Brad Wardell's views about technology, politics, religion, world affairs, and all sorts of politically incorrect topics.
Published on February 21, 2012 By Draginol In FE Sneak Peeks

I’m one of those people who have traditionally disliked tactical battles because I hate having to fight all of them out.

In FE, my goal is to make it so that the AI does a pretty good job of fighting the battle for you. You’ll still want to fight it out if it’s a close thing or if there are particular units you want to spare. But overall, it does a decent job.

Right now, I’m still toying with the concept of having units retreat from battle to let others take their place. I’m not sure if that’s a good thing or not in *this* game given the game mechanics. 

Here’s a battle where my side lost but in 0.77 would have been a walk over because the tactical AI sucked so badly.

http://screencast.com/t/Uj0Wp0DAU

Also, fun tactical battles:

image

 

[Warning: Explicit]

image

One of the big changes in FE beta 2 that wasn’t in GalCiv II was the concept of army groups. 

Who here is a WW2 history nut? If so, you know what army groups are.

Well, I put in army groups to fix a different flaw in the AI but I didn’t recognize the issue they created. The AI is now really good at having armies from different cities converge on a single city on the same turn (the AI treats army groups as a single unit).

I’ll await feedback but I’m thinking normal is too hard now.


Comments (Page 3)
5 Pages1 2 3 4 5 
on Feb 21, 2012

<Munching popcorn until Thursday >

on Feb 22, 2012

Frogboy
I would have won if I had rally points and could coordinate like that.

You may consider making this a tech or upgrade.

An armies ability to coordinate movement and communicate is extremely vital even in today's age of instant communication let alone the pseudo medieval world of Elemental.

One thing that sticks in my mind from reading Thomas F. Madden's New Concise History of the Crusades was how much time, men, and treasure was lost by armies sitting around waiting for other armies to show up (sometimes many months).  An army could be delayed for so long that their fellows would continue on without them to disastrous results.

Since being able to precisely coordinate multiple groups of soldiers is a pretty big deal in both real life and in game terms you may consider making it a tech or upgrade that must be earned instead of a "freebie".

 

on Feb 22, 2012

I don't think UI features should be techs.

 

One idea: have three difficulty settings

 

World difficulty- for monster difficulty

Strategic difficulty- for strategic AI

Tactical difficulty- for tactical AI

 

This way, if folks find the tactical AI unfun, they can tone it down, and if they like it, they can keep it.

 

 

on Feb 22, 2012

Alstein
I don't think UI features should be techs.

 

One idea: have three difficulty settings

 

World difficulty- for monster difficulty

Strategic difficulty- for strategic AI

Tactical difficulty- for tactical AI

 

This way, if folks find the tactical AI unfun, they can tone it down, and if they like it, they can keep it.

 

 

Ya, great idea if possible!

on Feb 22, 2012

Having different settings for Tactical and Strategic AI  .... an interesting concept.

Personally I don't care, but I can see how some might.

(the only types of bonuses I don't like for difficulty level is 'extra levels' for the AI units ... if they earn the levels then fine. But one thing about FFH2 deity is the free levels making it a bit unfun for me ... >.>  I don't mind resource bonuses for max-level difficulty, but level resources I don't like, especially in a fantasy setting such as this)

Eventually I may want to play with various AI settings, depending on what type of game I'd want to play. But for now, I'd rather just play with the Max level AI on all fronts.

on Feb 22, 2012
Seeing is believing. Please don't nerf the AI before beta 2. At least leave it full on challenging.
on Feb 22, 2012

Oh look, smack talk. Cool!

 

Frogboy
They always did.


A human being wouldn't be able to coordinate 4 different armies arriving at a target city in the same turn.

 

If the human being can't do that, he should stop playing strategy games. Hahahahahaha. I assume you are joking and just showing off features, because if you're going to tone down something that is clearly GREAT combat behaviour, I will be unhappy. Very. Unhappy.

 

Frogboy
My thoughts exactly.

Here's the thing I'm bitching about:

I've been able to write up all kinds of sick stuff for the AI to do that the human can't realistically do because there's no UI.

I've written up rally points, army groups (you guys deal with armies one unit at a time, I look at the map in terms of GROUPS of armies).

Even tactically you humans are pathetic.  You never look at the big picture. You worry about saving a precious unit here or maximizing a unit's HP there or what have you without looking at all your units all over and what's best globally.

And don't even get me started on your money handling or mana handling or experience.  

The AI will always have a better level of experience than you guys because you guys will insist on running around with several champions at once. You're too lazy to just send a single champion to take out a single spider or have a single champion move out of an army to take out a lair. 

The "l33t" will still beat the AI on normal and even challenging but I'll love to hear how they do it. Odds are, it'll be through a loophole. I'm still having a hard time getting the AI to counterspell spells.

 

(please substitute "you" with "AI" here)

Now you are being downright condensing and insulting. I guess you are trying to push buttons. I can beat the snot out of your fantasy AI and army groups any day. I don't need loopholes. I can, and will, coordinate mass attacks from five different fronts, cast a landbridge in your face and walk an army over it and smack your entire empire with spells when I assault, whilst simultaniously razing your assests with 20 single unit horse-archer stacks... and all my armies are backed up by spellcasters who are decked out with +Army buffs, shock troops, archers and tanky guys.  And then I'll  scorch the earth of anything I am not interested in, take over the two biggest cities and start rebuilding your empire before you've even had the chance to blink and surrender. Preferably with three enemy empires at the same time.

In other words: Please, give me that AI. Just make a difficulty setting named  "WantToDie?" with that army grouping, all those lovely routines and all that ass-whooping power, and make me giggle.

 -----------------------------------------------------------------------------------------------------------------------------------------------

 

 Therefore Melkor said to her: 'Do as I bid; and if thou hunger still when all is done, then I will give thee whatsoever thy lust may demand. Yeah, with both hands.

Time passes, the Trees falter and the Silmarils are taken.

 'Blackheart!' she said. 'I have done thy bidding. But I hunger still.'

What wouldst thou have more?' said Morgoth. 'Dost thou desire all the world for thy belly? I did not vow to give thee that. I am its Lord.'

'Not so much,' said Ungoliant. 'But thou hast promised me whatever I hunger for, and my Master, Baranor, Hungers for the World, and his hunger is mine.'

on Feb 22, 2012

Frogboy, can you not nerf the new high-functioning AI before players get a hold of it, to see how WE play against it?

on Feb 22, 2012

Frogboy

A human being wouldn't be able to coordinate 4 different armies arriving at a target city in the same turn.

I think... you have been working too hard. You and Tsunke need to share your ritalin prescriptions. Anyway... If the problem is that the human can't co-ordinate it.. why not put timers above the marching army that indicates the number of turns the unit will take to reach it's destination, or even more simple, list how long it will take the unit to reach the destination before you click on the square and when you reclick on that army it will show you the destination and the turns it will take again.

I'm pretty sure civ has a function identical to this already. Unless I am misunderstanding what you mean by this.

on Feb 22, 2012

Brad, please give us the best AI you've got for us to test it before you consider toning it down!

on Feb 22, 2012


I suggest not letting units retreat. 

Allow reinforcements but only when units are killed. 

Let massive battles be costly to all involved.

 

on Feb 22, 2012


Make a hardcore mode and we will whip the frog anyways

on Feb 22, 2012

Don't be afraid to release the beast!

on Feb 22, 2012

Don't worry. I won't be nerving the AI.  I was just enjoying seeing "it all come together" finally.

The really good strategy game players will be able to cream it still, I have no doubt about that.

But it's definitely playing much better than in 0.77.  And I'm not even close to running out of things to do to improve it.  Every time I play it I still see things that can be improved on.

It's still a challenge to determine when it's a good idea to have a lone sovereign (no units with it) go and wreak havoc versus waiting for other units to join it.  I know as a player the conditions but expressing it programmatically that isn't a hard coded script is a challenge.

The big difference I'm seeing here is the coordination of armies. It's the one thing humans tend to be pretty good about -- they'll send 5 or 6 armies together as a force.  AIs tend to do death trains (1 army at a time).  Having those 4 or 5 armies showing up together from different locations is a big deal.

The goal of normal is for it to be able to beat an average player 50% of the time.  Challenging should beat an average player 75% of the time.  I suspect most people in this forum aren't average players, so we have "Hard" and "Ridiculous" for them.  

With the performance improvements we've been making and with some time, I can code up more sophisticated analysis functions that are CPU expensive but would be really helpful.  This would let us put in more complex tactical battle maps (more terrain modifiers).

CPU time is the bane of AI programming.  As is, the AI spends a lot of time trying to keep its units from getting killed by the monsters. There are certain tricks for getting by monsters (they will attack weaker units that roam near them but it's not fully random and eventually players will see the patterns that will help them increase their odds of survival). The AI also spends a lot of time deciding who and when to upgrade in the shop (And what to buy). There's time spent figuring out how to optimize experience. 

On the other hand, with more CPU time I can still vastly improve the AI from where it is now.  The AI is still awful at trading equipment with other champions or selling items.  The strategic spell casting isn't nearly as sophisticated as the tactical AI spell casting. The diplomacy stuff can be made a lot more sophisticated. 

The #1 thing I still need are units to choose from.  The AI in beta 2 won't use units you designed  for other factions. It will use the units you designed for the particular faction it happens to be. That is' if you were playing as Magnar and made a bunch of units for him, in a future game the AI will use those units.  THe more units it has to choose from, the better it'll get.  Otherwise, you still see a lot of peons because it doesn't' design units.  Derek is adding in more units all the time but this is an area we can still do a lot more.

Sorry for the 4am wall of text guys.

 

on Feb 22, 2012

I don't get it.  Has the Stardock forums times in posts actually reverted to Stardock time as opposed to the time in your timezone?

5 Pages1 2 3 4 5