Results 1 to 16 of 16

Thread: Battles in Attila TW lack challenge (BAI analysis)

Hybrid View

Previous Post Previous Post   Next Post Next Post
  1. #1

    Default Battles in Attila TW lack challenge (BAI analysis)

    Hi, everyone! Some of you might remember a similar thread I made for Rome II where I compared battles in R2 and M2 noting that I was unsure as to why exactly battles were so easy in R2. This time I won't compare battles of Attila to other games, but I will do some analysis on very specific issues with BAI. Just to be clear, since there was a bit of confusions in the previous thread, I'm not trying to bash this game in any way whatsoever, in fact I like this game a lot and think it has a lot of potential for greatness. For me personally the AI is the main issue keeping me from fully enjoying this otherwise amazing game.

    I will discuss only open field battles in this segment, touching only the very basics of combat. All battles fought are with identical troops, no mods, normal difficulty, plains map.

    1. Clumping infantry.

    Any TW veteran or indeed any player who has at least played a campaign or two knows that the best way to engage enemy infantry is in a single straight line. The worst thing you can do is bunch up all your units and charge into the center of the enemy line, thus letting your soldiers become flanked and surrounded. This is and has always been one of the most basic rules of combat in TW. There are two levels of clumping: macro - where multiple units attack the center of the enemy line and get surrounded by the collapsing wings, and micro - when a unit has a smaller rank depth but greater width than the opposing unit which will become partially or completely surrounded when the two units clash. The AI in Attila is plagued by both of these issues.

    In the below video you will see this in action. Our units have longer ranks which is why we completely surround the enemy units upon clash. This means they take increased casualties due to fighting on multiple fronts, netting us an advantage where it should not exist. A depth of 3-4 ranks (on Smallest unit scale) is enough to surround the enemy who is 8 ranks deep and not worry about the enemy breaking through our center. A battle that should have gone down to the wire ends up being a Decisive Victory. I will note though that this issue is less painful on Large and Huge unit scales due to the gaps between AI units being smaller relative to unit size. The outer flanks still become significantly surrounded though.
    Spoiler Alert, click show to read: 

    What needs to be done:
    The AI must stop having gaps between its infantry units. The AI must stop using such unnecessarily deep ranks. The AI must stop randomly disengaging their infantry (losing men in the process) only to charge back in again after a short while (unrelated but shown in the video).



    2. Focusing missile fire.

    Often in TW Attila focusing all your missile fire onto a single unit is more effective than letting each unit target a different enemy unit. If there's enough firepower, this will let you quickly rout a single unit giving you immediate advantage. The AI almost never acknowledges this.

    In the video below you will see all our missile units target separate enemy units initially to break their "force fields" that all units in Attila have. Once they become vulnerable, we target a single unit with all our missiles. Moments later it is brought down to 7 men and routs. Rinse and repeat and a few left-clicks later all enemy missile units are destroyed or routed while our own units have suffered only minor casualties. Again, a battle that should have been close ends up being an easy win. Results might differ depending on what type of units you're using, the higher the missile damage, the better this will work.
    Spoiler Alert, click show to read: 

    What needs to be done:
    The AI should prioritize using several missile units on a single target. If the player has almost exclusively missile units, the AI should start the battle in loose formation by default. If the circumstances allow, the AI missiles should prioritze units that are NOT in Defensive Testudo or Loose formation. The AI should be less willing to friendly fire when there are targets that are not engaged in melee. (Latter two are unrelated but important)




    3. Missile cavalry - when to prioritize fighting over skirmishing.

    Cavalry - a very mobile and deadly force. Great at flanking enemy lines, performing repeated charges and running down routing enemies, but not so great in prolonged melee especially when surrounded or outnumbered. The AI in Attila more often than not knows how to use its cavalry when you have none yourself, you will get flanked almost every time. However, if you have equal numbers of different types of cavalry, the picture becomes different. The AI is scripted to keep its horse archers at bay whenever possible which essentially means they will try to avoid melee even if it means losing the battle.

    In the below video you can see this happen from the very start of the fight. At 00:20 the AI sends one of its cavalry units into two of ours who are backed up by a third moments later. On the other side, another of the AI's units has charged into two of ours as well. All the while the two AI cavalry archers are hanging back shooting toothpicks at us. At 00:55 both surrounded units are destroyed, all our units are free to do whatever they want with the now outnumbered enemy. Now and only now does the AI decide to send one of its horse archers to melee us, we counter-charge with two of our own and enter a prolonged fight during which the AI's second horse archer unit stands almost next to their surrounded comrades doing basically nothing. Yet another Decisive victory made possible by the AI's terrible decision making.
    Spoiler Alert, click show to read: 

    What needs to be done:
    The AI missile cavalry must know when to prioritize melee, especially when the AI's other units are being outnumbered or flanked, and especially when the player's missile cavalry has already engaged in melee.




    4. The big picture.

    All of this is very bare-bones and captured in a very controlled environment. However, I can prove that all of the above is very true in any battle. Below is a video with the last custom battle. It has all the variables thrown together - infantry, missile infantry, cavalry and missile cavalry. At 00:40 you can see the AI throwing its cavalry into a fight they can't win again, while their missile cavalry hangs at the back. At the same time, you can see how long our infantry line is compared to whatever that clump is that the AI is using. Notice at 01:00 we issue all the attack orders for the infantry and have two units left over on either side that can do free flanking. We have the same number of infantry as the AI, yet we're able to flank them... with infantry. 01:15 - AI's general completely ignores two spear units running at it and gets caught. He continues to fight outnumbered against our cav and spears for the rest of the fight, not even trying to disengage. By 01:35 three of their outnumbered cavalry units have been destroyed, finally one of their horse archer units decides to charge into our cavalry... nope, they stop 3 feet from us and continue to fire. 2:00 - our focused fire has broken one of their foot missile units, another one breaks at 2:21, shortly followed by a third one. I'm not sure what the enemy archers were targeting but our archers haven't incurred a single casualty. This means the AI did a lot of friendly fire because whatever they fired at was fighting in melee. Not that it matters in the end, the whole battle was just a formality after the AI charged its cavalry into our greater numbers and got annihilated.
    Spoiler Alert, click show to read: 


    That'll be it for now. This might get updated with more analysis at a later time. For now feel free to comment away.
    Last edited by Neige; March 30, 2015 at 04:19 PM.

  2. #2

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    Quote Originally Posted by Neige Noire View Post
    What needs to be done:
    The AI must stop having gaps between its infantry units. The AI must stop using such unnecessarily deep ranks. The AI must stop randomly disengaging their infantry (losing men in the process) only to charge back in again after a short while (unrelated but shown in the video).
    No, this tactic should work, it's an historic tactic used by generals on the battle field since classical times (it allowed the sides of the infantery on the front line to flank engaging infantery), the engine should be fixes so as this tactic actually could work!

  3. #3

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    Quote Originally Posted by Generaal Van Heutsz View Post
    No, this tactic should work, it's an historic tactic used by generals on the battle field since classical times (it allowed the sides of the infantery on the front line to flank engaging infantery), the engine should be fixes so as this tactic actually could work!
    Actually I doubt this one. I believe this be to a mistaken view about the so call "wedge" formation. The Romans normally formed a Century to 20 frontage and 4 deep. Typically behind this was a second line of the same. However in the case of an all out attack the Century formed to 10 frontage and 8 deep, this allowed the units from the rear line to form up at the front alongside the others. Such a formation would be visually deeper than its frontage, hence the term "wedge".
    Proculus: Divine Caesar, PLEASE! What have I done? Why am I here?
    Caligula: Treason!
    Proculus: Treason? I have always been loyal to you!
    Caligula: [laughs insanely] That IS your treason! You're an honest man, Proculus, which means a bad Roman! Therefore, you are a traitor! Logical, hmm? Ha, ha, ha!

  4. #4

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    It's ironic in any case because I'm pretty sure there's a loading screen that says having gaps in your infantry line is very bad lol.

  5. #5

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    Interesting thread, wish CA would actually respond to this...

  6. #6
    Humble Warrior's Avatar Vicarius Provinciae
    Join Date
    Oct 2005
    Location
    Great Britain.
    Posts
    11,147

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    Some interesting points there. I certainly find that focusing all missiles on one target is effective.

  7. #7

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    Interesting analysis and good background information for each issue.

    On the subject of clumping infantry, I'd like to point out or rather add a few things. It's true that the AI has a tendency to mass infantry at certain points in the line - often the center - although it's not always the case. This by itself is a problem, since only so many troops can engage one another at any one given time of combat. Essentially, clumping together troops like that is just a waste of reserves. It exhausts more troops quicker than necessary and doesn't do very much to quicken the combat. More importantly, it leaves the soldiers in a very exposed position as they can easily be flanked. The only particular advantage by this approach is the casualty distribution between units, which is of limited help if they rout.

    Another part of the problem is really the fact that there's a notable difference between the player and AI behaviour of organizing troops and formations. The player will often suffice with 3, 4, possibly 5 men deep units. The AI, on the other hand, isn't able to change the depth at all and will therefore always use the default depth of a unit. Hence, seen per unit, fewer men can actively participate in combat unless the unit gets flanked or enveloped by a thinner opposing unit.

    As of late, the AI has seemingly become better at keeping reserves, at least from what I've seen. There is more to be done, as always, though. Since the issues are not one but multiple, I think several measures to counter the issues need to be taken as well. Part of the solution lies within the AI behaviour, it does need to maintain formation integrity seen from the wider perspective. Carelessly throwing in a number of units here and there against the enemy formation will not suffice for victory, more often than not the disadvantages outweight the advantages. This would supposedly be the more complex part of the solution given the nature of AI coding. The other part is really more simple - thinner lines per default. Four or five ranks deep melee units - which may well not be particularly historical - will on the other hand even out the differences between player and AI. Since most units have a minimum depth of around two/three ranks, the difference is much smaller than if the AI would use eight ranks.

    Then of course there should actually be some advantage of using deeper formations; cohesion, mass and morale come to mind. As it stands now these have a limited correlation to unit depth depending on scenario.


    I will respond to the points on missile fire as well. I think your intentions here are good, but the solution not what it might be. This discussion refers us to the damage and health systems in the game. Those are the issues, not so much the AI's behaviour compared to the player. The whole balance/design concept here is unnatural. Damages should be reasonably consistent given circumstances don't change, but that's not a scenario the game as of now will ever present to us. The only good solution, as I see it, is to rework the whole system and try to bring back consistency to the degree possible. This wasn't a problem in Shogun 2, and the reason is because it had a completely different base combat design.
    Last edited by Sheridan; March 31, 2015 at 03:58 AM.
    Campaign modder for Ancient Empires


  8. #8

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    I agree, the health system is very unnatural and off-putting, I don't understand its purpose at all. Hopefully the devs will drop it at some point.

  9. #9

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    Quote Originally Posted by Neige Noire View Post
    What needs to be done:
    The AI missile cavalry must know when to prioritize melee, especially when the AI's other units are being outnumbered or flanked, and especially when the player's missile cavalry has already engaged in melee.
    This one is hard to judge even for human player. Need some hard rules if you want the AI to deal with this. Suggestions?

  10. #10

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    I know nothing about AI programming but I assume the AI should be able to calculate when it's going to lose a melee fight unless they send in backup? In any case, missile cav is too passive in this game, they act like regular foot archers. They should use their mobility and try to be as annoying as possible, instead they just passively stand behind their lines and fire. If the player has only infantry units, the AI immediately sends two of its missile cavalry on a flanking charge, which is excellent. But the timing is usually off, they should first wait for their own infantry to engage the player's and then flank. All of this is very easy to notice once you do a little bit of testing, something that I wish the devs would do a bit more.
    Last edited by Neige; March 31, 2015 at 06:53 AM.

  11. #11
    Campidoctor
    Join Date
    May 2010
    Location
    Danmark
    Posts
    1,507

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    In the video below you will see all our missile units target separate enemy units initially to break their "force fields" that all units in Attila have.
    Units have force fields now? That's ridiculous. What happened to simply tracking each arrow to the target and having a chance for it to kill the guy?
    The game development business is one of bottomless greed, pitiless cruelty, venal treachery, rampant competition, low politics and boundless personal ambition. New game series are rising, and others are starting their long slide into obscurity and defeat.

  12. #12
    Senator
    Join Date
    Mar 2009
    Location
    Tulifurdum
    Posts
    1,317

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    Nice analysis. Some problems however are artificial I think, like unit line length. If you know the problem why do you not use the same unit depth as the AI? Why using very thin lines and complaining that the AI cannot use similar unhistorical lines?

    What do you mean with "force fields"? Is it the stupid fact that damage first is absorbed by the hp?
    Last edited by geala; April 01, 2015 at 01:43 AM.

  13. #13

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    Quote Originally Posted by geala View Post
    Some problems however are artificial I think, like unit line length. If you know the problem why do you not use the same unit depth as the AI?
    Think about it. If one has to adapt one's play style and deliberately handicap oneself in order to even out the playing field for the AI, then the problem is not artificial, "the solution" is.

    Quote Originally Posted by geala View Post
    What do you mean with "force fields"? Is it the stupid fact that damage first is absorbed by the hp?
    I'm not fully familiar with how exactly it works but yes, the health system is the issue from what I've heard and seen.

  14. #14
    Senator
    Join Date
    Mar 2009
    Location
    Tulifurdum
    Posts
    1,317

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    I thought a little bit about it. It's not the dump AI, it's the player cheating using too thin a line to be practical in reality. Of course you can say why did CA not implement disadvantages for thin lines or why isn't there a command like "always copy player unit line lengths". But although you can blame CA (as usual) you can also easily solve the problem by using the given unit depths. Or change it generally like in the WOFE mod. Otherwise it's a bit like R2TW players complaining about the soo easy battles they have with their all elite units armies begging for AI changes instead of using more naturally composed armies themselves.

    There are of course some BAI problems for sure, f.e. when the AI attacks only the center of a long line of several units. Don't know wether it still happens in ATW, I have yet to start a campaign.

  15. #15

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    I agree it is cheating. It's cheating that should not be possible but it is. That's all I'm saying, I'm not blaming anyone like you say, I'm merely pointing out this issue in hopes that it will be fixed. You think the issue is not the AI? That's fine by me, we can agree to disagree then.

  16. #16

    Default Re: Battles in Attila TW lack challenge (BAI analysis)

    I really enjoyed reading this analysis and agree with many points. The BAI can always be improved.

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •