Matrix Games Forums

ORIGINAL: a white rabbit
why should Elmer have more information than we do ?

Unless cheating is desired, AI shouldn't have any more information than a human player in the same situation would have. However, in AI theory, "complete information" and "incomplete information" have special meanings. Complete means roughly that both players have all the information available. Incomplete means there are unknowns. Trivial examples are chess (you see all the pieces in game) and poker (you see only some of the cards in game). It should be obvious that games with perfect information are easier to play for computers. There are so much less possibilities that searching for optimal solutions is feasible. For incomplete information, something smarter is required.

Pattern matching might be useful, but there are limitations. Human brain is massively parallel and very good in such activities. You just have to take a glance on the game map in order to see whether a move makes sense. Usually it doesn't make any sense to move a HQ unit all alone next to enemy armor. You just see it immediately. Not so with AI. It has to do lots of work in order to decide that the move would be suicide for the HQ. Is it a trap or opportunity? How can you tell?

ORIGINAL: a white rabbit
if it's played a lot he'll become virtually unbeatable, if the player dls the totality of scen-models for that scen.

Sure, if the human players are following the common patterns. Should they start doing something really unexpected moves, the library becomes no good until end of the game. Eventually it should improve, but unbeatable it won't be. The random chance still has quite an effect. If attack rolls are always poor, no matter the strategy, one cannot win.

-P

ORIGINAL: vonPryz

If attack rolls are always poor, no matter the strategy, one cannot win.

In TOAW, the sheer number of "rolls" made in the game is so vast that it the probability of having bad luck throughout the game is statistically insignificant.

ORIGINAL: vonPryz

ORIGINAL: a white rabbit
why should Elmer have more information than we do ?

Unless cheating is desired, AI shouldn't have any more information than a human player in the same situation would have. However, in AI theory, "complete information" and "incomplete information" have special meanings. Complete means roughly that both players have all the information available. Incomplete means there are unknowns. Trivial examples are chess (you see all the pieces in game) and poker (you see only some of the cards in game). It should be obvious that games with perfect information are easier to play for computers. There are so much less possibilities that searching for optimal solutions is feasible. For incomplete information, something smarter is required.

Pattern matching might be useful, but there are limitations. Human brain is massively parallel and very good in such activities. You just have to take a glance on the game map in order to see whether a move makes sense. Usually it doesn't make any sense to move a HQ unit all alone next to enemy armor. You just see it immediately. Not so with AI. It has to do lots of work in order to decide that the move would be suicide for the HQ. Is it a trap or opportunity? How can you tell?

ORIGINAL: a white rabbit
if it's played a lot he'll become virtually unbeatable, if the player dls the totality of scen-models for that scen.

Sure, if the human players are following the common patterns. Should they start doing something really unexpected moves, the library becomes no good until end of the game. Eventually it should improve, but unbeatable it won't be. The random chance still has quite an effect. If attack rolls are always poor, no matter the strategy, one cannot win.

-P

...sorry, maybe i missed yr point, the computer has to do more work than we do ? to match patterns..

..come on, no but really, come on, we all match patterns, more or less the same, a computer may be slower having a silicon and not an organic "brain", but we all need the patterns to match with, otherwise, no thought, an experienced player has more patterns, an experienced computer has more patterns, is all...

..and what's chance got to do with anything ? we roll the dice, we recalculate based on the results, just as Elmer would..

..and on to the unexpected..sure, but that would take us all by surprise, i've still got a player bitching about me forcing him into designer no-go zones, i didn't play by the unwritten "rules', so ? i'll willingly give this info to Elmer, kick ass my silicon buddy..

..if it's any help, i trained as a teacher of the mentally handicapped, and worked with the same, before dropping out, and becoming a farmer, moving to France, and getting a degree (it's complicated, equivalent stuff) in French in Agriculture..

..don't ask what i did during my dropping-out period, it was fun, s'all you need to know..

ORIGINAL: golden delicious

ORIGINAL: rhinobones

First you said “overall situation”, then you added “individual units” to the overall situation and now you’re asking about “getting the most” out of the individual units . . . geeze kid, you keep moving the target all over the place.

No, not at all really. My point is about getting the most out of individual units with regard to the overall situation. This has been my point all along.

However, since neither you nor your buddy Collin can figure out an easier way to do things, I figure it must be a cultural perplexity endemic to TDG. Maybe you’re just trying too hard; not everyone seems to share your problem.

I dunno. See the way most people play the large scenarios. It's striking how little relation their play in one sector has to another; making fruitless and costly assaults in one area while another successful sector is starved of reinforcements, etc.

I'm not really interested in discussing my own record as a player- but I'm prepared to go as far as to say I'm not incompetent.

My own experience has been that there are a lot of players who do not look at the big picture as they push units, or don't properly manage the units they do push -- or both. Not looking at the big picture is a particularly endemic flaw -- I'v won several matches over tactically competent players simply because they never looked at the big picture -- kept pushing where they were pushing for no better reason than that's where their units had piled up, etc.

What I've found with playing the monsters is that if one does try to keep the big picture in mind as one moves specific units, and at the same time manage those specific units so as to conduct movement most efficiently, avoid density penalties in movement, mount attacks properly, etc -- that then turns take an extraordinary length of time. Like, one beast I was playing I would routinely start the turns on one day and finish them the next.

This with a scenario that is actually pretty modest in size compared to some of the behemoths out there. I wouldn't play those guys -- how can you?

I don't want to wantonly flame all who play the beasts. However, I can't help thinking of a player who seems to be a lot better than me at this game -- at least, he convincingly trounced me when we played once. Perhaps not coincidentally, he habitually designs rather small scenarios with a modest unit count.

One can have a huge area where various flank activities can go on. I'll also grant that doing the entire Eastern front at 5 km a hex has a certain appeal. However, I'm not sure an active combat area of more than about fifty hexes square is really desirable. As one exceeds that limit, the quality of your play is inevitably going to decline.

Colin,
 
I disagree.  It all comes down to situational awareness and cognitive complexity.  Some people can handle larger loads competently than others.
 
For instance, it was said that one of the reasons Napoleon was so superior to other commanders was because he could grasp and manage larger situations and larger battles much quicker and more effectively.
 
Thus, it depends on the individual.  Some players will never be Napoleons on the TOAW battlefield.  Others will be.

ORIGINAL: vahauser

Colin,

I disagree. It all comes down to situational awareness and cognitive complexity. Some people can handle larger loads competently than others.

For instance, it was said that one of the reasons Napoleon was so superior to other commanders was because he could grasp and manage larger situations and larger battles much quicker and more effectively.

Thus, it depends on the individual. Some players will never be Napoleons on the TOAW battlefield. Others will be.

No comment.

..i'll freely admit to loosing it on some of the monsters, this before the new zoom levels, the things are just so big that it gets difficult even knowing which direction you're supposed to be going, let alone how the current combat fits the over-all picture..

ORIGINAL: a white rabbit

..take a serious look at the future of a game with an AI that improves automatically most times it's played, "most" 'cos some won't take the auto-update per scen' option, "most" 'cos most will, a better AI, yahhh how good am i ? really ?..

..we all have the machinery and connections to the internet, and given the lack of screams, we're also ready to give up to 12 hours thinking time (optional length) and the computer space/ storage memory, and you have a game that'l do 6-hours to a month moves, a solid game, tried and tested...

..think "to be", not now, and really not 'then'..

AI deliberation without arbitrary time and resources constraints, Internet connections, pooled experience, emergent AI, ...

Yes, now we're thinking! Outside the box!!

Not all of these crazy ideas will be (soon) feasible, but some might, and they are all worth considering.

..Thank you berto..

..it all exists now, 24/7 internet at high speed, anti-virus programs that phone-home and auto-update, so why can't Elmer contact a patterns library ? computers that run 24/7, home computers that could run a moon landing and a pooled military/wargames experience that must run into 100s of years actual playing time..oh, and an engine that despite our bitching, actually works well..

..just gotta join up the dots..

ORIGINAL: a white rabbit

so why can't Elmer contact a patterns library ?

Because we don't want to play Fite against Elmer if he has access to Karri's database.

ORIGINAL: a white rabbit

..Thank you berto..

..it all exists now, 24/7 internet at high speed, anti-virus programs that phone-home and auto-update, so why can't Elmer contact a patterns library ? computers that run 24/7, home computers that could run a moon landing and a pooled military/wargames experience that must run into 100s of years actual playing time..oh, and an engine that despite our bitching, actually works well.

..just gotta join up the dots..

You'd have to be careful because of privacy concerns, but tracking all moves and either analyzing the games off-line on the player's PC or optionally sending a summary of the games moves to a central point to be analyzed (or some combination) is theoretically possible. You need to be careful about this, because it's easy to unbalance the AI, especially with Neural Networks.

Heck, just getting Elmer (Future name Bugs(?)) a strategic layer that's good enough that he could play against himself to try to determine the best strategy against the human's plays might be helpful in building that up too.[;)] If you can determine that on turn 10, an apparent strong concentration on the left flank can probably be successfully be responded to by ... he's got a leg up. Figuring out how to successfully do that, however, is a tough problem to solve.

If nothing else, having a finer layer of tracking than the current one-turn VCR might help with figuring out ways to send in error reports and a better way to tune Elmer/Bugs/Scenarios.

For the near future, I'm keeping my options open about what the best way to do these things is. I suspect that there isn't going to be just one answer, but a combination of different approaches.

Ralph

I welcome a new version but that Xbox 360 stuff scares me. That is arcade level, not my thing at all.


Regards,

Grell

ORIGINAL: vahauser

For instance, it was said that one of the reasons Napoleon was so superior to other commanders was because he could grasp and manage larger situations and larger battles much quicker and more effectively.

Well this would depend on the scenario. Is Napoleon controlling twenty divisions or two hundred battalions?

In TOAW, one doesn't have subordinates. If the scenario includes the battalions, you have to micromanage them whether it's appropriate or not.

This is the crux of the question. How many units can a commander control without beginning to lose effectiveness? Well, how many units were they historically expected to command? OKH didn't direct the movement of individual regiments on the Eastern Front because it was impractical for a single HQ to do so. Unless you have a more efficient staff than OKH, you're going to lose efficiency by controlling the Eastern Front at regiment level.

ORIGINAL: ralphtrick

ORIGINAL: a white rabbit

..Thank you berto..

..it all exists now, 24/7 internet at high speed, anti-virus programs that phone-home and auto-update, so why can't Elmer contact a patterns library ? computers that run 24/7, home computers that could run a moon landing and a pooled military/wargames experience that must run into 100s of years actual playing time..oh, and an engine that despite our bitching, actually works well.

..just gotta join up the dots..

You'd have to be careful because of privacy concerns, but tracking all moves and either analyzing the games off-line on the player's PC or optionally sending a summary of the games moves to a central point to be analyzed (or some combination) is theoretically possible. You need to be careful about this, because it's easy to unbalance the AI, especially with Neural Networks.

Heck, just getting Elmer (Future name Bugs(?)) a strategic layer that's good enough that he could play against himself to try to determine the best strategy against the human's plays might be helpful in building that up too.[;)] If you can determine that on turn 10, an apparent strong concentration on the left flank can probably be successfully be responded to by ... he's got a leg up. Figuring out how to successfully do that, however, is a tough problem to solve.

If nothing else, having a finer layer of tracking than the current one-turn VCR might help with figuring out ways to send in error reports and a better way to tune Elmer/Bugs/Scenarios.

For the near future, I'm keeping my options open about what the best way to do these things is. I suspect that there isn't going to be just one answer, but a combination of different approaches.

Ralph

..re privacy concerns..i wasn't really thinking of a p2p, eg emule, approach, more a central library similar to the scen libraries already in existance, The latter gives Elmer a lot more reference data always available and makes dling optional. If he only stores automatically in the player's computer the actual game, then compares the developing patterns to previous games in the library, he should be able to work out what's most likely to happen in future moves...

..re strategic layer.. i don't see it being possible without a second map, possibly a third at max zoom out for big scens, simplified at the relevant scale, and the formations simplified to one piece representing the center of the formation ? Don't forget this map level isn't a combat level map, it's a planning map, the one used to compare patterns, he get's his own technical data from the combat map, by assessing as now but simplified, eg loss = -1, draw = 0, win = +1 , no combat= 0, add these for each unit in the formation and then " is the formation's attack possibility greater than x (pick a number), y/n ?' store result, go on to next formation and when all formations have a result compare with the patterns till one fits (gets most correlation points) and is a game winner..execute..

..re stategic layer 2..i'm not suggesting that we try to make Elmer intelligent *, just that we find a way to give him memory/ chance to study, just as the best chess and iGo games have. Pattern comparison he can do, a simplified zoom-out map is all he needs to do it quickly and efficently, after all on the more popular scens he's gonna have a fair few pattern maps..

..* not immediately anyway..

..an after thought, is there any reason why the library couldn't have pbem and vs PO games in it ? that would seriously increase his abilities..

ORIGINAL: Grell
I welcome a new version but that Xbox 360 stuff scares me. That is arcade level, not my thing at all.

Regards,
Grell

No, it'd be the same game, mainly played with smaller scenarion, more online play. The interface would have to use only a joystick, the mechanics of the interface would definitely have to change for the lower resolution, but the game would be the same game.

Jesus Chrysler, Ralph!

Say it's not so!

Going xbox/arcade killed the last chance for MechWarrior to be a 'serious' game, eh? The whole 'mythos' is now 'twitch-time gamey'.....

Your scaring me........................

ORIGINAL: golden delicious
How many units can a commander control without beginning to lose effectiveness? Well, how many units were they historically expected to command?

The answers differ a bit on that one. Organization theory apparantely says seven subordinates is the maximum for any level. Clausewitz wanted an army divided into four to eight pieces, corps or lower into four to six; the Swiss writer Ruestow recommended six to eight sub-units on the army level. The German army in the 19th and early 20th centuries on the whole preferred much smaller numbers, usually two to three major sub-units on each level. British and Americans in the Cold War opted for four or five. In any case, it's usually taken for granted that anything in excess of the given numbers is too much for a single HQ to control. The eight infantry corps and five cavalry divisions plus artillery reserve, i.e. fourteen major sub-units, that the Austrian North Army commanded directly in 1866 were too much to keep track of by any standard.

ORIGINAL: Heldenkaiser

The answers differ a bit on that one. Organization theory apparantely says seven subordinates is the maximum for any level. Clausewitz wanted an army divided into four to eight pieces, corps or lower into four to six; the Swiss writer Ruestow recommended six to eight sub-units on the army level. The German army in the 19th and early 20th centuries on the whole preferred much smaller numbers, usually two to three major sub-units on each level. British and Americans in the Cold War opted for four or five. In any case, it's usually taken for granted that anything in excess of the given numbers is too much for a single HQ to control. The eight infantry corps and five cavalry divisions plus artillery reserve, i.e. fourteen major sub-units, that the Austrian North Army commanded directly in 1866 were too much to keep track of by any standard.

Yeah. Note that commanders are generally expected to be in contact with officers two levels below them. For a lot of scenarios, a player at high command (one level above army group) is commanding units five steps below them (army group, army, corps, division, regiment).

ORIGINAL: L`zard

Jesus Chrysler, Ralph!

Say it's not so!

Going xbox/arcade killed the last chance for MechWarrior to be a 'serious' game, eh? The whole 'mythos' is now 'twitch-time gamey'.....

Your scaring me........................

Interesting ... a reference to an ancient game I played on-line so much it led to carple-tunnel ... What a great game Mechwarrior was. Alas, I no longer play it ... my health suffered too much.

Did you play MW a lot? On-line? as part of a clan? That's refreshing to see other wargamers played MW too.