RE: AI Extra Cost When It's Released?

Matrix Games Forums

Forums Register Login Photo Gallery Member List Search Calendars FAQ

My Profile Inbox Address Book My Subscription My Forums Log Out

RE: AI Extra Cost When It's Released?

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 >> RE: AI Extra Cost When It's Released?

Page: << < prev 1 [2]

Message

<< Older Topic Newer Topic >>

RE: AI Extra Cost When It's Released? - 11/23/2013 9:40:55 PM

Shannon V. OKeets

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

quote:

ORIGINAL: celebrindal

quote:

ORIGINAL: Shannon V. OKeets

quote:

ORIGINAL: Dabrion

I would freeze the AI project and concider put it into the public domain. That will decrease development time drastically, for MWiF and for a computer opponent. The gained dev-time could be spent on improving the boardgame sandbox and UI/UX!

If you were going to build a 30 story condominium, would you want hundreds of untrained people to help you? For the AIO, some helping hands would be nice, but involving everyone who has a stray idea to write a script would be foolish. Sorry, but that's a solid truth in my opinion - see the next paragraph.

There must be an underlying structure and organized implementation layered onto that structure. That means the group has to work as a team, not a mass of individuals with each working on what he thinks is a good idea. All of that implies each person has an understanding of the structure and his role in the overall implementation. There will be a need to do some learning before writing scripts.

On the other hand, I am always willing to read other players ideas about how to play WIF well. The existing AI Opponent threads, one per major power plus dozens for setting up the minor countries, attest to that. Writing in those threads does not require any particular skill other than knowing how to play the game. You don't even have to be a genius at playing WIF, since many great ideas have strange points of origin.

I'd be willing to PM a team of folks working on it if you need.

As for AI, i'd prefer to see it going faster than PBEM.. i'm just not satisfied that PBEM lends its self well to something that has so much interaction between folks in various phases...

When the time comes, I'll ask for people who want to help with the AIO and organize the group according to each individual's interests and skills. It shouldn't be difficult to find a role for anyone who is interested. But there will be an upper limit to how many cooks can gather around the broth.

_____________________________

Steve

Perfection is an elusive goal.

(in reply to celebrindal)

Post #: 31

RE: AI Extra Cost When It's Released? - 11/24/2013 12:53:02 AM

Neilster

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

quote:

ORIGINAL: 76mm

No disrespect intended toward Steve at all, but I find it difficult to believe that this will be the first commercial wargame ever to have a competent AI, especially when you consider the implications of the various combinations of optional rules which may/may not be activated. Maybe possible OK for shorter scenarios but hardly for the campaigns

Ever played TOAW 3 or the Command Ops series?

Cheers, Neilster

(in reply to 76mm)

Post #: 32

RE: AI Extra Cost When It's Released? - 11/24/2013 12:59:49 AM

Neilster

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

quote:

ORIGINAL: pzgndr

quote:

ORIGINAL: Neilster
Allowing AIs to "cheat" is a bad idea. It warps the game. Better to use optional rules for play balance.

Ideally, if the AI is really good, then the optional rules should provide sufficient flexibility. But whatever rules are selected, they apply for all players. Since MWiF could have up to 6 players, there could be multiplayer games of less than 6 where some players could be set to AI or human players get replaced by AI. In these cases, depending on how the AI turns out, some specific bonuses and/or difficulty levels for individual AI players may be needed or desired. I don't necessarily see cheats as a bad idea, they're just compensation for AI weaknesses. And for some newbie human players in multiplayer games, this might also be a nice feature to have.

As far as I know, the AI will either play the Axis or the Allies, not individual powers. Hence optionals that clearly help one or the other side will be able to be chosen.

Very early on on I suggested giving the AI an advantage but I was completely dissuaded. I can't remember the details but I think it involved the humans changing the way they would normally play because they knew the AI had an unfair advantage. That's why I suggested it warps the game. From what I've seen, I'm confident the AI will be quite strong. Expert players will load up the AI with advantageous optionals or play each other.

Cheers, Neilster

(in reply to pzgndr)

Post #: 33

RE: AI Extra Cost When It's Released? - 11/24/2013 3:10:41 AM

76mm

Posts: 4688
Joined: 5/2/2004
From: Washington, DC
Status: offline

quote:

ORIGINAL: Neilster

Ever played TOAW 3 or the Command Ops series?

Dabbled TOAW 3 years ago, never played Command Ops.

I should also say that I have seen competent AIs for shortish scenarios, but for all the games that I buy, I typically play the longest scenarios available, and inevitably the AI breaks down sooner or later as the scripting runs out, the game takes an unforeseen turn, etc. To the extent that an AI can stay in a long scenario at all, it is generally because of "AI cheating".

(in reply to Neilster)

Post #: 34

RE: AI Extra Cost When It's Released? - 11/24/2013 10:36:04 AM

Rasputitsa

Posts: 2903
Joined: 6/30/2001
From: Bedfordshire UK
Status: offline

All of wargaming is a cheat, not only for the AI, but also for human players, no real commander ever had the amount of information that human players have access to in these games, movement allowances set for weather effects that haven't happened yet, instant obedience from subordinate units, etc..

A realistic representation for the Soviet command in June 1941 is that you wouldn't see the correct positions for most of the axis units and neither would you see where a good number of your own units were. A large proportion of your moment orders would not be carried out, but no game reproduces this fully, because nobody wants to play that way, games have to be playable and enjoyable, so we cheat history to get a better game.

The AI is unlikely to be good at imaginative blitzkrieg actions, but those were rare, most of WW2 was an attritional slugging match with logistics and planning more important (WIF territory).

The AI doesn't have to be a genius to simulate Allied performance in the early years, or the Axis digging in to hold their gains for the later years. By choosing the right scenarios and setting suitable options, it is easily possible to get a challenging game out of the AI in many game titles and hopefully WIF with offer the same.

The AI will probably not be able to do everything a human opponent could achieve, but it doesn't have to, with a little imagination a good game can still be had.

< Message edited by Rasputitsa -- 11/24/2013 2:16:08 PM >

_____________________________

"In politics stupidity is not a handicap" - Napoleon

�A people which is able to say everything becomes able to do everything� - Napoleon

�Among those who dislike oppression are many who like to oppress" - Napoleon

(in reply to 76mm)

Post #: 35

RE: AI Extra Cost When It's Released? - 11/24/2013 11:03:38 AM

Neilster

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

quote:

ORIGINAL: 76mm

quote:

ORIGINAL: Neilster

Ever played TOAW 3 or the Command Ops series?

Dabbled TOAW 3 years ago, never played Command Ops.

I should also say that I have seen competent AIs for shortish scenarios, but for all the games that I buy, I typically play the longest scenarios available, and inevitably the AI breaks down sooner or later as the scripting runs out, the game takes an unforeseen turn, etc. To the extent that an AI can stay in a long scenario at all, it is generally because of "AI cheating".

The consensus is that the AI in TOAW 3 is good. The AI in the Command Ops series is the best in wargaming. I really suggest you have a look at that series. It's really excellent and you might be surprised.

Cheers, Neilster

(in reply to 76mm)

Post #: 36

RE: AI Extra Cost When It's Released? - 11/24/2013 11:11:19 AM

Neilster

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

Chess AIs often don't do anything spectacular but are hard to beat because they don't make silly mistakes and they don't miss oportunities. This is an aspect of the MWiF AI that will be important.

Their human opponent may be much better at strategic planning and lots of other things but may overlook that a paradrop on some port is possible, has a high probability of success and can lead to more units then being shipped in with dire consequences etc etc. The very complication of the game and wealth of possible moves will assist the AI in this respect.

Cheers, Neilster

(in reply to Neilster)

Post #: 37

RE: AI Extra Cost When It's Released? - 11/24/2013 1:48:44 PM

pzgndr

Posts: 3170
Joined: 3/18/2004
From: Maryland
Status: offline

quote:

ORIGINAL: Neilster
As far as I know, the AI will either play the Axis or the Allies, not individual powers. Hence optionals that clearly help one or the other side will be able to be chosen.

I noticed this reading through the manuals the other day. Ditto for PBEM too apparently. So multiplayer will only work with NetPlay, when all players must be logged in at the same time. So not much flexibility switching between NetPlay, pbem, and swapping players with AI? That may need reconsideration in the future.

But since plan is for AI to play an entire side, the optional rules should provide sufficient play balance. That should work for most players, but again the multiplayer community may want some more flexibility down the road.

(in reply to Neilster)

Post #: 38

RE: AI Extra Cost When It's Released? - 11/24/2013 2:02:02 PM

76mm

Posts: 4688
Joined: 5/2/2004
From: Washington, DC
Status: offline

quote:

ORIGINAL: Neilster
The consensus is that the AI in TOAW 3 is good. The AI in the Command Ops series is the best in wargaming. I really suggest you have a look at that series. It's really excellent and you might be surprised.

Excellent, thanks. I've been waiting for Command Ops to come out with a East Front game and then will check it out. I'm not big on TAOW's interface and am not sure I'll pick that one up again.

(in reply to Neilster)

Post #: 39

RE: AI Extra Cost When It's Released? - 11/24/2013 2:06:13 PM

brian brian

Posts: 3191
Joined: 11/16/2005
Status: offline

There are actually only a very few optional rules that have a dramatic impact on balance. The majority of them add detail ("chrome"), with slight subsidiary effects on game results. Most of those are because if a rule favors the attacker, this can favor the Axis, which then has an easier time expanding. And the Allies get that attacking benefit in the 2nd half of the game.

One 'balancing' rule already in the game could be considered an AI cheat - the Food in Flames optional. It generally gives the CW 3 extra production points per turn as long at they maintain a convoy link to the other CW Home Countries. There isn't really much on the Axis side that is similar; perhaps playing without an Oil rule would help the Axis as much, though probably more. But ignoring oil in a game about WWII, even with an AI opponent, might not appeal to people interested in history, rather than just some complicated gaming.

The designer of the game has suggested from the beginning that the game can be balanced by adding or subtracting resources from the USA. Not sure if this would be considered an "AI Cheat", but it would be a very simple way to do it for either side in either direction, without changing any of the decision logic the AI uses at all.

(in reply to pzgndr)

Post #: 40

RE: AI Extra Cost When It's Released? - 11/24/2013 3:15:33 PM

Centuur

Posts: 8802
Joined: 6/3/2011
From: Hoorn (NED).
Status: offline

I don't agree. The Oil rule balances the game towards the Allies (especially when playing with the destruction of factories/oil rule also)... The No ZOC on surprise impulse rule balances the game towards the Axis. Both of these have quite a large (and I think even a dramatic) impact on balance...

_____________________________

Peter

(in reply to brian brian)

Post #: 41

RE: AI Extra Cost When It's Released? - 11/24/2013 6:07:14 PM

kkinoss

Posts: 2
Joined: 9/21/2009
Status: offline

Resigned to it not appearing, but not being able to play as a particular nation while all the other nations were AI is going to be a shame.

Im more than prepared to accept that because of the cooperation between the US and CW, for instance, that it would be a mine field to code and involve potentially a lot of unsatisfactory compromises...... but i still say its a shame

(in reply to Centuur)

Post #: 42

RE: AI Extra Cost When It's Released? - 11/24/2013 9:20:22 PM

Numdydar

Posts: 3211
Joined: 2/13/2004
Status: offline

The current plan is for all the optional rules to be included at some point. So not sure why you are resgining to something not occuring what it has been stated it will happen.

(in reply to kkinoss)

Post #: 43

RE: AI Extra Cost When It's Released? - 11/24/2013 9:37:22 PM

paulderynck

Posts: 8201
Joined: 3/24/2007
From: Canada
Status: offline

I believe he was referring to having the AI play specific major powers while humans play others.

_____________________________

Paul

(in reply to Numdydar)

Post #: 44

RE: AI Extra Cost When It's Released? - 11/25/2013 1:18:10 AM

Numdydar

Posts: 3211
Joined: 2/13/2004
Status: offline

Ah. More forum confusion I see.

I doubt that will happen. At least not at the start. Since the entire is based on just two sides. But who knows, maybe this is easier to code than we think

(in reply to paulderynck)

Post #: 45

RE: AI Extra Cost When It's Released? - 11/25/2013 11:51:38 AM

pzgndr

Posts: 3170
Joined: 3/18/2004
From: Maryland
Status: offline

quote:

ORIGINAL: brian brian
The designer of the game has suggested from the beginning that the game can be balanced by adding or subtracting resources from the USA.

It may also worth considering simple resource adjustments for all major powers; e.g., Allied AI for France might need a boost in 1939/40 and USA adjustments wouldn't help with that. And again, for multiplayer games, it would also be nice for play balance to have something adjustable for individual players. It's not just an AI issue.

(in reply to brian brian)

Post #: 46

RE: AI Extra Cost When It's Released? - 11/25/2013 4:12:31 PM

Shannon V. OKeets

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

quote:

ORIGINAL: pzgndr

quote:

ORIGINAL: brian brian
The designer of the game has suggested from the beginning that the game can be balanced by adding or subtracting resources from the USA.

It may also worth considering simple resource adjustments for all major powers; e.g., Allied AI for France might need a boost in 1939/40 and USA adjustments wouldn't help with that. And again, for multiplayer games, it would also be nice for play balance to have something adjustable for individual players. It's not just an AI issue.

New optional rules are unlikely to be programmed. Play balance can be handled in a variety of ways.

_____________________________

Steve

Perfection is an elusive goal.

(in reply to pzgndr)

Post #: 47

RE: AI Extra Cost When It's Released? - 11/26/2013 12:08:41 PM

pzgndr

Posts: 3170
Joined: 3/18/2004
From: Maryland
Status: offline

quote:

ORIGINAL: Shannon V. OKeets
New optional rules are unlikely to be programmed. Play balance can be handled in a variety of ways.

I think we've got that, no new optional rules. But it's fair to discuss how players can adjust play balance, for either inexperienced human players or for possible AI weaknesses (TBD).

Toggling some optional rules on/off may be sufficient, but maybe not so much.

How about modding the scenarios for play balance, can resources be added/subtracted either directly or by aligning/de-aligning some minor countries, to provide either an Axis or Allied advantage? Can USA entry markers be added/subtracted? Not trying to beat a dead horse here, just exploring what options may be possible.

(in reply to Shannon V. OKeets)

Post #: 48

RE: AI Extra Cost When It's Released? - 11/26/2013 6:25:57 PM

Shannon V. OKeets

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

quote:

ORIGINAL: pzgndr

quote:

ORIGINAL: Shannon V. OKeets
New optional rules are unlikely to be programmed. Play balance can be handled in a variety of ways.

I think we've got that, no new optional rules. But it's fair to discuss how players can adjust play balance, for either inexperienced human players or for possible AI weaknesses (TBD).

Toggling some optional rules on/off may be sufficient, but maybe not so much.

How about modding the scenarios for play balance, can resources be added/subtracted either directly or by aligning/de-aligning some minor countries, to provide either an Axis or Allied advantage? Can USA entry markers be added/subtracted? Not trying to beat a dead horse here, just exploring what options may be possible.

The data files are described in the appendices. If there isn't a data file for what you have in mind, then the data is embedded in the program. I've put a lot out there as CSV files which can be edited using a spreadsheet, a text file editor, or a database. But my task was not to create a WIF construction kit, so a lot of the finer details are internal to the program.

_____________________________

Steve

Perfection is an elusive goal.

(in reply to pzgndr)

Post #: 49

RE: AI Extra Cost When It's Released? - 11/26/2013 7:51:50 PM

abulbulian

Posts: 1047
Joined: 3/31/2005
Status: offline

For the time being I think we should all just be grateful that such a massive and impressive game has finally been released. Very pleased with the amount of work and effort put into not only just the game mechanics, but all the fantastic documentation too. As a developer myself, I understand and appreciate how much of an undertaking MWiF project must have been. The AI is definitely something I look forward to as well. In order to release something in say mid-late 2014 (just speculating) it would be best to

- limit the AI to just Axis or Allied powers
- only for a set of standard variants (easier to code AI for) - can expand this as AI code grows.

Let's not start worrying about play balancing for the AI until we have some grounds to do so. I for one would like to see AI sooner than later.

< Message edited by abulbulian -- 11/26/2013 9:10:28 PM >

_____________________________

- Beta Tester WitE and ATG
- Alpha/Beta Tester WitW and WitE2

"Invincibility lies in the defence; the possibility of victory in the attack." - Sun Tzu

(in reply to Shannon V. OKeets)

Post #: 50

RE: AI Extra Cost When It's Released? - 11/27/2013 7:18:03 AM

RSGodfrey

Posts: 147
Joined: 6/21/2004
Status: offline

I have to say that I am intrigued by the conceptual problems of simulating strategic and tactical decision-making for the AI under
uncertainty. The 'intelligence' available for the AI would have to be limited by the fog of war. Off the top of my head the AI
would have to consider optimal plans based upon its chosen maximum likelihood assumptions. The verisimilitude of the information
available to the AI would have to be variable depending upon time decay, communications (maybe a function of the robustness of supply lines?)
The programme would have to behave like an honest broker; knowing the truth of every unit's disposition and throwing a dice to determine the
outcome of each players moves.

Sounds more complex and interesting than programming chess!

Richard

(in reply to Neilster)

Post #: 51

RE: AI Extra Cost When It's Released? - 11/27/2013 7:48:13 AM

Dabrion

Posts: 733
Joined: 11/5/2013
From: Northpole
Status: offline

You dont have supply in Chess ;) And it is probably not possible to traverse the event space for even several impulses if you want to be feasible on a normal laptop.

You also need a (decision making) model for you agent to begin with in order to have ML or MAP criterions to base your decisions on. Question would be what a good scope for an agent would be. An HQ? Then you could solve a chess-like/-scoped problem for each HQ.

(in reply to RSGodfrey)

Post #: 52

RE: AI Extra Cost When It's Released? - 11/27/2013 2:13:22 PM

Neilster

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

A chess-like minimax approach won't work for MWiF. I asked Steve about this many years ago. Too many terrain and unit types and too many possible moves. The AI thread from yonks ago would have more details of the Approach Steve is planning to take.

Cheers, Neilster

< Message edited by Neilster -- 11/27/2013 3:14:18 PM >

(in reply to Dabrion)

Post #: 53

RE: AI Extra Cost When It's Released? - 11/27/2013 3:55:02 PM

Shannon V. OKeets

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

quote:

ORIGINAL: Neilster

A chess-like minimax approach won't work for MWiF. I asked Steve about this many years ago. Too many terrain and unit types and too many possible moves. The AI thread from yonks ago would have more details of the Approach Steve is planning to take.

Cheers, Neilster

The simple counter example for why minimax doesn't work is to consider your decisions making when going to the store. Too many variables explode the search space into infeasible. And for WIF, just consider how many possibilities there are for moving land units in one phase.

_____________________________

Steve

Perfection is an elusive goal.

(in reply to Neilster)

Post #: 54

RE: AI Extra Cost When It's Released? - 11/27/2013 4:07:28 PM

Solaristics

Posts: 195
Joined: 2/20/2002
From: UK
Status: offline

quote:

ORIGINAL: Shannon V. OKeets
Too many variables explode the search space into infeasible. And for WIF, just consider how many possibilities there are for moving land units in one phase.

Roll on quantum computing!

(in reply to Shannon V. OKeets)

Post #: 55