Strategy 101 resources wanted

Matrix Games Forums

Forums Register Login Photo Gallery Member List Search Calendars FAQ

My Profile Inbox Address Book My Subscription My Forums Log Out

Strategy 101 resources wanted

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

Logged in as: Guest

Users viewing this topic: none

Printable Version

All Forums >> [Current Games From Matrix.] >> [World War II] >> Norm Koger's The Operational Art Of War III >> Strategy 101 resources wanted

Page: [1] 2 3 next > >>

Message

<< Older Topic Newer Topic >>

Strategy 101 resources wanted - 1/31/2014 6:37:19 PM

ralphtricky

Posts: 6685
Joined: 7/27/2003
From: Colorado Springs
Status: offline

I'd like to try to improve Elmer (the AI) at a strategic level, where he has a complete lobotomy right now. He's controlled strictly by the objectives. I don't mind that idea, but I'd like to figure out some way to change the objectives so that he plays better.

The strategic level is about moving groups of formations, not individual units. Do I attack left, right, middle? Do I defend strongly here or draw back to where it's more defensible? Unlike Chess where the computer can look many moves ahead, TOAW has to try to figure out what the position looks like and what a coherent strategy is.

Does anything know of any resources (books, blogs, etc.) that talk about that analysis at the level of TOAW? There is plenty at the squad level, but I'm not seeing anything at a bigger scale.

I've read everything I can find on AI, I've even got a new book arriving today, but most of the articles are about FPS games right now. I can do the map analysis and figure out the chokepoints, I think I understand enough to start on a strategic AI, but deciding when to try to flank, how to respond to an attempted flank attack, and when to thrust up the center are something that is a bit vague still. Even something as simple as when to mop up the stragglers and when to let them go in favor of advancing as far as possible is fuzzy.

Thanks,
Ralph

Post #: 1

RE: Strategy 101 resources wanted - 1/31/2014 7:46:43 PM

Zovs

Posts: 6668
Joined: 2/23/2009
From: United States
Status: offline

When I worked with GG on WITE, I was playing against his AI routines which were scripted for the Germans for 41, 42 and 43 for the first 3-5 turns. After that it was almost impossible to script any more then that.

For a game like TOAW its darn near impossible because there are so many options, so many scenarios and so many ways to solve problems and that is the crux of AI and game AI in particular.

If you have a game such as WITE it's a bit more controlled and much easier to write AI routines or scripting language for it. Tactically games are easier cause there is so many finite things to take into consideration.

Sorry to not be much more help then that, but I have read and worked on AI stuff for years and it's no easy task to teach the computer how to think like a war gamer.

_____________________________

Beta Tester for:
Flashpoint Campaigns: Sudden Storm
War in the East 1 & 2
WarPlan & WarPlan Pacific
Valor & Victory
DG CWIE 2
SPWW2 & SPMBT scenario creator

(in reply to ralphtricky)

Post #: 2

RE: Strategy 101 resources wanted - 1/31/2014 8:18:48 PM

ralphtricky

Posts: 6685
Joined: 7/27/2003
From: Colorado Springs
Status: offline

Thanks,
It's definitely not easy, but games like Civ 5 can do a reasonable job, so I know I can do better.

One bottleneck I've got is that I'm trying to understand the strategies well enough to explain how to apply them and to choose between them to the computer, I don't remember ever seeing anything about how to do that for anything above small unit tactics.

(in reply to Zovs)

Post #: 3

RE: Strategy 101 resources wanted - 1/31/2014 8:52:06 PM

Telumar

Posts: 2236
Joined: 1/3/2006
From: niflheim
Status: offline

Re scripting as mentioned by dlazov66: Wouldn't it be possible to implement 'designer scripting' for the first turn? The scenario designer could 'play' the AI's first moves and order combats which could be stored in the scenario. Maybe there could be several scripted openings so that one could be chosen at random (also set by the designer).

Another idea that i have is to expand on the .ai file and strategic bias system in doing something like "AI doctrine" that can be switched during the scenario (by events). The AI doctrine could contain settings like agressiveness, air settings (combat support vs interdiction), emphasize on destroying enemy forces or capturing objectives. Starve cut off enemies or destroy asap. Stubborn defense or elastic defense. Acceptable Loss ratios. etc.

_____________________________

http://thetoawbeachhead.wordpress.com

(in reply to ralphtricky)

Post #: 4

RE: Strategy 101 resources wanted - 1/31/2014 9:03:16 PM

Zovs

Posts: 6668
Joined: 2/23/2009
From: United States
Status: offline

Well if we were to talk in human terms and lets just say for the sake of argument we are talking about a plan old board war game then...

First as a player you need to know the rules, the rules are the mechanics of how to play it. Then sometimes depending on the game, you have tactics that 'emerge' as result of those rules that in turn dictate a tactic (or gamit if you will of that game).

Next you have strategies, operations and tactics. The tactics are what you use to implement your strategies and the operational aspects are the "inbetween" stuff that may or may not need to be thought of or implemented.

So lets say a certain strategy in a certain game is to invade France. When do you do it? When is the right time? When do you have enough resources to do so? How much resources do you need? When or where is the best point in time to decide this, or when do you take a risk? How much risk? From a human perspective, these can take millisecs to a few hours to evolve, develop and plan for. It all depends on a lot of factors, how do you teach the computer these things? How can you program the AI to "recognize" a weak point? How do you teach a computer that the supposedly weak point is really a trap? The human mind can see these things but translating them for the computer is not an easy task.

Now for the operational, you have HQs that must supply artillery and other supplies to your divisions, each division must in turn dole these out to the regiments and the battalions must do the fighting. Since each hex is 2.5 km, each battalion can travel many hexes, but how far is the supply line and communication line to the HQ and how brittle are my units, how brittle are the enemy units, can I surround him to eliminate him? What is the over all plan? What is the main strategic goal?

See it gets complicated really quickly.

Really not trying to piss in your beer, there is just so much to factor in, the best you can do (again I base this with working with GG on this) is to have a human do detailed movements and have them scripted. Even the AI for chess is based on a chess game which has only so many squares and so many pieces and so many possible moves. In a war game where there are possible thousand of hexes, units, conditions, tactics, operations and strategies its almost insane to think of all the AI options.

In my opinion, it would take many programmers, and many in a team to detail it all out and then to process all that info you'll need lots of CPU horsepower.

Just my random thoughts...

_____________________________

Beta Tester for:
Flashpoint Campaigns: Sudden Storm
War in the East 1 & 2
WarPlan & WarPlan Pacific
Valor & Victory
DG CWIE 2
SPWW2 & SPMBT scenario creator

(in reply to ralphtricky)

Post #: 5

RE: Strategy 101 resources wanted - 1/31/2014 9:11:27 PM

Curtis Lemay

Posts: 12969
Joined: 9/17/2004
From: Houston, TX
Status: offline

quote:

ORIGINAL: ralphtrick

Even something as simple as when to mop up the stragglers and when to let them go in favor of advancing as far as possible is fuzzy.

So have you tried....fuzzy logic?

(in reply to ralphtricky)

Post #: 6

RE: Strategy 101 resources wanted - 2/1/2014 11:54:39 AM

USXpat

Posts: 388
Joined: 8/26/2010
Status: offline

You've already helped Elmer develop into a respectable opponent in several aspects. My inclination is that it is up to the designer to implement and "design for" strategy; and that it is Elmer's job to "implement it based upon an essentially mathematical understanding of the conditions." 3 articles loosely on this topic can be found at http://www.wargamecollege.net/?s=Elmer (some of this is dated).

Core factors are for Elmer to understand his formations relative to

A) space (frontage) - how many hexes it is expected to hold.

B) unit strength/density (friend & foe)

C) time (# of turns) - how far it is from its last objective and how many turns it has to get there.

D) sustainability (losses/replacements) -

B-D would serve to modify defensive/offensive posture and aggressiveness.

As an aside, one option that would be useful is an "Assist" order that could be further associated with a specific formation (even if it needs to be defined via an event) or as a global variable that would let Elmer define which formation/s need the most help.

The designer can create enough formations to cover total frontage - and stagger objectives so that their "areas of responsibility" overlap. Whether Elmer applies or is able to apply to this is the main issue (if a formation is essentially destroyed, that's going to create a gap). Elmer does reasonably well provided it maintains a continuous line.

As the player's objective is to break that line, Elmer needs additional ability to recognize it when it happens or even respond to the possibility of it happening.

This could be where the HQ level of formations come into play, but is complicated by the different scenario scales (5 km vs 15 km or 25 km per hex). This might be resolved if the designer can define the formation's expected frontage - or ability to override any default.

As far as I can determine, Elmer treats units and formations tagged as a regiment, division, corps or army as the same. I'm not a programmer but suspect it would be difficult, but maybe not -- is for higher level HQs/formations to consider its objectives based upon a radius around an objective.

In example - most formations (say a corps) would be concerned with specific objective hexes (or objective with a radius of 1); an army might then be concerned with 5 hexes and a front or army group with 9.

All this might be completely missing the mark, but it seems this is the layer that Elmer is presently missing that has the most impact on the majority of PO play -- where the basic objective is to break a line and exploit it.

< Message edited by USXpat -- 2/1/2014 1:15:01 PM >

(in reply to Curtis Lemay)

Post #: 7

RE: Strategy 101 resources wanted - 2/1/2014 12:21:55 PM

Zovs

Posts: 6668
Joined: 2/23/2009
From: United States
Status: offline

Now that got me thinking.

We already have set objectives in for the scenario designer so he can assign objectives to a formation.

I don't know the guts yet, but why not add Set Strategic Objective option?

I would limited it to 2-4 per side or for only Army HQ. Maybe it's like a Theater Option or something for the HQ or the Player?

Basically set objectives tells the formation to 'travel' (and I presume) and attack these hexes on this way point.

Set Strategic Objective could possible be:

Attack this river line there ->
Defend this Victory Hex here ->

I don't know it gets hard after a bit because you have to mix and match strategies with tactics sometimes.

From a purely board war game perspective (or any game for that matter) after you setup you have to study the map, understand the rules (again the mechanics) and then you start to form a plan on how your going to achieve your objectives.

See what I did there?

Its the reverse of how a scenario is designed. A designer will setup the map and units and then set objectives for those units (maybe he is guiding the scenario path, maybe he is adding his won bias in on how he wants the AI to proceed, maybe its his thought that this is the best course to follow).

So to me the strategy is setup in my mind like so:

Look at the map
Look at my units units positions
Look at the enemy units positions
Study the terrain I must cross
Find the Victory locations/hexes
Then formulate a plan of action to get to the Victory Hexes

Once the basic strategy is set in my mind I implement it as best I can and then re-evaluate it based on the opponents reactions to my moves, plans and the fortunes of war.

Rinse, repeat

_____________________________

Beta Tester for:
Flashpoint Campaigns: Sudden Storm
War in the East 1 & 2
WarPlan & WarPlan Pacific
Valor & Victory
DG CWIE 2
SPWW2 & SPMBT scenario creator

(in reply to USXpat)

Post #: 8

RE: Strategy 101 resources wanted - 2/1/2014 12:56:23 PM

golden delicious

Posts: 5575
Joined: 9/5/2000
From: London, Surrey, United Kingdom
Status: offline

quote:

ORIGINAL: Telumar

Re scripting as mentioned by dlazov66: Wouldn't it be possible to implement 'designer scripting' for the first turn? The scenario designer could 'play' the AI's first moves and order combats which could be stored in the scenario.

Wouldn't it be more elegant to simply start the scenario after these moves have been made?

_____________________________

"What did you read at university?"
"War Studies"
"War? Huh. What is it good for?"
"Absolutely nothing."

(in reply to Telumar)

Post #: 9

RE: Strategy 101 resources wanted - 2/1/2014 12:59:06 PM

golden delicious

Posts: 5575
Joined: 9/5/2000
From: London, Surrey, United Kingdom
Status: offline

quote:

ORIGINAL: dlazov66

Set Strategic Objective could possible be:

Attack this river line there ->
Defend this Victory Hex here ->

Remember, the object of war is to destroy the enemy's ability to resist. Going after geographic objectives might work in certain shorter scenarios, but in most it will be easily exploited by the other force. Strategy needs to be about bringing the enemy's forces to battle in conditions which allow them to be destroyed.

However, now that I think about it (see below) it's massively easier to program an AI with geographic objectives.

< Message edited by golden delicious -- 2/1/2014 2:11:30 PM >

_____________________________

"What did you read at university?"
"War Studies"
"War? Huh. What is it good for?"
"Absolutely nothing."

(in reply to Zovs)

Post #: 10

RE: Strategy 101 resources wanted - 2/1/2014 1:10:24 PM

golden delicious

Posts: 5575
Joined: 9/5/2000
From: London, Surrey, United Kingdom
Status: offline

In general I think that the current objective chain system works well for the lower level, the problem comes in making this behaviour joined up across the whole map.

I'd suggest a system in which the designer maps out a relatively small set of objective strings which tells the game what the good paths of attack are and where they can potentially be resisted. Then each formation would be assigned to one of these paths- so far this is what we have with a slightly different structure (formations assigned to paths rather than vice versa)

Then I'd build in a calculation at the beginning of the PO turn where the AI looks at each of these paths and decides;
1) Am I being/about to be overwhelmed along this path?
2) Am I/do I have an opportunity to be exploiting along this path?

Formations would be assigned to 1 and 2 with priority depending on the strategic bias (i.e. aggressive AIs will ignore risks in favour of opportunities). Formations in areas where things are quiet are likely to be pulled out either to these other paths or to reserve.

Making the decisions 1 and 2 above is a complex problem but broadly speaking there are three things to consider;
1) The positioning of the defending forces (e.g. fortified status, difficult terrain)
2) The relative combat factors of the two forces (note here that the AI ought to be limited to units which it can actually see under the normal recon rules)
3) Whether ground has been lost or gained on this path in the last couple of turns.

_____________________________

"What did you read at university?"
"War Studies"
"War? Huh. What is it good for?"
"Absolutely nothing."

(in reply to golden delicious)

Post #: 11

RE: Strategy 101 resources wanted - 2/1/2014 2:37:55 PM

ralphtricky

Posts: 6685
Joined: 7/27/2003
From: Colorado Springs
Status: offline

Some of these are working, possibly the need to be tweaked, but...

quote:

ORIGINAL: USXpat
As an aside, one option that would be useful is an "Assist" order that could be further associated with a specific formation (even if it needs to be defined via an event) or as a global variable that would let Elmer define which formation/s need the most help.

Independent orders are supposed to try to do something like that, I believe it's the formation with a similar track of safe objectives that's least far along it's path.

quote:

As the player's objective is to break that line, Elmer needs additional ability to recognize it when it happens or even respond to the possibility of it happening.

If the objects are at least 4 hexes apart, he actually should understand the concept of a 'front' and reacts to it being breached. The front is defined as a a line through the front objective perpendicular to the objective behind it. If the hexes are three or less, then it's a circle of 1/2 the radius of the front. I can't guarantee that he does it well, but he's supposed to recognize it by either advancing to the next objective or retreating to the last one.

quote:

As far as I can determine, Elmer treats units and formations tagged as a regiment, division, corps or army as the same. I'm not a programmer but suspect it would be difficult, but maybe not -- is for higher level HQs/formations to consider its objectives based upon a radius around an objective.

In example - most formations (say a corps) would be concerned with specific objective hexes (or objective with a radius of 1); an army might then be concerned with 5 hexes and a front or army group with 9.

Objective control is based on a radius right now, but it's dynamically set each turn. I've considered adding the radius (+/-1) to each objective or formation. Right now, it's 5-9 + 4 for the smartest intelligence option.

(in reply to USXpat)

Post #: 12

RE: Strategy 101 resources wanted - 2/1/2014 3:28:33 PM

Telumar

Posts: 2236
Joined: 1/3/2006
From: niflheim
Status: offline

quote:

ORIGINAL: golden delicious

quote:

ORIGINAL: Telumar

Re scripting as mentioned by dlazov66: Wouldn't it be possible to implement 'designer scripting' for the first turn? The scenario designer could 'play' the AI's first moves and order combats which could be stored in the scenario.

Wouldn't it be more elegant to simply start the scenario after these moves have been made?

Actually, yes. But that would result in two scenario files, one for PBEM and one (one turn ahead) for play vs the AI.

quote:

ORIGINAL: golden delicious
I'd suggest a system in which the designer maps out a relatively small set of objective strings which tells the game what the good paths of attack are and where they can potentially be resisted. Then each formation would be assigned to one of these paths- so far this is what we have with a slightly different structure (formations assigned to paths rather than vice versa)

This is a really good idea. It would also prevent never closing gaps in AI lines which usually happens when the AI loses a formation.

_____________________________

http://thetoawbeachhead.wordpress.com

(in reply to golden delicious)

Post #: 13

RE: Strategy 101 resources wanted - 2/1/2014 3:58:23 PM

Oberst_Klink

Posts: 4778
Joined: 2/10/2008
From: Germany
Status: offline

quote:

ORIGINAL: ralphtrick

I'd like to try to improve Elmer (the AI) at a strategic level, where he has a complete lobotomy right now. He's controlled strictly by the objectives. I don't mind that idea, but I'd like to figure out some way to change the objectives so that he plays better.

The strategic level is about moving groups of formations, not individual units. Do I attack left, right, middle? Do I defend strongly here or draw back to where it's more defensible? Unlike Chess where the computer can look many moves ahead, TOAW has to try to figure out what the position looks like and what a coherent strategy is.

Does anything know of any resources (books, blogs, etc.) that talk about that analysis at the level of TOAW? There is plenty at the squad level, but I'm not seeing anything at a bigger scale.

I've read everything I can find on AI, I've even got a new book arriving today, but most of the articles are about FPS games right now. I can do the map analysis and figure out the chokepoints, I think I understand enough to start on a strategic AI, but deciding when to try to flank, how to respond to an attempted flank attack, and when to thrust up the center are something that is a bit vague still. Even something as simple as when to mop up the stragglers and when to let them go in favor of advancing as far as possible is fuzzy.

Thanks,
Ralph

Ralph,

I can highly recommend a waragming system from the 90s that was never being released commercially but which had the best AI scripting far ahead of its time. It's called CRISIS - by Dutch Owen. It includes the design thoughts of the author, the scripting and of course the game. It's available here:

https://dl.dropboxusercontent.com/u/56215959/crisis73.rar

Extracts from the doc:

CCL is a language for issuing commands to CRISIS headquarters and
units when a side is under control of the AI. The strategic level
"AI" of a CRISIS scenario *is* the CCL written for that scenario, in
a file named <scenario>.CRL where <scenario> is the first part of the
scenario save file (example: PRACTICE.CRS would look for it's CCL in
PRACTICE.CRL).

Operational level AI is built into the game -- it knows how to defend,
attack, patrol, handle logistics, etc. But the direction of that
operational level via the HQ units to select objectives and chart
a path to victory is the function of the strategic level, hence, it
is the function of the CCL writer.

A given CCL file is for one scenario only, and it references data in
that scenario, by map coordinates and by city/unit/square names. The
map and units of a scenario and the CCL that drives the AI are
closely interrelated, and often changes to one means changes to both.

Klink, Oberst

_____________________________

My Blog & on Twitter.
Visit CS Legion on Twitter & Facebook for updates.

(in reply to ralphtricky)

Post #: 14

RE: Strategy 101 resources wanted - 2/1/2014 4:43:23 PM

Zovs

Posts: 6668
Joined: 2/23/2009
From: United States
Status: offline

Nice find Oberst Klink!

I have always been fascinated by AI when I was in CIS back in the nineties. I am also a C, C++ and java programmer and I know how hard it is to program AI stuff, so this is a neat find!

_____________________________

Beta Tester for:
Flashpoint Campaigns: Sudden Storm
War in the East 1 & 2
WarPlan & WarPlan Pacific
Valor & Victory
DG CWIE 2
SPWW2 & SPMBT scenario creator

(in reply to Oberst_Klink)

Post #: 15

RE: Strategy 101 resources wanted - 2/1/2014 6:02:31 PM

Oberst_Klink

Posts: 4778
Joined: 2/10/2008
From: Germany
Status: offline

quote:

ORIGINAL: dlazov66

Nice find Oberst Klink!

I have always been fascinated by AI when I was in CIS back in the nineties. I am also a C, C++ and java programmer and I know how hard it is to program AI stuff, so this is a neat find!

I'd call Dutch Owen the master of AI scripting. The program IMHO is still a classic and I am still fascinated by the simplicity of the AI script and how it gives the AI, more or less, options in order to follow its orders and react to new challenges. Even though I am cr*p at scripting or programming languages in general, I am sure you agree it's highly recommend for future wargaming developers, no?

Klink, Oberst

_____________________________

My Blog & on Twitter.
Visit CS Legion on Twitter & Facebook for updates.

(in reply to Zovs)

Post #: 16

RE: Strategy 101 resources wanted - 2/1/2014 9:44:45 PM

rhinobones

Posts: 1540
Joined: 2/17/2002
Status: offline

quote:

ORIGINAL: ralphtrick
I'd like to try to improve Elmer (the AI) at a strategic level, where he has a complete lobotomy right now. He's controlled strictly by the objectives. I don't mind that idea, but I'd like to figure out some way to change the objectives so that he plays better.

A couple of suggestions:

1. Add a feature where an individual unit, such as a HQ or supply unit, can be designated as an objective. As the unit moves, so does the objective.

2. Add events for when an objective is taken multiple PO objective track probabilities are triggered. If a PO objective is captured, multiple PO formations might be diverted to recapture the objective. If the PO captures an objective, options could be triggered to send PO formations on alternative objective tracks.

3. Create a new class of "objective like" hexes which act as triggers for influencing PO reaction but are not counted towards victory points. The reaction could be a change in objective priorities and/or objective track. Recommend that these new hexes are not visible during play.

For all of the above add in some event probability variables for extra confusion. Additional objective tracks may also be required.

Regards, RhinoBones

_____________________________

Colin Wright:
Comprehensive Wishlist Forum #467 . . . The Norm (blessed be His name, genuflect three times and accept all values in the program as revealed truth)

Pre Combat Air Strikes # 64 . . . I need have no concern about keeping it civil

(in reply to ralphtricky)

Post #: 17

RE: Strategy 101 resources wanted - 2/2/2014 4:20:56 PM

ogar

Posts: 297
Joined: 9/6/2009
Status: offline

I kept writing, and re-writing, a long post on strategic goals/directions/constraints vs operational level goals/methods/resources... And peppering Ralph with extra questions.

Short version --
what Rhinobones just wrote above. Humans react to changes at the operational level (capture Tobruk; massive losses to 2nd Tank Army; etc.) by adjusting their operational plans, and those changes can impact the overall strategy (how do I win this game ?). The suggestions above give some quantifiable or measurable points by which the PO can possibly change strategy.

(in reply to rhinobones)

Post #: 18

AI web resources and TOAW AI in large scenarios. - 2/2/2014 5:37:36 PM

governato

Posts: 1079
Joined: 5/6/2011
From: Seattle, WA
Status: offline

Just a few resources I stumbled upon

Gamasutra: designing an AI for strategic games

fuzzy logic (real time AI)

A good article specific to strategic games. It points to some interesting computer science resources...

I find the AI for Command Ops really brilliant. I would buy any game at the divisional level that had such a great C&C structure.

So far the weakest parts of the TOAW AI are:

a) the ability to keep a continuous front. I think it is partially caused by me not being good at programming the AI...but also from the fact that different formations act independently of each other.

b) the AI moves units all the time, and units rarely stop to keep decent supply/readiness levels. Players do the same and I think that is partially due to a lack of penalties for low supply.

(in reply to ogar)

Post #: 19

RE: Strategy 101 resources wanted - 2/2/2014 5:56:03 PM

Lobster

Posts: 5104
Joined: 8/8/2013
From: Third rock from the Sun.
Status: offline

After reading Rhinobones post #16 suggestion: Keep in mind humans (some rookies at programming) will be programming this stuff for their scenarios and it can be very easy to make spaghetti. So some kind of structured architecture would be desirable. The current editor is a good spaghetti maker. But he has some good points.

< Message edited by Lobster -- 2/2/2014 6:58:01 PM >

(in reply to ogar)

Post #: 20

RE: AI web resources and TOAW AI in large scenarios. - 2/5/2014 6:02:07 PM

ralphtricky

Posts: 6685
Joined: 7/27/2003
From: Colorado Springs
Status: offline

quote:

ORIGINAL: governato
Just a few resources I stumbled upon

...

So far the weakest parts of the TOAW AI are:

a) the ability to keep a continuous front. I think it is partially caused by me not being good at programming the AI...but also from the fact that different formations act independently of each other.

b) the AI moves units all the time, and units rarely stop to keep decent supply/readiness levels. Players do the same and I think that is partially due to a lack of penalties for low supply.

Thanks for the links.

I've also been following the 'At the Gates' game and have access to the design documents (part of Kickstarter) which also has some ideas.

On the first one, that's part of why I need advice on the strategy layer. I don't know enough about when Elmer should try to do a breakthrough and encircle some units (without being cut off) and when to head for the victory hexes and when to do something else. If I don't know those things, it becomes pretty hard to advise ELmer.

There are all kinds of resources on small unit tactics, but very little on strategy.

On the second one, Elmer does that because the players do ;) 3.5 adds an 'OverExtended' supply level if defined by the scenario designer. There is code to rest units when the formation supply levels go below 25%, I'll need to enhance that if the overextended supply levels are in place. I've tried to make Elmer dance less, but I know that I can do more. Some of the moves can definitely be eliminated, part of the problem is that he doesn't have an overall plan.

(in reply to governato)

Post #: 21

RE: Strategy 101 resources wanted - 2/5/2014 6:16:19 PM

ralphtricky

Posts: 6685
Joined: 7/27/2003
From: Colorado Springs
Status: offline

quote:

ORIGINAL: Telumar
Another idea that i have is to expand on the .ai file and strategic bias system in doing something like "AI doctrine" that can be switched during the scenario (by events). The AI doctrine could contain settings like agressiveness, air settings (combat support vs interdiction), emphasize on destroying enemy forces or capturing objectives. Starve cut off enemies or destroy asap. Stubborn defense or elastic defense. Acceptable Loss ratios. etc.

Some of those are easy.
Aggressiveness is captured already, I think.
Combat support vs interdiction is possible, right now, it's random.
Destroying forces or capturing objectives. I'm not sure how to do that one.
Starve cutoff units or destroy should be possible.
Stubborn defense or elastic is simple, I think.
Acceptable loss ratios - what do you do when they are exceeded? withdraw the formation?

_____________________________

Ralph Trickey
TOAW IV Programmer
Blog: http://operationalwarfare.com
---
My comments are my own, and do not represent the views of any other person or entity. Nothing that I say should be construed in any way as a promise of anything.

(in reply to Telumar)

Post #: 22

RE: Strategy 101 resources wanted - 2/5/2014 6:22:44 PM

ralphtricky

Posts: 6685
Joined: 7/27/2003
From: Colorado Springs
Status: offline

quote:

ORIGINAL: Lobster

After reading Rhinobones post #16 suggestion: Keep in mind humans (some rookies at programming) will be programming this stuff for their scenarios and it can be very easy to make spaghetti. So some kind of structured architecture would be desirable. The current editor is a good spaghetti maker. But he has some good points.

I think part of the problem is that there are two different things that the editor is trying to do. One is the scenario design and editing. The other is the AI design.

Even though they may use events under the hood, there is no reason that you might not be able to edit an 'objective' and bring up a list of things that can happen when it's won/lost, including triggering an event for more complex things.

(in reply to Lobster)

Post #: 23

RE: Strategy 101 resources wanted - 2/5/2014 9:16:13 PM

Telumar

Posts: 2236
Joined: 1/3/2006
From: niflheim
Status: offline

quote:

ORIGINAL: ralphtrick

quote:

ORIGINAL: Telumar
Another idea that i have is to expand on the .ai file and strategic bias system in doing something like "AI doctrine" that can be switched during the scenario (by events). The AI doctrine could contain settings like agressiveness, air settings (combat support vs interdiction), emphasize on destroying enemy forces or capturing objectives. Starve cut off enemies or destroy asap. Stubborn defense or elastic defense. Acceptable Loss ratios. etc.

Some of those are easy.
Aggressiveness is captured already, I think.
Combat support vs interdiction is possible, right now, it's random.
Destroying forces or capturing objectives. I'm not sure how to do that one.
Starve cutoff units or destroy should be possible.
Stubborn defense or elastic is simple, I think.
Acceptable loss ratios - what do you do when they are exceeded? withdraw the formation?

Yes, aggressiveness is captured by the Strategic Bias.

Combat support vs interdiction - get me right, i don't mean either this or that, i mean an emphasize on the one without omitting the other.

quote:

Acceptable loss ratios - what do you do when they are exceeded? withdraw the formation?

You know this is difficult and no simple answer can be given. Yes, withdraw is an option, but so is exchanging a formation by another or if on the attack cancelling the offensive etc. Things that i (and others) do are looking at own losses, i.e. in infantry, and compare it to the replacements. Something that good players in longer campaigns do look at (i.e. FitE). But as said, in longer camapigns. In a 15 turn game this is not as significant as in 200+ turns scenario.

The longer i think about it, the less i like my idea about the doctrine screen. A good AI probably wouldn't need it. And most is so situation dependent that it's impossible to catch all with something as generalized as a global AI doctrine.

quote:

ORIGINAL: ralphtrick

On the first one, that's part of why I need advice on the strategy layer. I don't know enough about when Elmer should try to do a breakthrough and encircle some units (without being cut off) and when to head for the victory hexes and when to do something else. If I don't know those things, it becomes pretty hard to advise ELmer. There are all kinds of resources on small unit tactics, but very little on strategy.

So, let's get that clear. You are looking for player advice on how and why they do this and that? Tactics, operational evaluations etc. No nerdy tech stuff, just nerdy armchair general stuff? If so, have you looked in the AAR section? Though i doubt that there is such an AAR with a player going into 'deep musing'.. but one never knows. There has been a TOAW workshop at gamesquad.com when TOAW3 came out which might be worth a look, i think the old threads should still be there in the AAR section.
Maybe it would be better if you asked a specific question, maybe even present a certain situation and simply ask what and how to do and what and why not etc. You'll get as much opinions as players answering, but it wouldn't hurt. The other thing that would be even better is play, play and play, but that would eat up your precious programming time...

_____________________________

http://thetoawbeachhead.wordpress.com

(in reply to ralphtricky)

Post #: 24

RE: Strategy 101 resources wanted - 2/6/2014 3:24:28 AM

governato

Posts: 1079
Joined: 5/6/2011
From: Seattle, WA
Status: offline

quote:

ORIGINAL: ralphtrick

On the first one, that's part of why I need advice on the strategy layer. I don't know enough about when Elmer should try to do a breakthrough and encircle some units (without being cut off) and when to head for the victory hexes and when to do something else. If I don't know those things, it becomes pretty hard to advise ELmer. There are all kinds of resources on small unit tactics, but very little on strategy.

quote:

ORIGINAL: Telumar
So, let's get that clear. You are looking for player advice on how and why they do this and that? Tactics, operational evaluations etc. No nerdy tech stuff, just nerdy armchair general stuff? If so, have you looked in the AAR section?

.

I like the idea of general settings that drive the global strategy of the AI, say:

- propensity of getting encirclements vs frontal attacks (German vs Red Army doctrine)
- effort to keep a continuous front (large scale scenarios vs small scale or WWII vs Napoleonic era)

That would enable the scenario designer to set the general behaviour of the AI in a way that is consistent with the scale of the scenario and the C&C style of the forces involved. The blessing/curse of TOAW is that it is much broader in scope than other good PC games (as War in the East or Command Ops), that is quite a challenge. So I think some way of helping the PO prioritize depending on the scenario will be absolutely necessary.

Regarding Telumar's point of reading material for real forces (rather than discussion of AIs behaviour) here is a link to a Red Army doctrine paper that discusses
some practical applications of strategic doctrine.

As an intermediate step, I would *love* to see the game develop further what has been introduced in v3.4: the PO takes strategic directions from the player through a series of intermediate objectives (that can be modified during the game) but then the PO does all the tactical moves for each formation. This is what 'Command Ops' does perfectly, but at scales below the operational level of TOAW.

I think this approach, if improved, 'd make some monster scenarios a lot more fun and playable (and realistic).

< Message edited by governato -- 2/6/2014 4:29:24 AM >

(in reply to Telumar)

Post #: 25

RE: Strategy 101 resources wanted - 2/6/2014 4:38:45 AM

larryfulkerson

Posts: 39932
Joined: 4/17/2005
From: Tucson, AZ
Status: offline

quote:

ORIGINAL: governato
....it'd make some monster scenarios a lot more fun and playable (and realistic).

I like this thought. Part of the problem w/ monster scenarios is that they are terrifically hard to playtest properly.
FITE has a very large unit count for both sides and there's like 400+ turns to play. Very few players have
time resources to spend on such a project. So Steve Sill has developed a beautiful scenario called
Sicily to Brenner Pass that has a medium unit count and plenty of turns but the turns pass quickly. It's actually
fun to play agan. And he has another one called D-Day to the Elbe that also is a monster that plays quickly
and is fun to play. This man is a gold mine I tell you. A national treasure. I've played both and I like both of them
and thought I'd say so .

(in reply to governato)

Post #: 26

RE: Strategy 101 resources wanted - 2/6/2014 5:28:17 AM

sPzAbt653

Posts: 9511
Joined: 5/3/2007
From: east coast, usa
Status: offline

Thanks very much ... it is the combined effort of many that makes it all possible, though.

I am dedicated to making solitaire scenarios that hopefully are worth playing. Its difficult because there is no 'PO Manual', so its all trial and error. Ralph helps out immensly, for example by increasing the PO Tracks from 3 to 5. I have an opinion that the addition of more tracks, used in conjunction with the Variable Value, could result in a decent PO on a semi-strategic level. Rick and I had some success with this in Directive 21.

But some of Elmers' basics lack, and I think this might create difficulties in getting a decent Strategic Elmer. I have no programming education so I can't speak intelligently on AI theory, but if everything that a human player sees can be transformed into math, I don't see why a PO can't reach a reasonable conclusion to situations. A human player can look at the front line and see where the weakness is, then can make adjustments. Overall this decision is based on mathmatical calculations of unit strengths. A program ought to be able to make the same calculations and arrive at reasonable conclusions, I would think. But Elmer often likes to overstack and not protect unit flanks. This degrades many scenarios into an excersise of denying supply to the PO in order to reduce its units, as opposed to being an excersise in strategy and tactics.

(in reply to larryfulkerson)

Post #: 27

RE: Strategy 101 resources wanted - 2/6/2014 2:07:21 PM

Curtis Lemay

Posts: 12969
Joined: 9/17/2004
From: Houston, TX
Status: offline

I still hold a scoffer position on all this. I fear that even if everything so far proposed were implemented and then actually used by scenario designers we would still have a PO that would barely suffice as a practice opponent. The PO, so far, has more or less been a black hole down which limitless code can be pored without any real improvement. And this stuff, altogether, would be a huge amount of coding effort.

The cost of pursuing the more elaborate of these will be in fewer actual game features. We don't have infinite coding time. That means that we have to budget what we want. In general, I would rather see more game features. We could afford a few PO ideas with the highest cost/benefit ratios, though, But we have to think in those terms.

(in reply to sPzAbt653)

Post #: 28

RE: Strategy 101 resources wanted - 2/6/2014 2:16:31 PM

Lobster

Posts: 5104
Joined: 8/8/2013
From: Third rock from the Sun.
Status: offline

I think Ralph is on the right track. Judging by the Opponents Wanted section it would appear most people play against the PO, not against human opponents. And if this is indeed true then wouldn't a better PO be important? All the great features in the world would not make a difference if there is no one to play the game because it lacks a decent non human opponent. On the other hand, if Elmer could be made into a decent opponent then the replay value of the game would be a huge selling point. Playing against a brain dead Elmer is not fun after the first few times.

(in reply to Curtis Lemay)

Post #: 29

RE: Strategy 101 resources wanted - 2/6/2014 3:02:12 PM

ralphtricky

Posts: 6685
Joined: 7/27/2003
From: Colorado Springs
Status: offline

quote:

ORIGINAL: Lobster
I think Ralph is on the right track. Judging by the Opponents Wanted section it would appear most people play against the PO, not against human opponents. And if this is indeed true then wouldn't a better PO be important? All the great features in the world would not make a difference if there is no one to play the game because it lacks a decent non human opponent. On the other hand, if Elmer could be made into a decent opponent then the replay value of the game would be a huge selling point. Playing against a brain dead Elmer is not fun after the first few times.

Matrix does have a PBEM system that I want to add support for that may help some with that. Bob is partially right in that an expert level AI would be incredibly expensive to build, but I think that getting an AI that is better than the current one is definitely possible, and worth doing for a number of reasons. I don't expect it to be an expert, just not as stupid, have a more consistent plan, and sometimes doing clever things.