Can Good AI Be Bad?


Although the central place for design discussion is ##crawl-dev on freenode, some may find it helpful to discuss requests and suggestions here first.

User avatar

Tomb Titivator

Posts: 857

Joined: Monday, 31st January 2011, 23:19

Post Sunday, 24th July 2011, 09:17

Can Good AI Be Bad?

Some enemies purposely have bad AIs. Orc priests and centaurs come to mind but there are others too. If their AI was improved the game would be too hard... just imagine centaurs that kite you and keep firing those branded arrows even when you're in melee range and orc priests that smite spam and never give you a break. What if every orc wizard blinked when you were in melee range and spammed throw frost when you aren't?

To be blunt, the AI improvement for ranged enemies did nothing more than remove tactical options from the player and widened the gap between melee and casters even more. If this was the goal then why not improve the AI for orc priests and centaurs?

I've been playing a lot of melee classes recently and even won with a MDFi. Before I would try to tactically use the environment to get an edge... now my only option is to mindlessly run up to enemies and hope for the best.

I know the change won't be reverted after existing for so long. I guess I just want to ask the point of it all. What was accomplished that wouldn't also be accomplished by improving orc priests and such?

Also, it would be interesting if every enemy played as tactically as the hypothetical orc wizard if melee characters had more options to deal with them like making wands of confusion more common or something or allow the throwing of potions (yes, I know it's never going to be done, but I'm just tossing ideas out there).

Abyss Ambulator

Posts: 1244

Joined: Thursday, 10th March 2011, 19:45

Post Sunday, 24th July 2011, 10:47

Re: Can Good AI Be Bad?

I agree that in a game where monsters "cheat" in so many ways, the developers should be very careful about any AI improvements - monsters don't need to be much more intelligent to be able to completely overwhelm the player with superior speed, unlimited MP, no hunger worries, or just greater HP and weight of numbers.

I am aware that in another major roguelike, when a new "improved" AI was developed it had to be abandoned almost immediately as a large number of monsters became literally impossible to kill. I know that what's true for another game isn't necessarily the case for Crawl, but I'd think that if monsters got significantly more intelligent they would probably have to be weakened in other ways.
User avatar

Dungeon Master

Posts: 4031

Joined: Thursday, 16th December 2010, 20:37

Location: France

Post Sunday, 24th July 2011, 11:09

Re: Can Good AI Be Bad?

We don't want monsters to be too intelligent, but we don't want them to be dumb as bricks either. With the old behaviour, the tactic against ranged AI monsters was extremely simple. Hide behind a corner and wait for them to run to you. Now, you can still ambush them, but it's a bit more tricky and more interesting IMO. You actually have to tactically use the environment, not just find a corner to hide. And the AI improvement applies to all monsters with ranged attacks, orc priests and centaurs included.
I totally agree that monsters using their spells optimally or kiting when you get in melee range wouldn't be fun at all. Don't worry, there's no plan to add that. The ranged AI improvement has made fighting them more interesting. Having monsters use spells optimally would just make them much more dangerous and predictable, but not more interesting.

For your information, here are the AI improvement I plan to do at some point:
  • If a monster has ranged attack and can see you but not fire at you (because of friendly, plants or status in the way), it will try to move into a spot with a line of fire.
  • Monsters will use pathfinding to avoid clouds. The pathfinding algorithm will evaluate damage taken vs path length and try to find a good compromise.
  • If a monster cannot attack you, cannot reach you and is taking damage from you, it will retreat. A behaviour similar to fleeing but which will stop on different conditions.
<+Grunt> You dereference an invalid pointer! Ouch! That really hurt! The game dies...

Bim

Crypt Cleanser

Posts: 700

Joined: Wednesday, 5th January 2011, 15:51

Post Sunday, 24th July 2011, 11:11

Re: Can Good AI Be Bad?

I agree, but I believe that it is just the direction that the devs wanted to take it. For instance, I feel the game is now (rather than a few years ago) a lot more centred towards using items correctly and planning ahead rather than 'on the fly' tactics (using wands instead of corners). This is great in some ways as it makes for a deeper game over the long term, but makes battles, as you say, slightly more simple.

Overall, I'd love to see more battle tactics, terrain types that give bonuses, furniture that can be used to the players dis/advantage and so on, however the dev's have already pretty much thought through a lot of these ideas and there are some good reasons why they don't work.

The main point though, is that a lot of the AI improvements were done to stop people having to use cheating tactics (hiding round corners) for every monster they see. Although some could argue it isn't cheating, it does become very repetitive. Plus, flavour wise, I can imagine centaurs being too proud to run away and beogh not always being the most co-operative :P
2012 Winner of fewest proposed ideas implemented by devs.

Bim

Crypt Cleanser

Posts: 700

Joined: Wednesday, 5th January 2011, 15:51

Post Sunday, 24th July 2011, 11:15

Re: Can Good AI Be Bad?

Galehar, about the retreating thing, I've been playing a lot of merfolk recently, and my only gripe with the current AI is that I can pretty much sit in deep water and pelt stuff with my superior throwing skills/spells to my hearts content. I recently dispatched a whole orc horde (including knights) which would have vastly, vastly overpowered me otherwise. I'm not sure what would be best, because if monsters ran away as soon as they couldn't get to you it would negate a lot of the fun of merfolk, but it also seemed strange that a knight was just stood around whilst I poisoned dated him.
2012 Winner of fewest proposed ideas implemented by devs.
User avatar

Dungeon Master

Posts: 4031

Joined: Thursday, 16th December 2010, 20:37

Location: France

Post Sunday, 24th July 2011, 11:52

Re: Can Good AI Be Bad?

Bim wrote:Galehar, about the retreating thing, I've been playing a lot of merfolk recently, and my only gripe with the current AI is that I can pretty much sit in deep water and pelt stuff with my superior throwing skills/spells to my hearts content. I recently dispatched a whole orc horde (including knights) which would have vastly, vastly overpowered me otherwise. I'm not sure what would be best, because if monsters ran away as soon as they couldn't get to you it would negate a lot of the fun of merfolk, but it also seemed strange that a knight was just stood around whilst I poisoned dated him.

What's your point? You start by explaining how the current behaviour leads to extreme abuse then you're saying it's a lot of fun. Any situation in which you can kill monsters with 0 risk is a problem which should be addressed. What you've described is exactly what I'd like to fix with the retreat behaviour. And it doesn't apply just to merfolk, but also to grey dracs, octopodes and anyone with a reliable source of swimming or levitation. You can abuse it with ranged attacks and reaching.
There is a plan to give reaching to octopodes, so this AI fix should be done before.
<+Grunt> You dereference an invalid pointer! Ouch! That really hurt! The game dies...

Vestibule Violator

Posts: 1567

Joined: Friday, 21st January 2011, 22:56

Post Sunday, 24th July 2011, 13:24

Re: Can Good AI Be Bad?

This might be hard to get right. Example: a char with reaching that can stay in/over water. I don't know the right glyphs because I play tiles, so I'll just use @, ~, g and . for player, water, enemy and land.
  Code:
~@~g.. --> player attacks (reaching), enemy retreats
~@~.g. --> player steps forward, enemy can reach the player and also steps forward
~~@g.. --> Player retreats, monster did not take damage since the last time it tried to close in and so doesn't retreat
~@~g.. --> repeat until the monster is dead

Retreating when an enemy takes damage from something it cant reach, attacking if it can reach the player and waiting if it cant reach the player and did not take damage since the last attack attempt is too simple. Coming up with conditions that avoid at least all of the simple abuses might not be so easy.
User avatar

Dungeon Master

Posts: 4031

Joined: Thursday, 16th December 2010, 20:37

Location: France

Post Sunday, 24th July 2011, 14:21

Re: Can Good AI Be Bad?

Galefury wrote:Retreating when an enemy takes damage from something it cant reach, attacking if it can reach the player and waiting if it cant reach the player and did not take damage since the last attack attempt is too simple. Coming up with conditions that avoid at least all of the simple abuses might not be so easy.

I've already partially implemented it, but didn't finished because there were some more urgent stuff. I just used some randomization to try to avoid those kind of silly dances. First, the condition "is taking damage" is more accurately phrased as "is wounded and damage was from the player". Now the logic to stop retreating is the following:
if we can reach the player and one_chance_in(3), attack.
else, if we're not in LOS and one_chance_in(3), wander.
<+Grunt> You dereference an invalid pointer! Ouch! That really hurt! The game dies...
User avatar

Dungeon Master

Posts: 4031

Joined: Thursday, 16th December 2010, 20:37

Location: France

Post Sunday, 24th July 2011, 14:27

Re: Can Good AI Be Bad?

galehar wrote:Monsters will use pathfinding to avoid clouds. The pathfinding algorithm will evaluate damage taken vs path length and try to find a good compromise.

Also, if a monster is in a dangerous cloud it will try to move out of it if it can do so while still attacking the player.
<+Grunt> You dereference an invalid pointer! Ouch! That really hurt! The game dies...
User avatar

Vestibule Violator

Posts: 1593

Joined: Thursday, 19th May 2011, 16:38

Location: Penza, Russia

Post Sunday, 24th July 2011, 15:39

Re: Can Good AI Be Bad?

Erm, you mean it will actually try to retreat and NOT to continue going through the clouds anymore trying to hit the adventurer?
Right now, a monster already in the clouds just forgets about his own safety and tries to hit the player desperately.
If you find any mistakes or typos in my post, feel free to PM me about it. Thanks in advance!

The Verse flows throughout Aquaria...
Through each ripple and wave...
Through every living being...
The Verse binds us all as one.
User avatar

Dungeon Master

Posts: 4031

Joined: Thursday, 16th December 2010, 20:37

Location: France

Post Sunday, 24th July 2011, 16:28

Re: Can Good AI Be Bad?

MyOtheHedgeFox wrote:Erm, you mean it will actually try to retreat and NOT to continue going through the clouds anymore trying to hit the adventurer?
Right now, a monster already in the clouds just forgets about his own safety and tries to hit the player desperately.

No, I said it will try to move out of the cloud if it can do so while still attacking the player. For example, a stone giant standing in a poisonous cloud and in melee range will look adjacent cells to see if moving into one of them would allow it to get out of the cloud while still being in melee range. If not, it will off course stand still and keep attacking since attacking the player has a higher priority than self survival (at least for intelligent monsters). And something similar can be done for ranged attackers.
<+Grunt> You dereference an invalid pointer! Ouch! That really hurt! The game dies...

For this message the author galehar has received thanks:
MyOtheHedgeFox

Crypt Cleanser

Posts: 718

Joined: Monday, 14th February 2011, 05:35

Post Monday, 25th July 2011, 14:03

Re: Can Good AI Be Bad?

I know my approval is unlikely to decide anything, but the improvement to ranged AI is the change I like the LEAST out of any in the last four releases. It has made the game significantly less fun for me (I do not mean harder or requiring more tactics; I mean less fun). Current popular strategies for pure melee characters include running in circles around a level until you find a place to ambush, finding a door and closing it, or blocking ranged attackers with other monsters. I'm sorry, but going to all this trouble so a centaur doesn't burn all my scrolls is tedious. Then there's the strategy snow has described - I imagine a lot of players new to melee are unable to come up with much beyond 'get shot a lot'.

The use of corners being so prevalent before was probably a symptom of ranged attacks being too strong. For instance, an orc throwing stones actually does more damage than the orc punching you in melee, with higher accuracy and without gdr. If anything, this strength should be lowered. Monster missile damage could be based on something other than melee damage + weapon damage - perhaps on HD.
mikee_ has won 166 times in 396 games (41.92%): 4xDSFi 4xMDFi 3xDDCK 3xDDEE 3xHOPr 2xDDHe 2xDDNe 2xDSBe 2xKeAE 2xMfCr 2xMfSt 2xMiAr 2xMiBe 2xNaTm 1xCeAr 1xCeAs 1xCeBe 1xCeEn 1xCeFE 1xCePa 1xCeTm 1xCeWz 1xDDAs 1xDDCr 1xDDHu 1xDDTm 1xDENe 1xDEWz

Ziggurat Zagger

Posts: 3037

Joined: Sunday, 2nd January 2011, 02:06

Post Monday, 25th July 2011, 14:12

Re: Can Good AI Be Bad?

Maybe you should stop playing pure melee characters? You're supposed to need more than three skills for the entire game, and it was a degenerate situation in the first place that you could clear the entire game by mashing the arrow keys. Try picking up slings, or crossbows, or evocations, or invocations.

Casters are certainly overpowered currently compared to non-casters, but funneling the entire game into a single-file tunnel so your mountain dwarf fighter can axe everything in melee is not a good solution.

Crypt Cleanser

Posts: 718

Joined: Monday, 14th February 2011, 05:35

Post Monday, 25th July 2011, 14:26

Re: Can Good AI Be Bad?

KoboldLord wrote:Maybe you should stop playing pure melee characters?

<Sequell> 163 games for mikee_ (xl>20): 4x DSFi, 4x MDFi, 4x DDEE, 3x DSBe, 3x DDCK, 3x HOPr, 3x KeAE, 2x MfCr, 2x MiBe, 2x OgFE, 2x DDHe, 2x NaTm, 2x MiAr, 2x VpIE, 2x MfSt, 2x DDNe, 2x CeAr, 1x DDAs, 1x HuBe, 1x MiCK, 1x DSHu, 1x KoHu, 1x GhWn, 1x MfPa, 1x HuVM, 1x HuAs, 1x TrAE, 1x HaBe, 1x TrAr, 1x MfBe, 1x DrSu, 1x HOHu, 1x OgCK, 1x HEFi, 1x KeWz, 1x DSGl, 1x HuWr, 1x GhGl, 1x DEWz, 1x DGRe, 1x HECj,
[22:08] <Sequell> 1x KoNe, 1x DrHe, 1x SEDK, 1x CeAs, 1x TrAM, 1x MfPr, 1x DSCr, 1x HEEE, 1x CeTm, 1x TrHe, 1x KeHe, 1x VpAr, 1x CePa, 1x OPAE, 1x DENe, 1x MiSu, 1x TrEn, 1x TrIE, 1x VpSu, 1x SENe, 1x OgPr, 1x KoBe, 1x OpMo, 1x KoTm, 1x SEAE, 1x SEVM, 1x HEWz, 1x HOAE, 1x NaPr, 1x HaFi, 1x MiMo, 1x DDTm, 1x DSWn, 1x MfTh, 1x CeFE, 1x VpWn, 1x SEHe, 1x DSVM, 1x MfCj, 1x DrWz, 1x DGWz, 1x DSCK, 1x HaDK, 1x MuVM, 1x
[22:08] <Sequell> HuCK, 1x MDGl, 1x MfWz, 1x DGCj, 1x HaVM, 1x HEAr, 1x TrPr, 1x MuWn, 1x DrIE, 1x HEGl, 1x GhAs, 1x SpEn, 1x DSMo, 1x DDCr, 1x SETm, 1x DrMo, 1x KeCK, 1x FeDK, 1x KeIE, 1x DrAE, 1x GhMo, 1x VpWz, 1x DSAr, 1x KeGl, 1x KeAr, 1x FeMo, 1x MDBe, 1x MuFE, 1x DrCr, 1x HaNe, 1x HaFE, 1x CeEn, 1x MDEE, 1x DrEE, 1x SEIE, 1x MiGl, 1x VpTm, 1x SpIE, 1x KoAE, 1x MDPa, 1x HOFE, 1x CeBe, 1x HaWn, 1x TrCr, 1x
[22:08] <Sequell> HOVM, 1x HaPr, 1x MuMo, 1x MuAr, 1x CeHe, 1x DGAs, 1x KoVM

Moreover, why should one have to stop playing pure melee characters? That's not meant to be a viable build? If one wants to play berserkers and doesn't want to use a crossbow one should find another roguelike?

You're supposed to need more than three skills for the entire game

according to whom? Maybe I just don't like mephitic cloud, which is perhaps why many others don't have a problem with this.
mikee_ has won 166 times in 396 games (41.92%): 4xDSFi 4xMDFi 3xDDCK 3xDDEE 3xHOPr 2xDDHe 2xDDNe 2xDSBe 2xKeAE 2xMfCr 2xMfSt 2xMiAr 2xMiBe 2xNaTm 1xCeAr 1xCeAs 1xCeBe 1xCeEn 1xCeFE 1xCePa 1xCeTm 1xCeWz 1xDDAs 1xDDCr 1xDDHu 1xDDTm 1xDENe 1xDEWz

Ziggurat Zagger

Posts: 3037

Joined: Sunday, 2nd January 2011, 02:06

Post Monday, 25th July 2011, 15:03

Re: Can Good AI Be Bad?

minmay wrote:The only ranged option non-ranged backgrounds usually get right away is chucking darts and stones, which is hardly sufficient for centaurs.


I'll happily agree that melee-oriented backgrounds are horrendous in the early game. However, I believe that the best approach to correcting this problem is to give them some options to start with, not create a blanket ban on interesting monsters in the early game to accommodate a crippled starting background. Ranged threats that mindlessly close in to melee range are not interesting.

If you make it past Temple and still have no solution for ranged threats other than to charge directly across open ground, then you get no sympathy from me.

mikee wrote:Moreover, why should one have to stop playing pure melee characters? That's not meant to be a viable build?


No, I'm pretty sure it's not. It was once, and it got specifically nerfed to remove that viability.

mikee wrote:If one wants to play berserkers and doesn't want to use a crossbow one should find another roguelike?


If you don't want to use a crossbow for your berserker, there's also the options of slings, thrown weapons and blowguns, evocations, and liberal use of Brothers in Arms. If literally no option other than mashing arrow keys is acceptable to you, then possibly roguelikes in general aren't for you. Even commercial roguelikes for general audiences are stricter than that.

The new ranged AI makes for a better game. It's unfortunate that you lost the set of exploits that you were abusing, because you had fun with those exploits and now they're gone, but I'd rather find a new way to have fun with the better version of the game than leave the problems in forever just so I don't have to change.

Vestibule Violator

Posts: 1567

Joined: Friday, 21st January 2011, 22:56

Post Monday, 25th July 2011, 15:19

Re: Can Good AI Be Bad?

I don't think that much changed with the new ranged AI. Many corners still "work", there are just a few more situations where you cant trivialize ranged threats just by taking one step back. And with good AC and a shield simply walking up to a centaur and chopping its head off usually works unless it is a very early centaur. Maybe drink heal wounds at some point or something.

I don't think the new ranged AI is that big a deal, and in many situations it is a slight improvement.

Crypt Cleanser

Posts: 718

Joined: Monday, 14th February 2011, 05:35

Post Monday, 25th July 2011, 15:38

Re: Can Good AI Be Bad?

KoboldLord wrote:[If you don't want to use a crossbow for your berserker, there's also the options of slings, thrown weapons and blowguns, evocations, and liberal use of Brothers in Arms. If literally no option other than mashing arrow keys is acceptable to you, then possibly roguelikes in general aren't for you. Even commercial roguelikes for general audiences are stricter than that.

The new ranged AI makes for a better game. It's unfortunate that you lost the set of exploits that you were abusing, because you had fun with those exploits and now they're gone, but I'd rather find a new way to have fun with the better version of the game than leave the problems in forever just so I don't have to change.


WOW. Way to turn a discussion about how a feature may be tedious or annoying into how I must be an abusive player who shouldn't be involved in roguelikes. All for daring to not play the way you do.
mikee_ has won 166 times in 396 games (41.92%): 4xDSFi 4xMDFi 3xDDCK 3xDDEE 3xHOPr 2xDDHe 2xDDNe 2xDSBe 2xKeAE 2xMfCr 2xMfSt 2xMiAr 2xMiBe 2xNaTm 1xCeAr 1xCeAs 1xCeBe 1xCeEn 1xCeFE 1xCePa 1xCeTm 1xCeWz 1xDDAs 1xDDCr 1xDDHu 1xDDTm 1xDENe 1xDEWz
User avatar

Ziggurat Zagger

Posts: 5832

Joined: Thursday, 10th February 2011, 18:30

Post Monday, 25th July 2011, 15:55

Re: Can Good AI Be Bad?

Retreat behavior, IMHO should work similar to this abstract:
  Code:
boolean fearFlag = false;
boolean retreatingFlag = false;
boolean ralliedFlag = false;
int morale = ##; // value 1 - 100, higher is less chance of retreat.  Value never changes
int retreatDamageThresholdPercentage = ##; // value indicates chance under which morale check due to damage needs to be made

boolean isCapable() {
  // checks monster mental states to see if monster is generally normal and not charmed or confused or afraid or retreating or stunned otherwise incapacitated
}

boolean isAfraid() {
// checks states of magical fright; returns fearFlag
}

boolean isDamagedEnoughToRetreat() {
// checks current damage vs the retreatDamageThresholdPercentage
}

boolean isRetreating() {
// checks if monster was retreating previously; default false.
return retreatingFlag;
}

boolean isRallied() {
// checks if monster was just previously retreating but just recovered; is only true for one round.
return ralliedFlag;
}

int entityRadius(int distance, object alignment, object state) {
// returns the number of creatures of alignment (friendly/ally, unfriendly/non-ally) in stated distance from current entity with specified state
}

// Run prior to the start of a monster's turn
checkMorale() {

  // start retreat chance at -50 if rallied last turn, otherwise default to 0
  int retreatPercentageChance = ralliedFlag ? -50 : 0;

  // reset rallied
  ralliedFlag = false;

// check retreat
if (isRetreating() || isAfraid() || isDamagedEnoughToRetreat()) {
  // check morale
  retreatPercentageChance += 100 - morale;

  // check fear status
  retreatPercentageChance += isAfraid() * 50;  // isAfraid checks spell fear effects; 50 is arbitrary

  // affected by retreating allies, diminishing - would almost suggest using this ALL THE TIME to make healthy allies potentially retreat.
  retreatPercentageChance += 10 * entityRadius(1, ally, state.retreating) + 5 * entityRadius(2, ally, state.retreating) + 2.5 * entityRadius(3, ally, state.retreating);

  // affecting by non-retreating capable allies, diminishing
  retreatPercentageChance -= (10 * entityRadius(1, ally, state.capable) + 5 * entityRadius(2, ally, state.capable) + 2.5 * entityRadius(3, ally, state.capable));

  // affected by recently rallied allies, diminishing
  retreatPercentageChance -= (10 * entityRadius(1, ally, state.rallied) + 5 * entityRadius(2, ally, state.rallied) + 2.5 * entityRadius(3, ally, state.rallied));

  // affected by proximity of capable enemies, diminishing
  retreatPercentageChance += (10 * entityRadius(1, enemy, state.capable) + 5 * entityRadius(2, enemy, state.capable) + 2.5 * entityRadius(3, enemy, state.capable));

  if (rollUnder(retreatPercentageChance) && canEscape()) {
    // RUN AWAY
  retreatingFlag = true;
  } else {
    // NO RETREAT
    if (isRetreating()) {
      // NO LONGER RETREATING, RALLIED!
      retreatingFlag = false;
      ralliedFlag = true;
    }
  }

boolean canEscape() {
// returns true if monster can move away from player.
}


The concept I want to relate here is that ally states and quantities should affect whether a monster decides to retreat or continue retreating, and that recently recovered monsters should help rally their retreating allies.

Edit: added in additional to cover multiple player-aligned entities and genericized for use with player allies.
"Be aware that a lot of people on this forum, such as mageykun and XuaXua, have a habit of making things up." - minmay a.k.a. duvessa
Did I make a lame complaint? Check for Bingo!
Totally gracious CSDC Season 2 Division 4 Champeen!
User avatar

Abyss Ambulator

Posts: 1189

Joined: Friday, 28th January 2011, 21:45

Post Monday, 25th July 2011, 17:24

Re: Can Good AI Be Bad?

KoboldLord wrote:
mikee wrote:Moreover, why should one have to stop playing pure melee characters? That's not meant to be a viable build?


No, I'm pretty sure it's not. It was once, and it got specifically nerfed to remove that viability.


Then what's a Felid Berserker supposed to do then? Flip Trog the middle claw?


Anyway, ranged A.I. The new one is better and much more interesting. But if I don't want to just charge them in the open, I can still usually utilize the nearby terrain or other monsters to get that Centaur into a much better position. Unless I'm just on one of those very open levels, in which case it doesn't matter which A.I. the Centaur has.
The best strategy most frequently overlooked by new players for surviving: not starting a fight to begin with.

Dungeon Master

Posts: 1613

Joined: Thursday, 16th December 2010, 21:54

Post Monday, 25th July 2011, 18:11

Re: Can Good AI Be Bad?

Personally I'm fine with the new ranged AI, even from the perspective of pure melee characters (which should be (and are) viable, if sometimes difficult), but I'm not sure how that makes it okay to start attacking people for their quite reasonably-expressed opinions. Maybe it'd be good not to do that.

Bim

Crypt Cleanser

Posts: 700

Joined: Wednesday, 5th January 2011, 15:51

Post Monday, 25th July 2011, 19:08

Re: Can Good AI Be Bad?

galehar wrote:What's your point? You start by explaining how the current behaviour leads to extreme abuse then you're saying it's a lot of fun.

Sorry for not making that clear (and for the delay) I meant that although it's fun in a glitch sort of way, I'd really like to see it fixed.
It seems to break the 'tension' when you can just sit in a corner of deep water and pelt things without any fear from anything.

However, I also stated that I think it would be a mistake for monsters to run off as soon as they realised they couldn't get to you, as that would take away a lot of the interesting tactics that water brings in. I don't know what the best way to handle this is really, but it just seems a bit lame to see intelligent monsters sit and take 20 javelins to the face.
2012 Winner of fewest proposed ideas implemented by devs.
User avatar

Halls Hopper

Posts: 63

Joined: Monday, 6th June 2011, 21:28

Post Monday, 25th July 2011, 20:25

Re: Can Good AI Be Bad?

Bim wrote:it just seems a bit lame to see intelligent monsters sit and take 20 javelins to the face.


Obviously, they should pick them up and throw them back at you. (here's to me being hated by generation upon generation of crawlers!)

Bim

Crypt Cleanser

Posts: 700

Joined: Wednesday, 5th January 2011, 15:51

Post Monday, 25th July 2011, 20:34

Re: Can Good AI Be Bad?

I think they do sometimes? Could be imagining it though!
However, I don't find monsters ranged to be a threat unless it's a purpose made ranged monster (e.g centaurs), possibly a separate issue to be addressed?
2012 Winner of fewest proposed ideas implemented by devs.

Shoals Surfer

Posts: 321

Joined: Friday, 17th December 2010, 02:21

Post Monday, 25th July 2011, 20:42

Re: Can Good AI Be Bad?

Bim wrote:However, I don't find monsters ranged to be a threat unless it's a purpose made ranged monster (e.g centaurs), possibly a separate issue to be addressed?


It sounds like you haven't witnessed the horror of an Imp who found a bow and arrows :)
In fact, I think the only difference between "ranged" monsters and "melee" ones is the gear they spawn with. A melee monster with a ranged weapon will hurt as much as a ranged monster given equal HD and base damage.

Bim

Crypt Cleanser

Posts: 700

Joined: Wednesday, 5th January 2011, 15:51

Post Monday, 25th July 2011, 20:51

Re: Can Good AI Be Bad?

Very true Asdu,
I think it's mainly that ranged weapons are scarce (needing to presumably find both launcher and ammo) and spears/whatever don't tend to be a problem after the first few levels.
2012 Winner of fewest proposed ideas implemented by devs.
User avatar

Dungeon Master

Posts: 4031

Joined: Thursday, 16th December 2010, 20:37

Location: France

Post Monday, 25th July 2011, 22:24

Re: Can Good AI Be Bad?

mikee wrote:I know my approval is unlikely to decide anything, but the improvement to ranged AI is the change I like the LEAST out of any in the last four releases.

This is the first negative comment I see on the ranged AI improvement. When I made this, I said in the commit message that it's a buff to ranged monsters and that maybe some balancing would need to be done. Like buffing ?fog for example. Nobody commented on this specific suggestion. So let's bring it again. How about buffing scroll of fog? Currently, it's quite unlikely (impossible?) to block LOS in 1 turn, but almost guaranteed in 2 turns (if you step back). What if it makes a little bit more fog so it has about 50% chance of blocking LOS on the first turn? And/or make them more common.

XuaXua wrote:The concept I want to relate here is that ally states and quantities should affect whether a monster decides to retreat or continue retreating, and that recently recovered monsters should help rally their retreating allies.

Fleeing when your allies are fleeing is something intelligent monsters would do. But intelligent monsters don't flee when wounded, only animals do. The retreating behaviour proposed is a really special case that has completely different conditions than fleeing.

Bim wrote:However, I also stated that I think it would be a mistake for monsters to run off as soon as they realised they couldn't get to you, as that would take away a lot of the interesting tactics that water brings in. I don't know what the best way to handle this is really, but it just seems a bit lame to see intelligent monsters sit and take 20 javelins to the face.

Interesting tactics? Like go in water and kill them without any risk? So, they shouldn't run off, and they can't attack you and staying there doing nothing is a bit lame. So what do they do? Suicide? Making them go away is not the best way to handle this, it's the only way. Even animals wouldn't just stand there.

Zuboki wrote:Obviously, they should pick them up and throw them back at you.

Good idea. Currently, monsters will never pick anything thrown or dropped by you unless they are just wandering. So you can't distract them by dropping/throwing stuff and run away while they pick it up. Maybe this can be improved too.
<+Grunt> You dereference an invalid pointer! Ouch! That really hurt! The game dies...

Dungeon Master

Posts: 553

Joined: Wednesday, 22nd December 2010, 10:12

Post Tuesday, 26th July 2011, 00:38

Re: Can Good AI Be Bad?

galehar wrote:
mikee wrote:I know my approval is unlikely to decide anything, but the improvement to ranged AI is the change I like the LEAST out of any in the last four releases.

This is the first negative comment I see on the ranged AI improvement. When I made this, I said in the commit message that it's a buff to ranged monsters and that maybe some balancing would need to be done. Like buffing ?fog for example. Nobody commented on this specific suggestion. So let's bring it again. How about buffing scroll of fog? Currently, it's quite unlikely (impossible?) to block LOS in 1 turn, but almost guaranteed in 2 turns (if you step back). What if it makes a little bit more fog so it has about 50% chance of blocking LOS on the first turn? And/or make them more common.

I am not as opinionated on this as some people; I don't think the new ranged AI makes the game any harder or more interesting, but I don't think it makes the game that much worse either. Overall I don't really care, although I think I liked the old behaviour a bit more. I do understand why some people strongly dislike it though, so I'll try to explain what I think the main problem is.

The new AI isn't really a buff to ranged attackers. The tactic of backing around a corner has been removed, but there are still enough alternatives (going around a corner in a corridor so you can ambush them, using doors, etc). Once you know how to deal with ranged attacks, none of these strategies are hard to do.

The real difference is that the current tactics are (arguably) more tedious than the old method of just going around a corner. Before, if you found a group of yaktarus in the vaults, you could just move around a corner. Now, you have to retreat all the way to a door, close it, and then wait for them to come to you. This takes more time to do, but ultimately it's just as easy.

Bascially, they aren't harder to fight. They just take longer to fight.

There are of course ways the ranged AI can be further "improved", but doing too much of that would make the game less fun. For example, consider this:

If a monster has ranged attack and can see you but not fire at you (because of friendly, plants or status in the way), it will try to move into a spot with a line of fire.

Right now, one of the ways of dealing with ranged attackers is to position yourself in a way that their shots are blocked by other enemies. Some people might not like this, but I personally think it can be interesting. A bit of tactical movement will let you position yourself in a way that lets you fight ranged enemies without being in too much danger. If yaktaurs were to intelligently move so they always try to have a line of fire, this would just further encourage retreating-based tactics (to a corridor or a door). You'd just be removing yet another option from the player, rather than making an actual improvement.


Anyway, I suspect that this is one of those issues that have a bigger effect on more experienced players, probably because they are the ones who are most likely to notice the changes, and know how to easily adapt to them.

I should also say that even if you haven't heard negative comments on this before, such opinions aren't that uncommon. What I've written here is my own opinion, but it's not unique by any means. Most people just aren't that outspoken about it (nor am I, usually).

Dungeon Master

Posts: 553

Joined: Wednesday, 22nd December 2010, 10:12

Post Tuesday, 26th July 2011, 00:54

Re: Can Good AI Be Bad?

Oh, and regarding scrolls of fog, they could stand to be be buffed, but they are still good items as they are right now. I like using them more for offensive purposes than defensive ones - read fog, retreat a bit, then let the ranged enemies close the distance. It's especially good for stuff you can't outrun (such as centaurs), or stuff that casts torment. Buffing fog to block LOS in 1 turn wouldn't change this usage too much. What it would do is make them more useful for defensive purposes, imo.

I also agree with mikee on the issue of ranged enemies being a bit too powerful sometimes. See wiglaf with a crossbow, for example. Or pretty much any strong melee enemy using a ranged weapon. When good players see an orc warlord, the first thing they do is check if it has a crossbow. This is because it's an easy enemy if it has a melee weapon, but can be deadly with a ranged one.
User avatar

Mines Malingerer

Posts: 51

Joined: Friday, 21st January 2011, 08:22

Location: Mare Desiderii

Post Tuesday, 26th July 2011, 01:14

Re: Can Good AI Be Bad?

I enjoy it when different monsters have different behaviors, and the new ranged AI doesn't exactly break anything nor is it strictly worse than the old behavior. I like it, but that doesn't mean that crawl's AI couldn't use some more improvements in general. :)

Bim

Crypt Cleanser

Posts: 700

Joined: Wednesday, 5th January 2011, 15:51

Post Tuesday, 26th July 2011, 01:42

Re: Can Good AI Be Bad?

Interesting tactics? Like go in water and kill them without any risk? So, they shouldn't run off, and they can't attack you and staying there doing nothing is a bit lame. So what do they do? Suicide? Making them go away is not the best way to handle this, it's the only way. Even animals wouldn't just stand there.


I just meant that it would become pointless to go into water (other than to escape and heal) if they moved away as soon as you got into deep water. At the moment it's nice if you see a level with water as an aquatic because you know that you can use that to your tactical advantage. However, at the moment I feel it's TOO much of an advantage.

What I'd like to see is perhaps if monsters ran off after a couple of hits (even if each hit didn't do a large amount of damage) and they couldn't close the gap between you. If they ran away straight away I feel that would make it seem even more unnatural than them standing there, but as you say, it's not great to have them just suicide themselves.

Having them pick stuff up to throw back is also good and could ease up the problem, especially with javelins and darts. This might be enough to keep the 'water edge' without nerfing it too much.
2012 Winner of fewest proposed ideas implemented by devs.

Ziggurat Zagger

Posts: 3037

Joined: Sunday, 2nd January 2011, 02:06

Post Tuesday, 26th July 2011, 01:55

Re: Can Good AI Be Bad?

minmay wrote:Quite a lot of people on IRC (including many of the best players) dislike it, and the OP is expressing that too.


Eh, I'll take the opinions of the 'best players' with at most a single grain of salt unless they can back up their opinion with supportable reasons, and not just a measurement of their longest streak or list of games played online or whatever.

Nethack has an interesting feature called 'Elbereth'. For those who are unfamiliar, it's basically an Elder Sign you can doodle on the floor. 95% of all monsters in the game will refuse to cross it and will refuse to attack you if you're on it. Consequently, it's a fairly reliable means of nigh-invincibility that you can start using from turn 1. The 'best players' in Nethack love this feature. They can get crazy winning streaks of arbitrary length and can walk through the entire game without using basic features of their character, like not using any offensive attacks the whole game, and they can do these things reliably. These 'best players' bitterly oppose any talk of nerfing Elbereth into some semblance of sanity or maybe just removing the thing completely, arguing that making this change would somehow wreck the game.

The 'best players' are always going to be ultra-conservative. Any change is a bad change when you're currently on top and you have to claw your way back up to the top once your position has been put into question by a major change. You've worked everything out so you can practically play blindfolded, and now this change means you have to re-evaluate everything you knew! The 'best players' will often have valuable insight, but they do not have some uniquely privileged position that automatically overrides all other perspectives.

In this particular case, the ranged AI is strictly better from almost every reasonable perspective. Aesthetically speaking, the monsters act slightly less idiotic in that they don't barge directly into the player's sword every single time no matter what the surrounding terrain is as long as the player has a magic corner to work with. For gameplay needs, instead of having one tactic that works in every scenario you need to adapt tactics to the situation at hand, and not every pack of yaktaurs is going to lend itself to exactly the same solution. Certainly, it is nice when sometimes a victory falls into your lap, but it would stop being fun if it happened every time.

I'm perfectly willing to agree that there are probably some further adjustments to be made, however. Nerfing early ranged enemies is an option, of course. My own preference is to fix non-magic backgrounds to cover this weakness. There's a range of valid opinions here, but throwing away some honestly clever AI coding because change is axiomatically bad? That's not one of them.
User avatar

Tomb Titivator

Posts: 857

Joined: Monday, 31st January 2011, 23:19

Post Tuesday, 26th July 2011, 02:18

Re: Can Good AI Be Bad?

My own preference is to fix non-magic backgrounds to cover this weakness. There's a range of valid opinions here, but throwing away some honestly clever AI coding because change is axiomatically bad? That's not one of them.


I share this opinion too. Improving AI is only bad if players aren't given more tools to deal with it. Or, of course, the range of ranged attacks could be reduced so you can't be sniped as soon as you enter LOS. Other options could include reducing their ammo, giving monsters MP pools, and such.

Also I don't understand why an orc priests smite doesn't have a cool down instead of this silly cantrip thing that's currently being used.

Crypt Cleanser

Posts: 718

Joined: Monday, 14th February 2011, 05:35

Post Tuesday, 26th July 2011, 02:18

Re: Can Good AI Be Bad?

KoboldLord wrote:Eh, I'll take the opinions of the 'best players' with at most a single grain of salt unless they can back up their opinion with supportable reasons, and not just a measurement of their longest streak or list of games played online or whatever.

Nethack has an interesting feature called 'Elbereth'. For those who are unfamiliar, it's basically an Elder Sign you can doodle on the floor. 95% of all monsters in the game will refuse to cross it and will refuse to attack you if you're on it. Consequently, it's a fairly reliable means of nigh-invincibility that you can start using from turn 1. The 'best players' in Nethack love this feature. They can get crazy winning streaks of arbitrary length and can walk through the entire game without using basic features of their character, like not using any offensive attacks the whole game, and they can do these things reliably. These 'best players' bitterly oppose any talk of nerfing Elbereth into some semblance of sanity or maybe just removing the thing completely, arguing that making this change would somehow wreck the game.

The 'best players' are always going to be ultra-conservative. Any change is a bad change when you're currently on top and you have to claw your way back up to the top once your position has been put into question by a major change. You've worked everything out so you can practically play blindfolded, and now this change means you have to re-evaluate everything you knew! The 'best players' will often have valuable insight, but they do not have some uniquely privileged position that automatically overrides all other perspectives.


It looks to me like the ad homs against me failed to frame the argument the way you wanted to, so you constructed a straw man instead. No one is discussing anything similar to Elbereth in Nethack, nor are the people who don't like this feature opposed to it just because it's relatively new.

evilmike has done a good job recapping the points I semi-coherently made earlier. Please refer to those in future arguments, not to Nethack or other 'analogies'.

As for why I quoted some games I had played online, I'd like to remind you that was in response to you suggesting I didn't like a feature because I only played mdfi or whatever (where did that even come from?). It's not meant to imply some special authority of mine.
mikee_ has won 166 times in 396 games (41.92%): 4xDSFi 4xMDFi 3xDDCK 3xDDEE 3xHOPr 2xDDHe 2xDDNe 2xDSBe 2xKeAE 2xMfCr 2xMfSt 2xMiAr 2xMiBe 2xNaTm 1xCeAr 1xCeAs 1xCeBe 1xCeEn 1xCeFE 1xCePa 1xCeTm 1xCeWz 1xDDAs 1xDDCr 1xDDHu 1xDDTm 1xDENe 1xDEWz

Ziggurat Zagger

Posts: 6393

Joined: Friday, 17th December 2010, 18:17

Post Tuesday, 26th July 2011, 04:09

Re: Can Good AI Be Bad?

KoboldLord is just racist against Mountain Dwarves, that's all.
User avatar

Dungeon Master

Posts: 4031

Joined: Thursday, 16th December 2010, 20:37

Location: France

Post Tuesday, 26th July 2011, 07:55

Re: Can Good AI Be Bad?

evilmike wrote:The real difference is that the current tactics are (arguably) more tedious than the old method of just going around a corner. Before, if you found a group of yaktarus in the vaults, you could just move around a corner. Now, you have to retreat all the way to a door, close it, and then wait for them to come to you. This takes more time to do, but ultimately it's just as easy.

Bascially, they aren't harder to fight. They just take longer to fight.

Retreating all the way to a door is just as easy? Only if you have at least the same speed I guess and your example is yaktaurs. Anyway, if fighting them is more tedious as melee fighters, maybe giving them more tactical options is more interesting than going back to the dumb AI. A couple of special moves could help with that, like charge or sidestep.

evilmike wrote:Buffing fog to block LOS in 1 turn wouldn't change this usage too much.

I said 1 turn only 50% of the time. Don't want to make them too reliable.

evilmike wrote:I also agree with mikee on the issue of ranged enemies being a bit too powerful sometimes. See wiglaf with a crossbow, for example. Or pretty much any strong melee enemy using a ranged weapon. When good players see an orc warlord, the first thing they do is check if it has a crossbow. This is because it's an easy enemy if it has a melee weapon, but can be deadly with a ranged one.

That's a good point. Ranged weapon damage formula for monsters is quite silly. Can't dig it right now, but suggestions for improvement are welcome.

Bim wrote:I just meant that it would become pointless to go into water (other than to escape and heal) if they moved away as soon as you got into deep water. At the moment it's nice if you see a level with water as an aquatic because you know that you can use that to your tactical advantage. However, at the moment I feel it's TOO much of an advantage.

What I'd like to see is perhaps if monsters ran off after a couple of hits (even if each hit didn't do a large amount of damage) and they couldn't close the gap between you. If they ran away straight away I feel that would make it seem even more unnatural than them standing there, but as you say, it's not great to have them just suicide themselves.

galehar wrote:If a monster cannot attack you, cannot reach you and is taking damage from you, it will retreat. A behaviour similar to fleeing but which will stop on different conditions.
<+Grunt> You dereference an invalid pointer! Ouch! That really hurt! The game dies...

For this message the author galehar has received thanks:
MyOtheHedgeFox

Dungeon Master

Posts: 553

Joined: Wednesday, 22nd December 2010, 10:12

Post Tuesday, 26th July 2011, 10:19

Re: Can Good AI Be Bad?

galehar wrote:Retreating all the way to a door is just as easy? Only if you have at least the same speed I guess and your example is yaktaurs.

I used yaktaurs as an example because the game has such a vast number of them. But it's also worth noting, most ranged enemies are speed 10. This includes yaktaurs, but also other ranged enemies, such as orc priests, elves, the various giants, most dragons, most uniques, liches, draconians... well, you get the idea. Centaurs are the only notable exception I can think of. Overall I think it's a good thing for most enemies to be speed 10, and for fast monster to be rare (it makes centaurs unique amongst ranged enemies) but the point is, for most enemies in the game, retreating is quite easy and safe.

Oh, and I think a "charge" type of move where you move quickly towards an enemy (perhaps doing a bit of extra damage when you connect) would be great. It would be a viable way of dealing with centaurs/yaktaurs in certain circumstances, and seems like it would be pretty fun too.

Temple Termagant

Posts: 10

Joined: Saturday, 23rd July 2011, 11:31

Post Tuesday, 26th July 2011, 10:25

Re: Can Good AI Be Bad?

Maybe shields could be used to provide a tactic for melee characters against ranged attackers..?

Here's what I'm thinking:

Grant a bonus to blocking ranged attacks on turns where the player's last action was movement, based on shield skill and the size of shield equipped. This would simulate a situation where the player basically ducks behind their shield and charges forward, thus reducing the area overall area they're vulnerable to getting shot at. Bucklers would be almost useless for this, and large shields would be optimal. And, of course, a character without reasonable shield skill wouldn't have the sense or skill to pull it off effectively.

As far as I can tell, to avoid skewing things against using two-handed weapons over wearing a shield, the way shield equipment is handled would have to be reworked a bit. Perhaps allowing a shield to be equipped whether using a two-handed weapon or not(taking the normal amount of time to "wear" it), flag it as active or inactive depending on the handedness of what's being wielded(a.k.a. slung over the back when not in active use.) Then, provide an extra delay penalty when switching between one-handed and two-handed weapons?(The shield could perhaps be readied with a seperate action, to allow the player to choose whether or not to bother with the shield... That would allow better handling of one-a-half-handed weapons.)

I can foresee some situations where this bonus wouldn't make much sense, like being fired at from multiple directions... But, accounting for that would be complicated.

On a side note: This would have a side-effect of giving the Fighter background a nice early game perk in their favor, since they really don't have much going for them now.

Edit: Just a grammar fix.

Bim

Crypt Cleanser

Posts: 700

Joined: Wednesday, 5th January 2011, 15:51

Post Tuesday, 26th July 2011, 11:07

Re: Can Good AI Be Bad?

galehar wrote: galehar wrote:If a monster cannot attack you, cannot reach you and is taking damage from you, it will retreat. A behaviour similar to fleeing but which will stop on different conditions.


Then I'm afraid I don't think it's working. Just two days ago I was camped in the back of some deep water with a two headed ogre just waiting whilst I pelted it with throw frost. I also played as an OcHu a bit before that and slinged a whole bunch of orcs who were just standing at the edge whilst I stoned them. I'm playing on the last trunk before beta release.
2012 Winner of fewest proposed ideas implemented by devs.

Dungeon Master

Posts: 1613

Joined: Thursday, 16th December 2010, 21:54

Post Tuesday, 26th July 2011, 11:16

Re: Can Good AI Be Bad?

That's the proposed change, not the current behaviour.
User avatar

Tomb Titivator

Posts: 857

Joined: Monday, 31st January 2011, 23:19

Post Tuesday, 26th July 2011, 12:16

Re: Can Good AI Be Bad?

I hope the mechanics of any changes aren't too convoluted.

Take, for example, the formula for using shields. Anyone under the assumption that SH works like EV will be very surprised when they face a hydra. How is an unspoiled player supposed to know that the more attacks you take in a turn, the less your shield works? That's just not something you'd expect to be coded into the game.

Also having enemies retreat when they can't attack you is quite possibly the worst idea ever conceived. I've seen it implemented in other games... and it makes optimal play painfully tedious without removing any advantage whatsoever from, for example, using deep water as a merfolk.

Step into deep water, the ogre retreats.
Pelt it a few times, step out of deep water, the ogre moves towards you.
Step back into deep water when it gets closer, the ogre retreats.

So instead of being 100% safe in deep water you're 100% safe while tediously deep water dancing. These AI "improvements" add nothing but tedium to the game....
User avatar

Dungeon Master

Posts: 4031

Joined: Thursday, 16th December 2010, 20:37

Location: France

Post Tuesday, 26th July 2011, 13:06

Re: Can Good AI Be Bad?

snow wrote:Also having enemies retreat when they can't attack you is quite possibly the worst idea ever conceived. I've seen it implemented in other games... and it makes optimal play painfully tedious without removing any advantage whatsoever from, for example, using deep water as a merfolk.

Step into deep water, the ogre retreats.
Pelt it a few times, step out of deep water, the ogre moves towards you.
Step back into deep water when it gets closer, the ogre retreats.

So instead of being 100% safe in deep water you're 100% safe while tediously deep water dancing. These AI "improvements" add nothing but tedium to the game....

As I have explained the implementation will include a bit of randomization so that the ogre won't necessarily run back to you as soon as you step out of water. The goal is to make reaching harder to abuse so we can give it to octopodes without giving them an intrinsic abusable set of features. You'll still be able to do it with ranged attacks, but the new behaviour don't just add tedium but risks too. If it takes more time, it increases the chance that other monsters join the party. We can make the retreating monster shout for help too.
Anyway, I know this change won't completely solve the problem, but I believe it is a step in the right direction. I don't understand how you can argue that monsters staying in place doing nothing while being pelted to death is an acceptable behaviour, or how having means of killing monsters without taking any risk is good design.
<+Grunt> You dereference an invalid pointer! Ouch! That really hurt! The game dies...
User avatar

Ziggurat Zagger

Posts: 5832

Joined: Thursday, 10th February 2011, 18:30

Post Tuesday, 26th July 2011, 15:20

Re: Can Good AI Be Bad?

There was a great board game called "Advanced Hero Quest" (by Games Workshop) where the game master (aka the GM, a non-player who controlled the monsters) would randomly generate rooms on the fly for the players.

Monsters were not allowed to open doors. Certain rooms would contain monsters called "sentries" and those rooms would always require at least one additional door generated. Sentries could open doors, which means they could explore the game like the players, but sentries were controlled by the GM. Players or the GM would layout the room depending on random selection, then the non-layout side would have the option to move all monsters at least 1 space, then combat would commence.

The tactic of the smart GM would be to get the sentry placed as close to the opposite door as possible, to get the sentry to open the door and continue to explore the dungeon. The sentry would then "awaken" as many other monsters (and possibly other sentries) as possible to overwhelm the players. I once had 4 sentries running around the board and piles of Skaven coming after the players.

A dumbed-down version of the game called "Heroquest" was produced by Milton Bradley; it was a fold-out board game, and AHQ had rules for incorporating it into their game.

Anyway, "yelling" is as close to sentry option. I'd love to have a monster that ran away yelling and randomly exploring, waking up the dungeon.
"Be aware that a lot of people on this forum, such as mageykun and XuaXua, have a habit of making things up." - minmay a.k.a. duvessa
Did I make a lame complaint? Check for Bingo!
Totally gracious CSDC Season 2 Division 4 Champeen!
User avatar

Abyss Ambulator

Posts: 1189

Joined: Friday, 28th January 2011, 21:45

Post Tuesday, 26th July 2011, 17:48

Re: Can Good AI Be Bad?

What if the monsters simply backed out of the player's LOS until they've moved a certain distance from deep water and/or spent so many turns out of it? They're not afraid of you, but they don't want to take arrows to the face without being able to fight back. It'd also be neat if they could scramble for any ranged weapons and wands nearby, player owned or not, as mentioned above.

It also might help to prevent or reduce potential "water dancing" by making a step from deep water directly onto solid ground take extra time. Pulling you up to shore, especially when loaded with heavy armor and a pack full of potions, scrolls, wands, chopped up corpses, and who knows what else, would not be a quick task! Maybe even prevent such a transition if the player is encumbered or overburdened.
The best strategy most frequently overlooked by new players for surviving: not starting a fight to begin with.

Mines Malingerer

Posts: 42

Joined: Friday, 7th January 2011, 01:41

Post Tuesday, 26th July 2011, 18:11

Re: Can Good AI Be Bad?

The AI improvement is pretty damn good IMO, someone talked about having to look for an specific place in the level to ambush the centaur so it doesn't burn his scrolls, well that's not the only way to deal with it, you can either go down a level or just go straightfoward and kill him. Is it optimal? No, but its way more fun, I imagine if you are a "competitive" player you could be pissed off since the most optimal way of playing is tedious, but from the perspective of an ocasional player such as me who never had any streaks, the AI improvement adds more than it takes, it makes the game more fun since it puts you in some positions where you have to take a chance or do something you don't want to, it forces the player to think a bit more than just going to a corner and wait.



Randomized good AI (or better AI on rand-unique monsters like a "green centaur") sounds good to me since it could add interesting tension moments (like a centaur who runs back or a wizard with optimal spell casting). Of course it would have to be balanced but it could be very very fun since combat as it is could use more tactical options.
User avatar

Dungeon Master

Posts: 4031

Joined: Thursday, 16th December 2010, 20:37

Location: France

Post Tuesday, 26th July 2011, 21:51

Re: Can Good AI Be Bad?

minmay wrote:If deep water promotes tedious or abusive player tactics, it's much better to fix deep water itself than try to work around it with convoluted AI.

Fix deep water? Like what? Removing it? Allowing monsters to walk into it? Removing merfolks and levitation?
And how is the retreat behaviour convoluted? It's simple and intuitive, that's what you would expect them to do. Even an animal wouldn't stand in place while you keep bouncing stones off its head.
<+Grunt> You dereference an invalid pointer! Ouch! That really hurt! The game dies...
Next

Return to Game Design Discussion

Who is online

Users browsing this forum: No registered users and 24 guests

cron
Powered by phpBB © 2000, 2002, 2005, 2007 phpBB Group.
Designed by ST Software for PTF.