AI Question

[Fintilgin]

As fond as I am of the Strategic Command series, I do sometimes get bored with the AI always following the same script.

Will the AI attempt different grand strategies, or will it just go through the same sequence every time?

Will the Axis AI attempt Sealion?
Will the Allies ever skip Italy and try for an early Overlord?

Things like that?

[GenSlack]

From what I've seen of the one guy doing YouTube videos of games vs. the PO, it is supremely awful with no fun factor. Within a few turns of every playthrough that I viewed the PO is already outmaneuvered on a grand scale and the game ruined. Not even a remotely interesting opponent. Without the ability to follow the progress of any PvP games, it is impossible to assess game balance or tell whether the mechanics result in any kind of stimulating and/or plausibly realistic game. I was looking forward to this but no way it is a Day 1 purchase for me. I will await the reports of others.

[waynenh57]

How old are those YT videos?

[Fred98]

Only a few months

[stormbringer3]

I've been hoping for a game like this for a long time. I only play vs. the AI. I know that a decent AI is not easy to craft. I'm hoping for at an AI that is at least competent to above average and doesn't make serious blunders too many times per game.

[sol_invictus]

In some of his more recent videos the Russian AI pretty much stopped him in his tracks during the winter of '41. Admittedly, he did make mistakes but it seems that the AI is at least good enough to make the Player keep their head in the game and not try to sleep walk through play. PBEM is where this game will shine I think.

[Fintilgin]

ORIGINAL: GenSlack

From what I've seen of the one guy doing YouTube videos of games vs. the PO, it is supremely awful with no fun factor. Within a few turns of every playthrough that I viewed the PO is already outmaneuvered on a grand scale and the game ruined. Not even a remotely interesting opponent. Without the ability to follow the progress of any PvP games, it is impossible to assess game balance or tell whether the mechanics result in any kind of stimulating and/or plausibly realistic game. I was looking forward to this but no way it is a Day 1 purchase for me. I will await the reports of others.

Oh, I didn't realize there were long forms games on youtube. I will definitely check them out.

As someone who largely exclusively plays against the AI that's disappointing to hear, although not entirely surprising, because I know AI is hard.

[scout1]

ORIGINAL: GenSlack

From what I've seen of the one guy doing YouTube videos of games vs. the PO, it is supremely awful with no fun factor. Within a few turns of every playthrough that I viewed the PO is already outmaneuvered on a grand scale and the game ruined. Not even a remotely interesting opponent. Without the ability to follow the progress of any PvP games, it is impossible to assess game balance or tell whether the mechanics result in any kind of stimulating and/or plausibly realistic game. I was looking forward to this but no way it is a Day 1 purchase for me. I will await the reports of others.

Yeah, will wait this one out to hear feedback …. though I have high hopes ….

[pzgndr]

ORIGINAL: GenSlack
From what I've seen of the one guy doing YouTube videos of games vs. the PO, it is supremely awful with no fun factor.

The "First Look at Beta Build" YouTube video, from back in June almost 4 months ago? A lot of things were updated and improved since beta playtesting started at about that time.

For a new game and initial release, there will be opportunities for more improvements. The AI is in pretty good shape now, certainly much better than 4 months ago. Players should reserve judgement until release and see for themselves.

[AlvaroSousa]

Very difficult to code an A.I. for a game as deep as WarPlan. It takes time and patient beta testers. It took a very long time to develop the A.I. for SC. It is a very good A.I. for that size game. It works via fuzzy logic and scripting like WarPlan does.

As for different strategies....

The Axis A.I. has 2 major strategies and a couple minor ones. It might do somethings historical if the opportunity and time present itself and it might do other things non-historical. I tried to have the A.I. realize it has no intuition and work around that with its strengths.

The Allied A.I. is more reactionary. But generally there is no multiple path it can take on a large level. It does have some small level non-historical actions depending on the situation.

Like SC WarPlan can only get better.

[GenSlack]

No, I am talking about the series of them that he's done up to present. Prior to posting, the last one I saw was around 3 weeks old. Seems a little better in the latest 1943 one that I watched yesterday.

[overkill01]

ORIGINAL: Alvaro Sousa

The Allied A.I. is more reactionary. But generally there is no multiple path it can take on a large level. It does have some small level non-historical actions depending on the situation.

Like SC WarPlan can only get better.

Will the Allied AI be potent enough to launch a succesfull invasion into Europe when a German player conquered it as they have historically, or will it just sit there in England ? The game looks very promising, and I understand AI is a though nut to crack, but AI makes or breaks a game.

[AlvaroSousa]

The A.I. invades of course. It does what it is supposed to do.

Keep in mind no A.I. at this size game can beat a human at historical levels. Unity of Command can have a very good A.I. because it is a limited scope game on a small field of battle. But when as many hexes as Europe has with all the dynamics it is very difficult to program well. I am friends with Hubert Cater (SC owner) and we talk about this kind of stuff often enough. I beat SC3 decisively the first time I played.

WarPlan has 4 supply levels and 11 experience levels. One of the beta testers put it a little tougher and he said the A.I. took advantage of it.

[AlbertN]

In general strategy games of this type have far too many variables for the AI to actually work well with.
My opinion always stand that these games are player vs player games, and the AI is just serving the purpose for learning the ropes and to tutor a new player.

That is also proved by the fact that AI levels never make the AI smarter, simply grant the AI superpowers that a player normally would not have. (More resources, added combat modifiers, quicker production, free units, etc).

[AlvaroSousa]

A.I. doesn't have the complexity of humans. But giving the A.I. more... whatever... to make it challenging certainly brings a form of entertainment and achievement.

If anyone beats WarPlan at Ubermensch Optimal level I'd be incredibly impressed.

[OxfordGuy3]

In SC3 I've found the AI to be pretty good, considering the type of game that it is - one thing that I've found helps to make it a bit more challenging when playing vs the AI is to try to act as if it was a human player that I'm playing against, in terms of how careful I am, and not exploiting AI weaknesses deliberately (e.g. espc. in the naval aspect of the game).

[michaelincol]

Is AI on par with SC AI is a buy or no buy for me who only plays solo and is happy with AI on all SC games?

[AlvaroSousa]

I designed AoC and AoD for SC2. I used SC's A.I. model for WarPlan. Hubert helped solidify the ideas.

If you measure everything together they are equal IMO. Each has their own tricks and this is dependent on the game mechanics. WarPlan's A.I. was harder to construct considering the multi-hex attack and naval stacking.

[michaelincol]

Thx for quick reply and all your hard work. I will buy as soon as released.

[OxfordGuy3]

ORIGINAL: michaelincol

Thx for quick reply and all your hard work. I will buy as soon as released.

Same here, that does sound promising about the AI. I hope the AI is reviewed again post-launch, though, after more people have used it, to see if any further improvements can be made.