Best AI?

[Septimianus]

A quick poll. What is the best Battle AI in your opinion and why?:

- MTW vanilla
- MTW + XL Mod

- RTW vanilla
- RTW + BI
- RTW + Alexander
- RTW + EB
- RTW + BI + EB
- RTW + Alexander + EB

- MTW 2

I know that they are not state of the art software but my opinion is that MTW + XL Mod is the best and RTW + Alexander + EB the second one.

My feeling is that RTW AI is less complex than MTW one, for example, the way tired units are managed, if you wait in RTW they are again fresh and ready and in general I win a lot more battles in RTW + Alexander + EB than in MTW + XL Mod.

[General Appo]

The only reason the MTW AI appears smarter is because it´s the RTW AI that is a lot more complex. A lot more.
Battles are much faster in RTW, giving less chance for AI manuevers, the way the strategic map and everything around it works is a lot more complex in RTW. The truth is that the RTW is a lot smarter than the MTW one, but the MTW AI has a lot easier tasks to accomplice, and so appears smarter.

[Septimianus]

The only reason the MTW AI appears smarter is because it´s the RTW AI that is a lot more complex. A lot more.
Battles are much faster in RTW, giving less chance for AI manuevers, the way the strategic map and everything around it works is a lot more complex in RTW. The truth is that the RTW is a lot smarter than the MTW one, but the MTW AI has a lot easier tasks to accomplice, and so appears smarter.

I am talking about the Battle AI. I agree with you that the strategic AI in RTW is more complex because it must manage regions and paths to invade them but the Battle AI is more or less the same, two armies and a battlefield (sieges in RTW I agree that are more complex). The speed of EB (not RTW) battles are the same as the MTW ones.

[Kromulan]

Hmm. . . Battle AI. . .

All I can give you is my anecdotal evidence, which goes like this:

I got pretty good at Shogun, then I kicked the crap out of MTW, then I never lost a battle (the ones I fought) in RTW.

I went back to playing some Shogun and got my butt handed to me using my sloppy, new RTW tactics. Re-learned how to beat Shogun and now I'm winning battles in EB that I would likely struggle with in Shogun or MTW.

Interestingly, I found that M2TW reversed the trend a little bit. I actually lost several battles I would never have fought in MTW or Shogun, but that I would likely have won in RTW.

I'm not sure if you could call the AI smarter, but my experience has been that, in order of battle difficulty, they rank:
Shogun
MTW
M2TW
RTW

[Lysimachos]

One reason for Shogun being quite hard might be the very high grade of specialization of the unit, even more than in any other of the games.

But I also think the AI is better in some way in the older games of the series (STW and MTW) than in RTW, because I remember the AI using maneuvers that I haven't seen in RTW, at least not with rtw.exe.
An example is this: My army is slightly stronger than the computer's. At the beginning we both try to reach the favourable position in the center of the map. I'm faster and what does the AI? It retreats!
You might say: So what? especially because a lost battle in MTW or STW automatically means the loss of a province for a defender, but from a point of view of a commander it is a wise decision to abort a battle you probably can't win.

I think the battles in RTW are more action-oriented than in the predecessors. I recall reading something I don't really understand about the complexity of battle physics in the different games which underlines this hypothesis (I'm not starting to search, course I won't find it ).

[Gleemonex]

Some good points made here. I should also note that the RTW engine (I haven't played M2 yet) obviously wasn't designed with set-piece battles in mind, which makes the battle situation more complex for better or worse.

-Glee

[gamegeek2]

I think M2TW AI is the best overall. First of all, its battle maneuvers actually make sense most of the time. Second of all, M2TW has even more tasks than RTW (religion, merchants). Third, its armies tend to be well-made, with most 10+ stacks featuring a good balance of units (exception: Mongols, but that's completely historical).

[Lysander13]

I'm not sure if you could call the AI smarter, but my experience has been that, in order of battle difficulty, they rank:
Shogun
MTW
M2TW
RTW

My opinion would be that you have the right of it. There is no way the "Battle AI" is better in RTW & MTW2 than it's predecessors. I would agree with your rankings. Though MTW2 and RTW are about equal as far as I'm concerned. I'd give MTW2 the edge by a hair, only because the Battle AI appears to be a bit more moddable so to speak.

[gamegeek2]

For battle, Shogun is definitely the hardest, followed by MTW. M2TW Battle AI is more moddable, AND it seems to be smarter; its battle plans make more sense much of the time (like not sending their heavy infantry forward until their archers are done). Then again, it sometimes sits around and takes arrows when it has no archers, but usually it's smarter than RTW.

[Gleemonex]

There is no way the "Battle AI" is better in RTW & MTW2 than it's predecessors.

Well, I wouldn't necessarily use the term "better". As mentionned above, STW's AI was much simpler, as it only had to deal with a dozen or so unit types. It performs better than newer versions because the newer ones are just so complex! It would be like saying a sportsbike is better than a station wagon. Well, try strapping a kayak to the roof of your sportsbike ;)

-Glee

[Ludens]

Well, I wouldn't necessarily use the term "better". As mentionned above, STW's AI was much simpler, as it only had to deal with a dozen or so unit types. It performs better then newer versions because the newer ones are just so complex!

S:TW and M:TW's A.I. understood height advantage, missile fire, could set and search for ambushes. None of this applies to the R:TW A.I. Yes, the S:TW A.I. had a simpler job comparing units, but this does not apply to M:TW, and M:TW's A.I. still performed better than R:TW's. My guess is that CA tried to reduce system requirements by including less A.I. routines. Increased battle speed probably also played a role. Or they may simply have wanted to make the game easier and more accessible.

[bovi]

Well, try strapping a kayak to the roof of your sportsbike ;)

I think I prefer a couch (no, the Norwegian there is not me).

Anyway, as far as battle AI goes, I can't see much difference between RTW and M2TW, except that battles are harder to win by tactics in M2TW; they are over so quickly that you don't have much time for maneuvering, so you need superior numbers or troop quality to win.

Shogun and MTW had much better battle AI. Particularly when the AI was defending a bridge or hilltop in Shogun it was hard to overcome it. Possibly it was just well made for the rather limited set of battlefields, but the AI used the terrain to a degree.

[Ludens]

Shogun and MTW had much better battle AI. Particularly when the AI was defending a bridge or hilltop in Shogun it was hard to overcome it. Possibly it was just well made for the rather limited set of battlefields, but the AI used the terrain to a degree.

I doubt it, since both S:TW and M:TW contained a lot of maps, and custom maps could be made. The maps were also more varied than the R:TW maps, which tend to be dominated by a single feature.

[Gleemonex]

My guess is that CA tried to reduce system requirements by including less A.I. routines. Increased battle speed probably also played a role. Or they may simply have wanted to make the game easier and more accessible.

It's hard (but not impossible) to imagine them skimping on the feature they touted so proudly in S:TW -- an engine inspired by Sun Tzu! -- when most everything else going from S:TW to R:TW was a major revamp.

I think they wanted to make it more accessible. Especially since they included ways to trick the AI ('loose formation' units looking more numerous, for example) piling the natural AI disadvantage with extra, purposely-programmed "stupidity".

The S:TW engine didn't have such debilitating nuances, so I still posit that it had a simpler task overall.

-Glee

[We shall fwee...Wodewick]

I think they wanted to make it more accessible. Especially since they included ways to trick the AI ('loose formation' units looking more numerous, for example) piling the natural AI disadvantage with extra, purposely-programmed "stupidity".

The S:TW engine didn't have such debilitating nuances, so I still posit that it had a simpler task overall.

-Glee

See I don't mind if that was in RTW or M2TW, but that should be what easy battle difficulty is. I would much rather that Very Hard was realistic morale, but your fighting against a napoleonic Hannibal the Great of Macedon then what you do have is a general with the iq of a chicken and the will to live of a suicidal mental patience on the top of a large building who can miraculously rally his soldiers no matter what.

[Ludens]

It's hard (but not impossible) to imagine them skimping on the feature they touted so proudly in S:TW -- an engine inspired by Sun Tzu! -- when most everything else going from S:TW to R:TW was a major revamp.

I agree, but really, if you study the way the R:TW A.I. responds and compare it with the way the M:TW A.I. does things under similar conditions, you cannot but conclude that something has gone very wrong. Off course, because CA completely redid the R:TW engine, they probably had to rewrite the A.I. as well. Game A.I. doesn't necessarily work like the human player: its interface is highly dependent on the engine itself.

I think they wanted to make it more accessible. Especially since they included ways to trick the AI ('loose formation' units looking more numerous, for example) piling the natural AI disadvantage with extra, purposely-programmed "stupidity".

The S:TW engine didn't have such debilitating nuances, so I still posit that it had a simpler task overall.

I rather doubt that was implemented. A more important factor would have been that more CPU power was allocated to the graphics. Also the increased speed of movement and battle resolution in vanilla R:TW would have left the CPU with even fewer CPU cycles to formulate tactics. If the S:TW A.I. responded a bit sluggishly, you probably wouldn't notice and it wouldn't matter much because battle resolution is slower and quick strikes count for less. In vanilla R:TW quick strikes are all-important, however.

[Gleemonex]

See I don't mind if that was in RTW or M2TW, but that should be what easy battle difficulty is. I would much rather that Very Hard was realistic morale, but your fighting against a napoleonic Hannibal the Great of Macedon then what you do have is a general with the iq of a chicken and the will to live of a suicidal mental patience on the top of a large building who can miraculously rally his soldiers no matter what.

Very few people would be as happy as I, if (when?) that were to come about. But that's delving deep into neural net and fuzzy logic research. I can only hope that some budding PhD out there is a rabid EB fan.

It's hard (but not impossible) to imagine them skimping on the feature they touted so proudly in S:TW -- an engine inspired by Sun Tzu! -- when most everything else going from S:TW to R:TW was a major revamp.

I agree, but really, if you study the way the R:TW A.I. responds and compare it with the way the M:TW A.I. does things under similar conditions, you cannot but conclude that something has gone very wrong. Off course, because CA completely redid the R:TW engine, they probably had to rewrite the A.I. as well. Game A.I. doesn't necessarily work like the human player: its interface is highly dependent on the engine itself.

I never got to play much M:TW -- I had a CTD three years after getting my scrumptious and sorely-missed Ghazi infantry :(

I think they wanted to make it more accessible. Especially since they included ways to trick the AI ('loose formation' units looking more numerous, for example) piling the natural AI disadvantage with extra, purposely-programmed "stupidity".

I rather doubt that was implemented. A more important factor would have been that more CPU power was allocated to the graphics. Also the increased speed of movement and battle resolution in vanilla R:TW would have left the CPU with even fewer CPU cycles to formulate tactics. If the S:TW A.I. responded a bit sluggishly, you probably wouldn't notice and it wouldn't matter much because battle resolution is slower and quick strikes count for less. In vanilla R:TW quick strikes are all-important, however.

I don't know the specifics of the GPU/CPU performance breakdown for R:TW, but I suspect that the battle AI was given fewer and sparser time slices of CPU time. Maybe that counts as 'simplicity'.

-Glee

[Ludens]

I don't know the specifics of the GPU/CPU performance breakdown for R:TW, but I suspect that the battle AI was given fewer and sparser time slices of CPU time. Maybe that counts as 'simplicity'.

That is most likely. IIRC a stated goal of R:TW was to keep the system requirements similar to M:TW. Since it obviously had a heavier graphical engine, something else had to be scaled down.