Tactical AI (and how it breaks).

[Monk]

So. I got ETW today and the first thing I wanted to do was test the AI. It's been promised that it's improved and true enough it was. Discounting the RTI AI which would not even attack me, the Grand campaign AI (on hard) showed early signs of promise. However, after more and more battles I am starting to believe the Ai is not properly coded to handle the new additions to the tactical side of the game.

I'm talking about garisons and cover. The example of which is a recent battle between me (Austria) and Prussia. I pushed on Berlin after Prussia declared war on me, but Prussia intercepted my army. No problem, since I had a lot of artillery and strong line units, i'd make them march into a death trap! Right? Eh..

Prussia immediately deployed in a nice formation, but it didn't last. It soon sent a unit to some cover near by, sent one to garrison a building on the opposite end of the map, as well as sent the rest of its army in a peice meal attack against my line. while sending its cavalry on WIDE flanking maneuvers. Ones that, once in position, never moved. The garrison and cover systems seemed to divide the AI's mind, it couldn't decide whether it should attack me, flank me or make me come to it. It decided to try to do all at once and subsequently was defeated easily.

I wasn't very satisfied with my victory.

I've been experiencing this sort of thing a lot, and whenever cover or garrisons seem to be lacking on the battlefield the AI does amazing! But when they're present, it acts as if it doesn't know how to respond to in regards to them.

[Alexander the Pretty Good]

Probably has the priorities too high in the "goals" system.

[A Very Super Market]

As dissapointing as that may be, its still a step up from the earlier games.

Field battles are fine, right?

[Monk]

Probably has the priorities too high in the "goals" system.

I believe this is indeed the case, yes. Priorities are likely far too high for cover, and the Ai is unsure of how to react when it finds some. Sometimes it ignores it altogether, which can be a good thing, other times it just huddles near it despite its objective (me) being far away.

Field battles tend to be much more cohesive.

[Polemists]

Well I think it sounds like a loose system.

I mean if you were in a small city setting, the wide flanks around and garrisoning buildings in the center would be a legit tactic.

Yet most of the time you are out in the open, during a field battle


The odd time you run into a building on end of the map though, the Ai seems to say, Garrison priority number 1.

Nothing's perfect, i'm sure Ai fixes will be in patches. It's not a life killer.

[Monk]

Well I think it sounds like a loose system.

I mean if you were in a small city setting, the wide flanks around and garrisoning buildings in the center would be a legit tactic.

Yet most of the time you are out in the open, during a field battle


The odd time you run into a building on end of the map though, the Ai seems to say, Garrison priority number 1.

Nothing's perfect, i'm sure Ai fixes will be in patches. It's not a life killer.

Buildings and stone-wall fences (cover) on your field maps are more likely than you think. I wasn't suggesting it was a death knell, but its something that needs to be addressed.

It's not something that cannot be patched or modded and may be as simple as a priority listing.

[AussieGiant]

Hi Monk,

From your description it does seem these "New" tactical components are causing a few incoherent choices.

Hopefully it's just a question of rebalancing priorities and reaching a few thresholds earlier or later in the decision making process.

It certainly seems like it is "thinking in the right direction" which is a huge thing in itself.

*fingers crossed*

[Monk]

Hi Monk,

From your description it does seem these "New" tactical components are causing a few incoherent choices.

Hopefully it's just a question of rebalancing priorities and reaching a few thresholds earlier or later in the decision making process.

It certainly seems like it is "thinking in the right direction" which is a huge thing in itself.

*fingers crossed*

Indeed. It's good to see it trying to balance numerous tasks, but not when those task contradict each other's purpose. I read on TWC that CA is already going to release a second patch next week that will deal with many graphical improvements as well as a crash fix.

If they keep up this sort of support, I see no reason why an AI fix cannot be implemented sooner rather than later.

[Polemists]

Agreed. I have no idea what the release date patch was, but assuming they will get quite a few of the graphical issues hammered out next week, I would assume the next focus would be AI or something similiar (Since they have a whole ai team now, shouldn't take to long.)

I give CA credit where credit is due, the game may not be perfect but it is nice to see them supporting us with a patch every week. One can only hope this keeps up :)

[AussieGiant]

I'd say we are in good shape guy's.

The thread titled: "Official: Empire - The first 24 hours" has CA committed to ongoing support and patching. I expect the AI to have ongoing time spent on it.

There's nothing quite like 30 000 people play testing for you!!

[Monk]

Hm.

I bumped the AI up to VERY HARD and gave the campaign another go <H/VH>.

Color me impressed that the AI has not once in my 10 turns and subsequent battles displayed the behavior that I saw on Hard. Its very strange. Perhaps whatever perimeters that lead to its behavior are not present on Very hard. Or maybe I have just yet to put it in the right conditions to lobotomize itself. Time will tell!

It even fought me to a grueling standstill in Western Poland by use of formations and clever adaptions. I won, but barely.

[AussieGiant]

Well then, perhaps the difficulty setting changes certain aspects and it fights as perfectly as it can on that setting, rather than making mistakes...which could be a deliberate part of the sliding scale of difficulty.

[Oleander Ardens]

Very good news, Monk.

Perhaps the setting of the difficulty does finally not change stats but the capability of the AI. This would be amazing. It seems more and more that CA did walk the talk and that the team devoted to the AI did do good (or even great) work. I will perhaps buy a need rig to play it.

OA

[TB666]

The AI isn't very good at sieges I have noticed.
It seems to have a hard time understanding how to best scale the walls and kill those inside.
It never goes for the best wall to scale but always the one at the back, which is a good theory in general however since they pretty much stay within firing range the whole time it kinda renders it pointless.

However on field battles is another story where it behaves far better and really makes you work for a victory.
And this is on VH.

[PBI]

How often do sieges happen? Is it just when you build a fort or is it every city you attack? My understanding was that city sieges now are really field battles on the outskirts of the city.

[TB666]

How often do sieges happen? Is it just when you build a fort or is it every city you attack? My understanding was that city sieges now are really field battles on the outskirts of the city.

Well capital cities are indeed field battles with a city in the middle.

[Malkut]

That's funny. I besieged Berlin on Normal, and it reacted entirely differently. They rallied around the fort, and left some garrisons in the town to inconvenience me.

What they did not do, however, was garrison the command HQ building inside the fort. After I breached the wall, I raced a unit of line infantry through, and into the building. They took some losses in the process, but they easily survived the battle.