AI- ?? (Full Version)

All Forums >> [New Releases from Matrix Games] >> Desert War 1940 - 1942



Message


AP514 -> AI- ?? (3/12/2018 6:35:59 AM)

So Will the AI be any good ?? Or will it be like most games made after the mid 90's where Graphic's come first and the AI is an after thought ?




wodin -> RE: AI- ?? (3/12/2018 11:41:26 PM)

Funny, A silver lining being useless gamer in general is on the whole AIs are fine for me. The big hinderence though is when it comes to online FPS games, being useless then is a drag!

Sorry for going on tangent, the little I've seen it seems OK, but I'm prob not the best to ask!




bcgames -> RE: AI- ?? (3/17/2018 7:08:20 AM)


quote:

ORIGINAL: AP514

So Will the AI be any good ?? Or will it be like most games made after the mid 90's where Graphic's come first and the AI is an after thought ?

The AI has two levels; self defense...and then--you versus me.

Self Defense. Units have inherent capabilities to defend themselves. Reconnaissance units are the best example. Their role (unless you change it--which you can) is to collect information and report it; their job is NOT to fight. Therefore, recce units by default will seek to avoid combat.

You vs Me. As it says, the AI is 99.9% you versus me--literally. Desert War 1940-42 uses a scripting language to describe the actions to be performed by the AI. I have described those actions. So if you do a good job and the AI loses--you beat me. Let me know how you won and I'll make it harder the next time around. This is the power the scenario writer can wield with the AI. If this, then that...deep dive into the decision tree! It's the most exciting part of the engine.

And so it follows...if the AI sucks, so does the scenario designer. If the AI issues an ass-whoopin'? Then full props to the scribe.




Solaristics -> RE: AI- ?? (3/22/2018 5:22:54 PM)


quote:

ORIGINAL: bcgames
This is the power the scenario writer can wield with the AI. If this, then that...deep dive into the decision tree! It's the most exciting part of the engine.


You have my attention. I was going to pass on this, but I will follow the forum with interest to see how the AI performs.




KarlXII -> RE: AI- ?? (3/22/2018 5:28:56 PM)

Can anyone describe the benefits of scripted A.I ? I have always been very sceptical about it. It seems like the lazy way to go when not managing to create a dynamical A.I. A scripted A.I for me is an opponent which are very inflexible in handling unpredicted player moves. Or one that is cheating by already knowing where, when and which strength the player will have at a give point.




Saint Ruth -> RE: AI- ?? (3/23/2018 1:37:23 PM)

quote:

ORIGINAL: KarlXII

Can anyone describe the benefits of scripted A.I ? I have always been very sceptical about it. It seems like the lazy way to go when not managing to create a dynamical A.I. A scripted A.I for me is an opponent which are very inflexible in handling unpredicted player moves. Or one that is cheating by already knowing where, when and which strength the player will have at a give point.

Hi, good question.

Well, first off, the AI doesn't cheat, and has no more knowledge than you do when planning its moves.

I did original try and create a dynamic AI and did it for the defense (it choose its own line of defense using terrain etc) but defense is one think but attack? That's an entire different kettle of fish. That'd be a Rommel AI in effect.

So why scripted AI?
By "scripted" I do not mean every unit has its orders plotted in advance.
The scenario creator does the AI 'strategic plan'. The AI has tactical controls over the units.
The script orders "Organisations" to move / attack (e.g. a Brigade or entire Division -> though it can also give orders to individual units)
So the AI might have a script to say "21 Panzer attack north" and all individual orders for each unit would be done by the AI (deciding what enemies to attack, how to position its units, how to allocate air / artillery etc).
The script is also responsive. E.g. it could be "if British attack south then 21 Panzer... " etc.
And also has priority scripts and random chance of scripts being activated etc.

Frankly, now that I've done it, I prefer it because it means the Scenario Creator can make (or not) Rommel attack in exactly the same way as he did. The Scenario creator can make the AI as simple or as complicated as he likes.

Sure, it'd be nice to have the AI do the whole shebang regardless of terrain, deployment, number of units and their strengths etc, but its strategy would never match that produced by a good Scenario Creator...(one can of course "force" an non-scripted AI to do certain things like in attack in a direction by using Victory Locations to influence the AI, but then if you're going to do that, then why not talk to the AI directly with scripts rather than indirectly though victory locations)...




KarlXII -> RE: AI- ?? (3/23/2018 1:52:30 PM)

Thanks for the explanation.

Good to hear the A.I is not cheating by neglecting Fog-Of-War or whatever.

I now understand the balance between supporting scenario designers vs more A.I autonomy. The performance of the A.I is then very much the result of how good the scenario designer is. Still the tactical A.I seems to be very autonomous on the tactical level. I hope that part will be improved upon continously to get the best possible tactical A.I. Isnīt it a problem that the tactical A.I lack the strategic view of the battlefield and will be unable to cooperate with other units which has other scripted goals ?




Saint Ruth -> RE: AI- ?? (3/23/2018 3:18:33 PM)

Hi,
For Orgs with different scripted goals, they should still tactically co-operate. E.g. 2 Regiments attacking along a road with different orders will combine their attacks etc.
Is there room for improvement? Sure!
I do hope to continually improve the tactical AI. The more people who play it, the more feedback I'll get hopefully adn the better it'll get! [;)]
Thanks!
Brian




KarlXII -> RE: AI- ?? (3/23/2018 7:40:04 PM)

Well, now I want to play the game as well :-)




AP514 -> RE: AI- ?? (3/24/2018 1:15:22 AM)

You got me Leaning that way also...thanks for the reply....




giffin -> RE: AI- ?? (3/24/2018 3:51:08 AM)

Well I would say the AI is very Good I still haven't beat the Tutorial 1st Scenario. Anybody have any hints how to beat the Axis? Yes I have watched both Video's 1 on Youtube and the other on Twitch and they were No help :(




SlickWilhelm -> RE: AI- ?? (3/25/2018 1:43:50 AM)


quote:

ORIGINAL: giffin

Well I would say the AI is very Good I still haven't beat the Tutorial 1st Scenario. Anybody have any hints how to beat the Axis? Yes I have watched both Video's 1 on Youtube and the other on Twitch and they were No help :(


giffin, did you read my player's guide/tutorial for the first turn of the Bardia scenario? If you use all of your assets, it should be relatively easy to win that scenario.

Desert War players guide-tutorial




giffin -> RE: AI- ?? (3/25/2018 2:35:23 AM)


quote:

ORIGINAL: SlickWilhelm


quote:

ORIGINAL: giffin

Well I would say the AI is very Good I still haven't beat the Tutorial 1st Scenario. Anybody have any hints how to beat the Axis? Yes I have watched both Video's 1 on Youtube and the other on Twitch and they were No help :(


giffin, did you read my player's guide/tutorial for the first turn of the Bardia scenario? If you use all of your assets, it should be relatively easy to win that scenario.

Desert War players guide-tutorial


Thank you SlickWilhelm I will go over this again. I think I am not allocating proper supplies and grouping my units correctly to win battles.




ETF -> RE: AI- ?? (3/25/2018 8:03:06 PM)

Been wargaming on PC's since the 80's. Never seen an AI (non-cheating) better than a decent beginner player. Let's face it....if you want a challenge you can ONLY play against friends online.




Rosseau -> RE: AI- ?? (3/25/2018 11:24:22 PM)

True - the AI usually wins because it knows the rules better than you. But the scripting thing is a nice addition.




bcgames -> RE: AI- ?? (3/26/2018 5:39:55 AM)

The Desert War AI scripting language is our attempt to codify the US Army's wargaming methodology--from the Intelligence Preparation of the Battlefield (IPB) to Course of Action Development to the Decision Support Template (DST)--and then put it in the hands of the scenario designer. That's Y-O-U.

Even YOU the scenario designer can't create an AI that cheats. If the AI needs to know something to make a decision, then YOU the scenario designer need to allocate reconnaissance assets at the right time and at the right place to get the right info to trigger the decision YOU thought needed to be made.

I can't imagine a more funner (gooder?) place to play for a would-be modder/scenario creator.




daj -> RE: AI- ?? (3/26/2018 8:22:19 PM)

so far I can't say that the AI is very good.

After playing the introductory scenario I played into the blue, first as axis then allies. I lost by about 20 points as Allies, but given the lousy Italian units I should have been killed. I started to worry about both the strategy and tactics of the AI for allies. You would think that the main objective beyond destroying axis units would be to capture the objective hexes. By turn 5 the allies were within 2 squares of Sofafi but never attempted any further assaults, relying only on artillery of air attacks. Worse, initial attack at El Maktila drove to the sea and isolated my units to the east. I thought this is and I can kiss those units goodbye as supply was cut off. But no the allies sat there and made no further attacks. Instead I was able to surround the units cutting off their supply and eventually eliminate them on the 14th turn.

When playing the allies the AI axis strategy seemed to be even worse. Unlike the AI my allies did destroy all the units isolated at El Maktila. I captured Sofafi by turn 5 with little resistance and walked into Sidi Barani with no resistance at all. Given the point value these have for the axis I would have thought that the strategy would be to protect and hold these hexes at all costs.

So far I have to agree with ETF that AI is not good.




Rosseau -> RE: AI- ?? (3/27/2018 2:07:49 AM)

Thanks for sharing your experience. I'm sure you're a much better player than I, which can be a curse in versus-AI games!

However, if that happens to me, I will go into the editor and dump a new enemy unit in Sofafi and Sidi Barani to defend. At least we have the ability to "tweak" the scenarios. But for those unwilling to do some quick editing - and in my experience many are unwilling - it might be a problem.

I and a friend plan to support the game by buying it. So have you. Our hope is the devs and modders will come through with new scenarios or adjust existing ones if needed. I know I haven't played the game yet, but I feel it is a good investment.




Solaristics -> RE: AI- ?? (3/27/2018 11:34:19 AM)

quote:

ORIGINAL: daj

so far I can't say that the AI is very good.

After playing the introductory scenario I played into the blue, first as axis then allies. I lost by about 20 points as Allies, but given the lousy Italian units I should have been killed. I started to worry about both the strategy and tactics of the AI for allies. You would think that the main objective beyond destroying axis units would be to capture the objective hexes. By turn 5 the allies were within 2 squares of Sofafi but never attempted any further assaults, relying only on artillery of air attacks. Worse, initial attack at El Maktila drove to the sea and isolated my units to the east. I thought this is and I can kiss those units goodbye as supply was cut off. But no the allies sat there and made no further attacks. Instead I was able to surround the units cutting off their supply and eventually eliminate them on the 14th turn.

When playing the allies the AI axis strategy seemed to be even worse. Unlike the AI my allies did destroy all the units isolated at El Maktila. I captured Sofafi by turn 5 with little resistance and walked into Sidi Barani with no resistance at all. Given the point value these have for the axis I would have thought that the strategy would be to protect and hold these hexes at all costs.

So far I have to agree with ETF that AI is not good.


This is disappointing to hear. Keep us posted on your experiences with the other scenarios.




daj -> RE: AI- ?? (3/27/2018 12:36:18 PM)

I finished into the blue as the allies and scored about 1350 vs 180 for axis. I watched the game review video step by step to follow the AI axis strategy and tactics. In general it appears that the AI did not bring released units into the fray. The units south of Bugbear just sat there, didn't even take up what appears to be better defensive positions based on terrain. Same for the units near bir el khireqat, they never came to support units I was attacking near Sofafi. When the units fixed near Sidi Barrani were released THEY ALL RETREATED to the west to join the bugbear forces but allowing me to kill off the remaining forces in the battle and take Sidi Barrani without resistance.

Overall, the AI did not bring reinforcements into the battles so that I could outnumber, overpower, and wipe out those in battle.
The quick start guide recommends moving HQ's to keep them out of artillery range and fire. Yet I moved adjacent to a stack of 3 HQ's allowing me not only artillery attack but direct assault. So the tactics are questionable.

I hope my experience and comments might help in tweaking or improving AI.

I am starting Operazione E as axis. Given the poor quality of Italian forces it could be a tough slog.




bcgames -> RE: AI- ?? (3/28/2018 3:29:52 AM)

The AI is scripted to provide an historical experience--not Rommel every time/all the time. If you want a competitive, no-holds-barred-experience, then it is best to play head-to-head. But the tools are there in the engine (AI Tab) if you want to script an AI ass-whoopin'. Maybe the first patch can kick things up a notch...if the suits allow.




Solaristics -> RE: AI- ?? (3/28/2018 8:25:19 AM)


quote:

ORIGINAL: bcgames

The AI is scripted to provide an historical experience--not Rommel every time/all the time. If you want a competitive, no-holds-barred-experience, then it is best to play head-to-head. But the tools are there in the engine (AI Tab) if you want to script an AI ass-whoopin'. Maybe the first patch can kick things up a notch...if the suits allow.


It seems a surprising design decision not to create the most challenging (non-cheating) AI opponent possible if the game has that potential. Some players, me included, have no interest in head-to-head gaming and make purchases based on the single-player experience. I'll be holding off purchasing this game until its AI's potential has been demonstrated in practice.




Saint Ruth -> RE: AI- ?? (3/28/2018 11:04:02 AM)

Well, it's just that an AI will never match a good human, so PBEM will always be better than an AI.
However, we did try to make the best AI possible and we will continue to improve the AI as we get more feedback, thanks, Brian (I'm not Rommel either after all! [8D])




daj -> RE: AI- ?? (3/28/2018 1:10:16 PM)

First,I am sorry. I did not intend to be critical or complaining. I was just adding to the AI thread with my experience based on a single scenario.

Second, I appreciate historical accuracy and the difficulty a designer has in being accurate while creating a challenging game.

Third, I enjoyed the WEGO system which was novel for me and I agree with others that more games such as Battle of the Bulge would be great additions.

Fourth, back to "Into the Blue," not to be critical but for me to understand what happened or perhaps help with a patch. Although it is subtitled the battle for Sidi Barrani, in my scenario there was no battle with axis AI as all the units retreated. Sidi is the major source of points for the axis. They can't abandon it, nor can they allow Sofafi to be taken as it gives allies 15 points/turn.

Finally, I checked the editor but it appeared to be well beyond my abilities. Can your edit a specific scenario? I couldn't figure out how. If I were to edit into the blue, I would add some more experienced Italian units to Sidi and Sofafi and fix them until within enemy ZOC. One could also add fixed artillery a few hexes west of each for support, or make them forts.




bcgames -> RE: AI- ?? (3/28/2018 1:33:05 PM)

I did not explain my point sufficiently. The scenario is the Battle of Sidi Barrani. The AI is scripted to do what General Graziani did--sit or run away. That's what I mean by an historical experience. I originally scripted an aggressive Italian AI and the result was not historical--the British get swarmed by a thousand ants.




spinecruncher -> RE: AI- ?? (3/28/2018 3:37:19 PM)

"I did not explain my point sufficiently. The scenario is the Battle of Sidi Barrani. The AI is scripted to do what General Graziani did--sit or run away. That's what I mean by an historical experience. I originally scripted an aggressive Italian AI and the result was not historical--the British get swarmed by a thousand ants. "

Sounds like you got money waiting in the bank re a "What if" scenario: What if the Italian Commander was more aggressive? What if the ITalians really wanted to fight? Like I have written before, depending on the motivations of the designer, there is plenty of room going forward. We are at 1.0 here. 1.friggen 0. I hope the designer gets a solid enough community for addl material and improved AI, scripting etc. The truth is, the Italians moreover had little heart to fight their former allies, the British. El Duce was never really able to garner a lot of high morale in his troops. I bet the common Italian was thinking, "fight the Brits, they took over India no problem and we could barely take Abyssinia, I am outta here first chance I get."




daj -> RE: AI- ?? (3/28/2018 4:20:16 PM)

Frankly, I am not sure how to respond. Why do we play historical war games? We know what happened, who won, and who lost. If the computer AI is designed to follow the actual battle events, strategy, and tactics for historical accuracy, then what is the point of playing.




Solaristics -> RE: AI- ?? (3/28/2018 4:45:38 PM)


quote:

ORIGINAL: bcgames

The AI is scripted to do what General Graziani did--sit or run away.


While I applaud your attention to historical accuracy, I'd much rather be surprised by the AI doing something challenging and competent than slavishly following historical command decisions. Also, within the context of a wargame, which is always a "what if" situation, having the AI take alternative command decisions within what are otherwise the same historical constraints, does not undermine the historical nature of the scenario in my opinion.




Nico165b165 -> RE: AI- ?? (3/28/2018 5:28:39 PM)


quote:

ORIGINAL: spinecruncher

"I did not explain my point sufficiently. The scenario is the Battle of Sidi Barrani. The AI is scripted to do what General Graziani did--sit or run away. That's what I mean by an historical experience. I originally scripted an aggressive Italian AI and the result was not historical--the British get swarmed by a thousand ants. "

Sounds like you got money waiting in the bank re a "What if" scenario: What if the Italian Commander was more aggressive? What if the ITalians really wanted to fight? Like I have written before, depending on the motivations of the designer, there is plenty of room going forward. We are at 1.0 here. 1.friggen 0. I hope the designer gets a solid enough community for addl material and improved AI, scripting etc. The truth is, the Italians moreover had little heart to fight their former allies, the British. El Duce was never really able to garner a lot of high morale in his troops. I bet the common Italian was thinking, "fight the Brits, they took over India no problem and we could barely take Abyssinia, I am outta here first chance I get."


This.

We could end up with something like, for every scenario, 3 options:

- Historical AI: the AI tries to simulate what their historical counter part did, right or wrong.
- Best AI: the AI tries to use the best options with the units it has to counter the player, hindsight included.
- Random AI: maybe the AI does the historical mistakes/good ideas, maybe it tries to do its best regardless of history. Surprise !

Could this be achieved within the same scenario ? Or does it needs different versions ?




daj -> RE: AI- ?? (3/28/2018 5:49:28 PM)

I am in complete agreement with you. In my opinion historical accuracy ends at setup. If my actions as commander don't follow historical events why should the computer?

If the scenario designer insists on historical outcome then at least we should know ahead of time and the advantage under the scenario synopsis should be edited to indicate computer will lose if historically accurate.

We want a challenge not historical outcome accuracy.

Here is a suggestion: If the historical outcome is that important, fine keep it. But offer us more options of each scenario by changing the AI only. These could be: Easy, hard, and insane if you like.





Page: [1] 2   next >   >>

Valid CSS!




Forum Software © ASPPlayground.NET Advanced Edition 2.4.5 ANSI
0.03125