uses an algorith (aside for a statistical one) for deciding which main strategies to shoot for, i expect a game of this type to be based on some kind of decision tree, where each fork in the tree will have a hard-coded probability. For instance:

German 41 main strategy:
40% All out barbarossa 41, pulling all axis forces out of africa
20% All out med strategy, trying to seal the med, prepare for 42 barb or sitz
20% All out med strategy, attempt seelöwe, prepare sitz vs the USSR
20% Historical strategy of a medium effort middle east strat combined with a barbarossa 41

My suggestion, would be to, for most of these kinds of decisions, to designate one choice as "historical". It would then be possible (indeed quite easy) to make the AI favor the historical option by supressing alternative choices by a certain percentage. Imagine having an AI option, saying "historical AI", that supresses non-historical AI strategy selections by 75%. The above distribution would then transform to:

German 41 main strategy (historical AI):
10% All out barbarossa 41
5% All out med strategy, trying to seal the med, prepare for 42 barb or sitz
5% All out med strategy, attempt seelöwe, prepare sitz vs the USSR
80% Historical strategy of a medium effort middle east strat combined with a barbarossa 41

(The historicalness of the AI could even be made a slider, where non-historical strategies could be supressed by 0-100%)

So why do I want this option. Well, first of all I want to be able to turn it off :) But I realize that having the AI play more historically will be an advantage when facing critical reviews by non-wiffer-wargamers, as well as history-buff wargamers.

Second, I expect that an AI that plays more according to history will be more attractive if the player chooses to only control one of the smaller major powers, like china or france, or even Japan or the Commonwealth.

How much would this require in terms of work? Not much, I would imagine. The supression algorithm is trivial, so the problem is more or less reduced to creating an additional UI element at startup, as well as going through the main AI code, and adding a historical-flag to any historical decision. Of course, if the main strategy AI is not decision-tree based, it could be a lot harder.

Now what do you think?

(My applogies if this has already been discussed.)



A potential problem is that even if the AI embarks on a historical course, the game will usually quickly diverge from history. Even if the game doesn't, the player will have too much knowledge about what the AI is going to try next.

Many WiF scenarios start after the end of some major historical campaign and that channels the game into more historical avenues, at least for a while.

Many moons ago, when I suggested a video replay feature (which is still on the cards AFAIK), I also proposed that a saved game could be hard-coded that represented as closely as possible the historical war (with the appropriate abstractions and with WiF granularity). I thought it would be very interesting to watch this MWiF WW2, with it's stunning early Axis victories and the gradual turning of the tide, the massive Allied production build-up and their powerful counter-blows.

I then suggested that if it were possible to "jump into" a game being replayed at the beginning of any turn, it would be possible to enter this historical saved game and play on from there.

I think enough MWiF games will follow a roughly historical path (at least in 39-41 and with the exception of Norway) to satisfy the gamers you mention, but those that don't will provide essential variety and replayability.

Cheers, Neilster







(in reply to hakon)
Post #: 4
RE: Historical AI? - 11/15/2006 2:45:31 PM   
trees

 

Posts: 175
Joined: 5/28/2006
Status: offline
Fascinating strategic algebra, you may wish to compare it to the German AI strategy thread of several months ago...the wild card in those calculations is whether the Russians are "up" or not; they can be the ones to select whether Barbarossa is in 1941 or 1942 if Germany likes a sunny Med vacation.

(in reply to Neilster)
Post #: 5
RE: Historical AI? - 11/16/2006 9:11:48 PM   
cockney

 

Posts: 83
Joined: 11/15/2006
From: London
Status: offline
on the German invasions of Denmark and Noway, why are they not popular with WiF players?

_____________________________

never piss off a sgt major

(in reply to trees)
Post #: 6
RE: Historical AI? - 11/16/2006 9:27:49 PM   
Shannon V. OKeets

 

Posts: 22095
Joined: 5/19/2005
From: Honolulu, Hawaii
Status: offline

quote:

ORIGINAL: cockney

on the German invasions of Denmark and Noway, why are they not popular with WiF players?

Denmark yes, Norway no. I guess the latter has to do with the diversion of effort from the attack on France.

---------------------------

I ascribe to the belief one should never annoy a comedian: they are both skilled and trained to publicly humiliate people they don't like.

_____________________________

Steve

Perfection is an elusive goal.

(in reply to cockney)
Post #: 7
RE: Historical AI? - 11/16/2006 9:49:35 PM   
Froonp


Posts: 7995
Joined: 10/21/2003
From: Marseilles, France
Status: offline
quote:

ORIGINAL: Shannon V. OKeets

quote:

ORIGINAL: cockney

on the German invasions of Denmark and Noway, why are they not popular with WiF players?

Denmark yes, Norway no. I guess the latter has to do with the diversion of effort from the attack on France.

The 2 TRS and the score of CP / Tankers given for free to the CW has A LOT to do with this too.
The Norwegian OOB also has to do with that (one (4-4) MTN, one (2-4) SKI).

(in reply to Shannon V. OKeets)
Post #: 8
RE: Historical AI? - 11/16/2006 10:09:38 PM   
cockney

 

Posts: 83
Joined: 11/15/2006
From: London
Status: offline
You have a point, but the benifits of Norway under German control would out way this. Of particular importance is when you start the battle of the atlantic, Norwegian airbases, like that at Stavingar, allowed German reconnaisance aircraft to operate far out over the North Atlantic, without having to fly over, or near to, Britain, and of course the ports are ideal for your U-Boats.
Germany was decisively victorious in the conflict, and yes it might damage your navy, (if the cw player tries to intervine)
And you wouldn't need to keep that many garrison troops as Norway will allign under the Quisling govenment.
A quick victory and the off to the low countries in time for tea.

_____________________________

never piss off a sgt major

(in reply to Froonp)
Post #: 9
RE: Historical AI? - 11/17/2006 10:26:01 AM   
wosung

 

Posts: 692
Joined: 7/18/2005
Status: offline
Perhaps two AI settings for the player to choose? To please all of us?

Wildcard
Historical

But then again it's a question of what is defined as "historical": Was a German full medditerraen strategy really an historical option? An US Japan first?

(in reply to cockney)
Post #: 10
Page:   [1]
All Forums >> [New Releases from Matrix Games] >> World in Flames >> Historical AI? Page: [1]
Jump to:





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
 Reply to Message
 Post New Poll
 Submit Vote
 Delete My Own Post
 Delete My Own Thread
 Rate Posts


Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI

3.016