If I was the AI...

[Greyshaft]

WIF has very simple victory conditions based around occupying specific hexes on the map. So what should the AI's considerations be when determining strategy? I'll stick with Germany for the moment and for the purpose of this thread I'll define 'strategic victory' as the securing of an additional VP hex.

* Do I have the initiative - this is based on whether I have a 'surplus' of attacking power beyond that which is necessary to counter perceived threats.
* Can I redeploy and use that quantified 'surplus' to achieve a strategic victory before the enemy can redeploy to defend my target?
* Am I reaching a decision point in the game which requires additional considerations - Is the US or Russia about to declare war? Are we near the end of the game so there is no point in launching an offensive which will take six months to complete?
* Should I periodically make a decision to 'write off' certain VP hexes as unattainable - say it is 1944 in a fairly historical game and I have no AMPH or other Navy. Should I still be thinking of laying down new AMPH and invading England or should that particular evaluation subroutine be closed down?
* How do I determine which is the greater threat in 1945? The Russians on the Oder or the Allies on the Rhine?

I'm not looking at playing styles - that's covered in another thread.

[Mziln]

Include:

Political points: Early in the game (Allies or Invasion?).

Economics: A fuel surplus (save or send some to Italy?).

Of course this would be influenced by Greyshaft's previous formula.

[Neilster]

I wrote this as part of a post about an AI last year...

"I'm one of the people who wants an AI and it seems to me that there may be two approaches. Both would combine probability calculations for the unknowable but estimateable bits.

The top-down method....

1. Analyse possible strategic goals given the current situation and choose/modify the best one/s
2. Work out the most efficient operational/political/production sub goals to accomplish 1
3. Work out the best tactical moves to accomplish 2
4. Go to 1

The current situation could involve initial setup.

The bottom-up method...

1. From the current situation, choose moves/attacks that are likely to maximize tactical success.
2. Make operational decisions based on the results of 1 (or perhaps the probabilities of results). Produce units/choose political options that are likely to be useful for these plans.
3. Modify strategic goals according to 2.
4. goto 1

I think humans do a bit of both when they play WiF. Perhaps combining them is possible."

Back to the present...
I'll give an example of what I meant. Let's take the top-down method. Say the AI is playing the Axis. Let's consider Germany's options for Sep/Oct '39

1. It looks at the initial setups of the Allies and decides to, for example, knock out Poland and Denmark first.

2. The best way to do this is to capture their capitals. (Is a coup of Denmark a realistic possibility? I can't remember. I haven't played board WiF for ages). The bulk of the Whermacht should be setup to invade Poland with some fast moving forces for Denmark and a defensive position in the West. This plan probably wouldn't affect production priorities too much as these should be quick campaigns. Later on, a decision to go for Barbarossa for example would tend to shift production toward land units.

3. Setup armour, mechanised and motorised units in hexes where they are most likely to breakthrough to Warsaw/Copenhagen. Place the highest quality Luftwaffe units where they can best support the offensives.

4. After the first turn (or perhaps impulse?), re-evaluate the situation, changing the grand strategy as the situation demands.

The whole thing might be better implemented recursively. That is, can I win form the current situation by doing something? If yes, do it. Else, do something which gets me closer to victory, breaking that down into sub-goals.

The bottom-up strategy works in reverse. Good tactics lead to operational successes that, in turn, open up strategic possibilities.

I feel the structured turn mechanics of WiF may actually help an AI implementation. It certainly guided my thoughts but I'm not an AI expert by any means. Ingebrit (sorry if I got your name wrong) has talked about 15-30 Intelligent Agents cooperating to achieve goals. My effort is more of a thought exercise for my own amusement. I'm sure Matrix has clever boffins who know all the latest stuff.[:'(]

[terje439]

I see one thing I would miss with Greyshafts idea....
The tactical victories that can be made strategical..
What I mean is it might sometimes be worthwhile to abandon a drive for a victoryobjective
if you have the chance to lets say cut large forces off supply (ok not a good sentence I know).

So the AI should also look on tactical situations , or atleast that is my humble oppinion

[Greyshaft]

ORIGINAL: terje439
What I mean is it might sometimes be worthwhile to abandon a drive for a victory objective if you have the chance to lets say cut large forces off supply (ok not a good sentence I know).

I agree with this. So while the major goal is to gain VP hexes, there are a lot of minor goals along the way.

* aim to make individual enemy countries surrender - continuously re-evaluate strategy but don't continuously change strategy.
* gain factory/oil hexes - especially if you can take them off an opponent who is highly dependant on them
* gain strategic hexes - Gibraltar, Suez, Truk etc.
* destroy enemy forces - strategically aim for critical units - if going for supply war against UK then attack transports and light ships
* identify critical hexes and allocate defences - eg. how many attack points can the enemy muster against any particular hex on my defensive line? If greater than 3:1 then add more defences at that location
* how do I calculate is the Kiev pocket worth more than Moscow? (terje's point)

Then after all of this is done... allocate your production to support your strategy. That's a whole 'nother conversation
*

[Mziln]

[:)] A simple outline for a "AI Strategy Evaluation Phase" [:)]

(1 and 2 allow the AI to priortize 3, 4, and 5)

(1) Determine strategic objectives (not forgetting to consider weather conditions) - Encirclement of large concentrations of units (supply lines have to be calculated anyway), Capitol cities, Gibraltar, Suez, Truk etc.

(2) Determine possable capture/isolation of factorys and resources.

(3) Political evaluation - Allign Allies and/or Declare war.

(4) Assign units for defence of crucial strategic objectives.

(5) Assign possable attacks on critical enemy units.
I.e. Attacking Commonweath transports and light ships in a supply war.

(6) Assign production to support the evaluated strategy.

[meyerg]

I would like the AI to be as good as an average player, but I think it will have major problems. There is too much going on in Wif for it to play competently.

The only hope is good players "canned" strategy be programmed into the AI. Here is an example for Germany.

1) Split army equally on setup between Poland/Belgium. Attack Belgium second impulse if weather is good. Push quickly into France.

2) Build lots of INF, Uboats, fighters, Paras, Navs and TACs

3) Burn offensive chit early in France if necessary.

4) Sortie Kreigsmarine late in turn, especially to sea zones planes can't fly, to sea areas with convoys and subs

5) Dominate the MED with LBA.

6) Groundstrike first impulse during long weather clear turns against France and Russia. Maximize overruns and attacks vs out of supply flipped units.

Heck, many players don't even follow these formulas and we'll here some newbies saying things like, WOW, British CVs can really be toasted by land based air in the Med!! The best benefit is humans get better so PBEM games get better (I am still dreaming of PBEM).

[IronManBeta]

PBEM figures in my nightmares, not my dreams! How would you go about implementing such a complicated series of events without totally changing the game? Just curious, Rob.

ORIGINAL: meyerg

Heck, many players don't even follow these formulas and we'll here some newbies saying things like, WOW, British CVs can really be toasted by land based air in the Med!! The best benefit is humans get better so PBEM games get better (I am still dreaming of PBEM).

[SeaMonkey]

How would you go about implementing such a complicated series of events without totally changing the game?

That's just it RC, you can't. The only way to get a decent PBEM experience is prescripting the unphasing players activities for action during the phasing players' turn based on a set of AI priorities or randomized priorities. Since all players will have to think ahead and the AI will randomize the reactions it'll have to be deemed fair even though some players will not be happy with the AI's response. Explain it as a breakdown in command and control based on faulty intelligence....sound familiar?

[meyerg]

I couldn't have said it better Sea Monkey. First, hopefully the AI will be good enough to do obvious things (target fighters before bombers in a big nav air battle and burn four surprise points to choose a surface when you have no nav air factors)
Hopefully we can set some preferences. For example, abort in this hex when on -4 table or worse, always intercept German fleets in North Sea, initiate sub combat on this table or better, if attack 1 yields a B then move the armor here so for attack 2 the units are flipped and out of supply. We may have the inconvenience of declaring strategic bombing, port strikes, and ground support at the same time.
Hopefully we can still play TCP/IP the old fashioned way, but if WiF is to have any chance against Gary Grigsby, it better be PBEM (with not too many emails per impulse).
Count me in for online play (hopefully with a chess clock system to keep the slow ones moving) when feasible. Just don't permanently keep the US from entering the war as Russia when I am US/China by declaring war on Rumania and Finland on the first impulse.
Hopefully the AI will show the newbees enough so that they quickly hang with the face to face veterans. By the way, the Russians are way powerful in the later versions, so beware the bear!!!

[meyerg]

RC, I didn't realize by the "PBEM nightmare" you were one of the guys that has to do the AI--my sympathies. You can't do PBEM without seriously changing the game, but you can't get universal acceptance of CWiF without PBEM. Unfortunately, if PBEM isn't built in up front, it won't make it in later (it would be too hard to add). So an abbreviated sequence of play has to be built in that simplifies the interactions if PBEM is enabled.
I think you may have to have non-phasing naval decisions including adding planes to the sea box and intercepting port strikes be preset. You have to let the nonphasing player fly defensive air, but that may be it.
Just think of the complication two transports running around in a combined with an ATR and a PARA create, and it is enough to give you AI nightmares. Well RC, I guess we can continue to dream about PBEM. If you need any help, better make me a beta tester!!!![:'(]

[IronManBeta]

1. We are committed to providing PBEM and so we will and in a way that makes as many players happy as possible. It seems pretty clear to me now having reviewed the forum threads that there is no real consensus on how this might be achieved in an elegant way. Drat. What I am thinking of right now is a screen for 'Standard Operating Proceedure' that each player can fill out that will delegate to the computer authority to act on his behalf with respect to certain of the issues enumerated by the playtesters. If none are delegated then there are going to be a lot of files traded back and forth. If you check all or nearly all of them then it will be much quicker and more along the lines that Greyshaft so brilliantly outlined for us a while back. Different players can have different SOPs and they can change them at any time. We might also have SOPs for each player for each theater - if I can figure out how to define "theater"! The idea is to give a meaningful level of control to the player where it is needed and wanted, and otherwise speed up the game as much as possible so that the players don't feel bogged down all the time. This mostly meets SeaMonkey's points I believe. I don't want to change the core WIF game mechanics but I recognize that they were not created with computers in mind and I want to exploit the capabilities of the computer as much as possible.

2. Beta testers will be selected from this forum (amongst other places) but that is not my particular job. Bart was working on this a while ago but before we can really consider this we have a whole lot of figuring out and documenting to do ourselves. Everyone who looked at CWIF had ideas for things they would like to see changed or added to it - not least myself. We are at the stage of looking at all possibilities and trying to sort out priorities. Then we have a lot of writing and preparation to do, and then we need to start making the changes. When that process is a few months in, then we will be getting serious about beta testers. We will want to beta from an very early stage with this one but it is premature still. You will see a call for beta testers posted here when the time comes. To improve your chances just keep making thoughtful, useful posts here. I read everything and frequently cut and paste ideas directly into my MWIF notes.

Ta, Rob.

[wfzimmerman]

PBEM is a must for me.

[gbed]

I will be more than available as a playtester for you, Robert. I was one of many for Chris.
Good luck with PBEM though! You may be able to come up with something, but my opinion is that head to head will be the BEST way to go i.e. network play. The constant interaction between opposing sides is what makes WIF the game it is.

[IronManBeta]

Stay tuned then- but nothing is going to happen very soon.

Cheers, Rob.

ORIGINAL: gbed

I will be more than available as a playtester for you, Robert. I was one of many for Chris.
Good luck with PBEM though! You may be able to come up with something, but my opinion is that head to head will be the BEST way to go i.e. network play. The constant interaction between opposing sides is what makes WIF the game it is.

[gbed]

Robert, are you attending WifCON this month? Might be a great opportunity for ideas and feedback.

[YohanTM2]

ORIGINAL: gbed

Robert, are you attending WifCON this month? Might be a great opportunity for ideas and feedback.

I sure hope if you aren't Rob some of the guys from the development team are.

[IronManBeta]

WifCON would be a dream come true and at one point I was seriously hoping to do it. The day job interfered though - can you imagine that they think I am indispensible and they can't do without me for even a single day? The wife is not much less possessive either but that, of course, is a good thing.

Just going to Origins in June and my parents 50th anniversary in August leaves me overdrawn in the 'days away' department. How did day jobs get so all-consuming anyway?

I have the website bookmarked and will follow along if they post anything though. David is really excited about this project btw. I would love to meet the best of the best, but it looks like it will have to wait a bit yet.

Later, Rob.

[gbed]

I'm not sure the website will be updated much. Are you still on the WIFdiscussion list? Russ Craft, and others, usually post daily updates there. Too bad no one could attend for at least a day of observation! Would have been HIGHLY beneficial!
Peace!