Specially Designed Smart Monsters

[o_O]

So really smart centaurs would be impossible to kill in melee and smart orc priests would just smite everything, but it would be fun to have some really smart monsters that weren't so problematic for balance. What we find annoying may vary, but in general they shouldn't retreat, be uncatchable for melee or unkillable with spells. Maybe stuff like:

A unique band of weak Human adventurers with equipment that follows you upstairs Brogue style, and tries to keep its squishy casters behind tanky knights.

A fast monster that tries to hide out of LOS like a wandering mushroom, then attacks when you are fighting something dangerous.

Spellcasters that just let rip with magic dart and move back when they dont have a clear shot

Big melee dudes who hide behind doors

A shopkeeper that attempts to hoover a level and open a shop when it finds enough floor trash

A ranger who attempts to round up monsters into groups, then find you.

Herd/swarm monsters who refuse to split up or go into narrow hallways one at a time

[roctavian]

I try not to use the term 'spoilers' too much or at all, but I would be really concerned about creatures that behave in different ways because they might promote bizarre and degenerate player behavior. For example, the shopkeeper idea is terrible because it is countered by the player picking up everything. The 'monster maintains distance ' style of movement can be gamed, and IIRC there was a patch a while ago that smartened up ranged monster behavior that people didn't like. I think problems could additionally arise if monsters tried to avoid single tile corridors; it would be better for monsters to have abilities that make corridors seem undesirable for players.

That's what works well for slime creatures. I'd like to see a pack monster that does something nasty but loses the ability for a few rounds after being hit, so that players would prefer not to engage only one at a time.

I do personally feel that wandering mushroom style lurking could be used more -- something a lot like what you suggested would be interesting, though it should probably be confined to a branch or special vault.

[Davion Fuxa]

In a sense we have already started to get these - like the Orb Spider in version 0.11 and Saint Roka is already pretty Smite-happy. Main thing to note though is that these sort of enemies should appear in the later branches (like how Orb Spiders appear in Spider's Nest) whereas if they are smart enough to run away and keep handling you at range you can deal with it by running half the level to get them into a corner. It generally won't matter if the enemy is unkillable either if there are ways the character can survive as well - like making a target that only requires spells to be killed be auto-friendly to Trog Worshipers or you can lose it somehow by ways of time duration, staying out of an area, or like how with Player Ghosts you can simply run from them by jumping off the Dungeon Level they appear on.

[ebarrett]

Really smart posters would not propose to repeat tragedy (4.1) as farce (GDD)

[galehar]

I think there's room for AI improvements. Although, I don't like most of the OP's proposals, especially the ones which happen outside of LOS.

IIRC there was a patch a while ago that smartened up ranged monster behavior that people didn't like.

I coded the ranged AI and it got a mixed reception. Some love it, others hate it. I think it could be improved some more (improvement means more interesting, not more efficient). Make it a bit less predictable and also rarer. The chance for a monster to use it could depend on their intelligence and on a "ranged factor". The latter could be manually set for each monster or automatically calculated based on (ranged damage / melee damage). But then you have to give arbitrary weights to spells. On the other hand, monsters would automatically adapt their behaviour depending on how they are equipped. If this works well, the ranged AI could be generalized so that monsters have a chance to use it whenever they see you but can't shoot you (because of other monsters or plants in the way).

Another thing which I'd like to have for all monsters is smarter behaviour regarding clouds. Again, this should also depend on their intelligence, so animals would be likely to rush through a conjured flame but only if they can't easily step around it. Elves on the other hand would have a much higher chance of avoiding cloud especially if we can combine cloud and ranged AI in a flexible way. And I think it might be possible using brendan's diffusion implementation. This stuff looks like it could have a lot of potential for improving the AI in an interesting and flexible way.

Really smart posters would not propose to repeat tragedy (4.1) as farce (GDD)

The very first phrase of the OP makes me think he is aware of the problem of using 4.1 AI. But it doesn't mean there's nothing which can be done. I wouldn't go farther than the 2 proposals above for AI changes which apply to all monsters. However, giving special AI to some new monsters (or a few existing ones) can be interesting, and I know many devs agree with that. A good AI has the potential to create interesting tactical situations and a bad AI has the potential to be very annoying, so it needs to be handled with care.

The diffusion algorithm can be used to make a band of monsters cooperate intelligently. Some may try to flank you, others would maneuver intelligently to fire at you or support their allies. Killing the leader could disable the special AI. Used sparingly, it could be very nice.

[dpeg]

I believe the following approach is uncontroversial: uniques can get as good an AI as possible.

[dpeg]

minmay: Rest assured that I had developers in mind when I said "uncontroversial"; certainly not you.

It is plain to see that you're (a) a great player and (b) lack fantasy when it comes to design. So we will listen to one half of your comments, and ignore the others.

[Davion Fuxa]

On the topic with the ranged attacks coming at you - is it possible that the RNG could come into play in deciding what the enemies do instead of what their equipment tells them to do? Like if an Orcish Knight is wielding a Crossbow with Bolts of Pentration for Ammo, wearing pimped out Orcish Plate Mail, while having an Orcish Battleaxe in his inventory; maybe instead of players knowing that he'll want to use that Battleaxe, the Orcish Knight instead can decide that he'll play with the Crossbow if given the chance. Additionally perhaps he can switch his battle strategy while in combat and/or decide to use that Battleaxe next time he sees the player.

[crate]

um if you improve nessos's AI (or the AI of several other uniques who can be faster than the player, though nessos is the most obvious example since he's very fast and ranged) i'm pretty sure that's the worst thing ever

personally I mostly like how crawl's enemies behave because it's predictable and works fine for gameplay, isn't this supposed to be the point? I already don't like ranged AI much for various reasons.

[prozacelf]

Yeah, I've had an orc warlord swap out his great sword for a crossbow and straight up murder me....and I've seen numerous *taurs grab a weapon off the ground to melee from time to time.

[Davion Fuxa]

On the topic with the ranged attacks coming at you - is it possible that the RNG could come into play in deciding what the enemies do instead of what their equipment tells them to do? Like if an Orcish Knight is wielding a Crossbow with Bolts of Pentration for Ammo, wearing pimped out Orcish Plate Mail, while having an Orcish Battleaxe in his inventory; maybe instead of players knowing that he'll want to use that Battleaxe, the Orcish Knight instead can decide that he'll play with the Crossbow if given the chance. Additionally perhaps he can switch his battle strategy while in combat and/or decide to use that Battleaxe next time he sees the player.

Monsters with ranged weapons already use them...

I meant in more advance terms, like the Orc Knight will choose the strategy to keep running back if you chase him or maintain distance like an Orb Spider does now, in addition to now in how he blindly either shoots a crossbow until he is in melee with your character and moving forward if something is blocking is LOS between you and him; basically you can't predict what he's going to do because he has multiple attack strategies to pick from, though once you get into battle you should be able to identify what strategy he decided to pursue. It could also be linked to how his Orcish buddies are faring - maybe after you kill all the other Orcs he'll opt for a 'stand and fight' strategy as opposed to 'hit and run'.

um if you improve nessos's AI (or the AI of several other uniques who can be faster than the player, though nessos is the most obvious example since he's very fast and ranged) i'm pretty sure that's the worst thing ever

I agree that Nessos would be ungodly difficult - but then he could also be moved to appear later in the dungeon too (and personally even without the change I think he spawns too early in the Main Dungeon. A second buffer could be to reduce the ammunition Nessos carries (along with every other Ranged Enemy) so that lose the ability to continue loosing or firing ammo at you and have to resort to melee. Another buffer could simply be balancing - like perhaps Yaktaurs of Medicore Weapon skills but Excellent Crossbow skills, while Orc Knights have Decent Weapon Skills but lacklustre Crossbow Skills; this would all factor into their damage output with specific weaponry so Orcs will never deal a whole lot with Crossbows (or Bows, or Blowguns) while Yaktaurs will never really be all that dangerous in Melee.

For enemies that use spells, I don't think they necessarily deserve any change in behavior since they have the advantage of not needing a mana pool. If they had mana pools they would likely run up you melee you after their spells had depleted it beyond a certain threshold.

personally I mostly like how crawl's enemies behave because it's predictable and works fine for gameplay, isn't this supposed to be the point? I already don't like ranged AI much for various reasons.

Predictability is good and there is something to say about it. Recently I started doing a Let's Play for ToME and in one instance I started to play really fast and ended up getting killed because all of a sudden a really high-leveled enemy appeared with a 'kill set' where I had just mowing them down weaker types of the same enemy without worry that never spawned with anything that caused me worry.

However, the counter-argument here is 'decision-making' in all its glory.

[mikee]

I'd just like to suggest that when coming up with ideas you think about what optimal play in that situation would be, i.e., how players are likely to respond to your idea being implemented.

IDEA: monsters with ranged attacks should maintain range so you can't melee them! Maybe some spellcasters should stay at the edge of LOS and then leave if things get bad for the monsters. -> PLAYER RESPONSE: learn poison arrow and go vehumet on every single character.

IDEA: why don't all of the d:1 monsters wander towards the downstairs and cluster around them so you have to fight them all at once? Maybe they can pick up all the food on the way so you're forced to fight them sooner. -> PLAYER RESPONSE: delete crawl.

[galehar]

Let's make it clear. We don't want to generalise the maintain range behaviour. It will stay as a special monster behaviour and probably exclusive to orb spiders.
So maybe we could discuss other ways to improve the AI. The tweaks to the ranged AI I talked earlier doesn't involve monsters running away from the player when they can just shoot him. Only moving to a spot from where they can shoot instead of running up to you whenever there's something in the way.

[Galefury]

http://dictionary.reference.com/browse/fantasy

???

1.

[rebthor]

The diffusion algorithm can be used to make a band of monsters cooperate intelligently. Some may try to flank you, others would maneuver intelligently to fire at you or support their allies. Killing the leader could disable the special AI. Used sparingly, it could be very nice.

It could be thematic (if perhaps overly deadly) in Vaults if the new vault warden / protector whatever was able to apply some tactics with the other mobs there. Obviously the balance would need to be tweaked.

And I'd love to see this a little more with the late game orc packs that show up. Because right now, maybe the knight is still a little dangerous, but that doesn't really matter when the other orcs go down in one shot.

[BlackSheep]

Well, if you replace 'fantasy' in dpeg's post with the first word of definition one, it might make a little more sense.

[BlackSheep]

Apparently dpeg believes you lack it, at least as regards crawl design.

[BlackSheep]

...
Use your imagination.

[BlackSheep]

*sigh*
He is attributing your opinion of the concept as a "bad idea" to a lack of imagination on your part, thus dismissing your complaint. He values your input as a great player, but not as a designer.

[dpeg]

Giving in to thread pressure.

I believe the following approach is uncontroversial: uniques can get as good an AI as possible.

You believe incorrectly.

"Good" AI almost always makes encounters less interesting, not more; it removes options from the player. Monsters that never go into corridors? Great, now you're completely eliminating the player's ability to use terrain to their advantage! How interesting!

I thought it would go without saying -- but I believed incorrectly -- that AI improvements have to be applied judiciously. A pan lord who always start at the boundary of LOS, using haste and blink for that purpose, and spamming Fire and Ice Storms, element depending on your resistances would make near perfect AI, but lousy gameplay. Sure!

To counteract the onslaught of negativity and lack of imagination, I'll mention some successful instances of improved monster AI:

• Monsters making way for monsters of the same genus. This makes things more interesting tactically, and is great flavour-wise. Note that this kind of AI is painless enough that it can apply to all monsters, not just uniques.
• Ranged attackers trying to (a) reach LOS with you, but (b) keeping distance. This change made centaurs etc. more interesting because it has become less trivial to rid them of their ranged attack. Nonetheles,s it could be applied to all such monsters because you can generally still rely on terrain to deal with them, and in the rare cases where you can't, tools like fog and invisibility (in addition to the standard tools) are available.
• Maurice, the thief. He is using all kinds of nasty tricks, including theft, invisibility and blinking. It'd be very annoying if every humanoid could do this. Instead, it is interesting because it is the behaviour of a special opponent, a unique.

I brought up uniques, because most monsters in Crawl are dispensible. We should consider them as redshirts, and AI improvements can be done (like for ranged attackers) but should be pondered at length. On the other hand, with uniques we can run wild. Having a unique use controlled blink or silence or summoning etc. cleverly is okay. It may lead to a complicated or lengthy encounter, but that's what a unique is good for. Obviously, it can be overdone also for uniques, but then we either spawn it deeper or dress down its AI a bit.

If you think that "good AI makes monsters less interesting", i.e. the game worse, then you're shutting yourself out of one potential area where the game could be improved. This is not good -- this mode of thinking means stagnation.

To close with an example from Brogue monster AI, which surprised me a bit: Brogue monsters have perfect information. A fleeing monkey will never sleep and patiently run in circles. It is very hard to catch one, and it sounds as aggravating as it turns out to be if you try. Brogue centaurs shoot, and they keep distance: if you approach a step, they draw back a step. That is as evil as it sounds. Now, Brogue is well playable, so how can that be? For centaurs, they will approach if you wait behind a door (so in that regard their AI is intentionally not perfect). Apart from that, you're expected to use ranged damage, blink, beckoning (makes them come to you), discord (makes a monster attack other monsters, and centaurs always come in pairs) etc.
There are a number of reasons why none of this would work in Crawl, but it does teach us that we should design without fear. Bad ideas can and will be eradicated in the process -- at worst, when playing them. Good ideas will never see the light if we're afraid.

[roctavian]

I didn't say good AI in roguelikes was always bad! Just that it was almost always bad. I really should have been more specific about what "good" meant; certainly Crawl monsters shouldn't just do nothing. But I also don't think they should pick their spells based on your resistances, or have any movement habits more complex than following the shortest path to their target - these are spoilery and annoying for players to keep track of. And I certainly don't think they should do things that just remove options, like forcing you to fight on specific terrain...which seems to be what all the proposed AI changes are aiming for.

Doing something more complex than following the shortest path to the target isn't spoilery, so long as it is something that can be understood by playing the game. I hate Brogue monkeys and "Smart Kobolds" style guessing games because they're simplistic problems with uninteresting solutions. Not every potential change to AI would necessarily lead to that kind of shit situation, and it would be better to debate other points -- "ranged enemies maintaining distance" is the poster child of AI discussion and we could stand to try on some other concepts.

[pratamawirya]

I thought the AI for centaurs is quite good right now, but then why would some great players say otherwise...?

[o_O]

I should mention that its assumed: not all monsters will get AI upgrades, uniques would be more likely to get more/better AI, ranged monsters that outrun and avoid an average melee char are not the goal of this idea and any monster that gets highly effective AI would probly also be rebalanced to compensate. Very smart enemies would probably have less raw firepower or a notable weakness to compensate. I think you all know that AI's can be improved without going all the way to unkillable centaurs and Metagaming D1 mobs and no one is suggesting anything of the sort.

My examples were just quick ideas, but I also wanted to show that these improvements would not have to be focused on optimizing raw combat ability. Adding flavor and unique challenges would be possible as well. Also: This would be an excellent way to give the late game challenges other then just hordes of demons.

One last thing: A variety of AI's would be ideal, and good AI should not always be predictable AI.

[crate]

I thought the AI for centaurs is quite good right now, but then why would some great players say otherwise...?

The idea is that they make wide turns around corners so that they don't just turn the corner and end up next to you. However you can almost always tediously kite them around the dungeon to a spot where this "ranged AI" doesn't actually come into play (and in Vaults you can use doors). So for some players this ranged AI ends up being a large annoyance with no real benefits.

[dpeg]

I thought the AI for centaurs is quite good right now, but then why would some great players say otherwise...?

The idea is that they make wide turns around corners so that they don't just turn the corner and end up next to you. However you can almost always tediously kite them around the dungeon to a spot where this "ranged AI" doesn't actually come into play (and in Vaults you can use doors). So for some players this ranged AI ends up being a large annoyance with no real benefits.

(Emphasis mine.)

That's the crux of the matter, in my opinion. If you regard everything more complicated than "monsters will move to me using the shortest path" as tedious or annoying then there is no scope for change. Whether it is annoying or not is a personal matter. I never found it to be. I do find centaurs and orc wizards to be harder than before, so I like the change a lot. In my experience, kiting centaurs is not always trivial, because of their speed.

[galehar]

have any movement habits more complex than following the shortest path to their target - these are spoilery and annoying for players to keep track of.

What is spoilery about the ranged AI? If it is intuitive and can easily be learned from experiment, then it's not spoilery. Maybe the current implementation has some flaws in this regard, and this is why I'm suggesting improvements. I don't see how a well designed ranged AI would be necessarily spoilery.

And I certainly don't think they should do things that just remove options, like forcing you to fight on specific terrain...which seems to be what all the proposed AI changes are aiming for.

Sometimes one option dominates so much you almost never consider others. Before ranged AI, you'd never consider charging a centaur given how easy it was to trick it to close in. Now, it's not so easy, so you use different tactics depending on terrain and other monsters instead of the same one over and over.
Also, if all monsters in the game have the same behaviour, then you're going to use the same set of tactics to handle them. If some monsters behave differently, you have to invent new tactics to deal with them. Variety is good. Seriously, try to imagine an encounter with a band of intelligent cooperating monsters. Hide in a corridor, some try to flank you. In the open, melee ones surround you, ranged one don't close in and try to maneuver around their friends to shoot you or buff them. Would it be necessary annoying? Sure, you can't use the same move pattern you are used to to disable the ranged ones, so time to try something new or use some consumables.

[siprus]

I'm all for the idea that some monsters should have specific unique AI, (of course not every monester)

[pratamawirya]

IMHO, the current ranged AI might not be perfect, but the old one was stupid. Easily tricking centaurs to close in was stupid and didn't reflect the behaviors we expected from them (surely we expect centaurs to not give up using their bows so easily).

If the current ranged AI has flaws, then we should improve it, and not reverting to the old one. So, as far as I'm concerned, the current AI is a step toward the right direction.

[adozu]

i think centaurs are plenty smart already. sure it might be easy to close on them with the right layout but getting spotted by one in an area with little cover is just as bad as being in the wide open.

[some12fat2move]

Let monsters swap places with all other monsters, not just other monsters of the same genus.

How about if all intelligent monsters can swap places with all other intelligent monsters, but non-intelligent monsters never swap? That would be a less drastic change and would probably be more intuitive than the genus rule.

[crate]

Before ranged AI, you'd never consider charging a centaur given how easy it was to trick it to close in.

I don't agree with this. Except insofar that I almost never charge at a centaur anyway I guess, since it's so easy to trick them to close in in most dungeon situations where you find them now also. But I would "charge at" centaurs in pretty much the same situations before and after ranged AI.

[jejorda2]

Let monsters swap places with all other monsters, not just other monsters of the same genus.

I'd like to see flying monsters (or maybe large monsters) be able to displace other creatures. Or let a flying enemy occupy the same space as a standing or swimming creature, though that's almost certainly too complicated to make work. I'd just like to see flying matter more when there are no terrain features around.

[dpeg]

Hey, here's something I'd like that you'd consider an AI improvement. Let monsters swap places with all other monsters, not just other monsters of the same genus.

How about if all intelligent monsters can swap places with all other intelligent monsters, but non-intelligent monsters never swap? That would be a less drastic change and would probably be more intuitive than the genus rule.

Maybe you haven't been around back then, but when Lorimer's (the man behind Sporkhack, and who suggested swapping) idea was implemented, it applied to all monsters. That turned out to be confusing: the bee making place for the iguana turned the dungeon into something like a giant hive mind; so it was replaced with the current version. There is certainly still room for improvement but you shouldn't suggest that we didn't think about variants of the rule.

[mikee]

Sometimes one option dominates so much you almost never consider others. Before ranged AI, you'd never consider charging a centaur given how easy it was to trick it to close in. Now, it's not so easy, so you use different tactics depending on terrain and other monsters instead of the same one over and over.
Also, if all monsters in the game have the same behaviour, then you're going to use the same set of tactics to handle them. If some monsters behave differently, you have to invent new tactics to deal with them.

This is actually the reason I have never agreed with the ranged AI change - my tactics did not change at all. Like before, I mostly either: A) charge up to them if I'm in a hurry or can take the hits easily; or B) I maneuver them into a position where I take the least possible damage. The main difference is that B) feels a little confusing and awkward to me (how did a cyclops or a brain worm become a master tactician?) and that B) takes 5-10 seconds longer to execute. I don't ever feel clever at outwitting a monster with the new ranged AI, nor do I feel more in danger, because it's really solving the same puzzle.

I do of course believe that variety is good for the game, and maybe this is off-topic, but my guess is that a variety of individual monster threats would be simpler and more effective than cooperative band behavior. This could be as simple as adding a few attention-demanding monsters (e.g., hellion, tormentor, giant eyeball) to some spawn lists. Maybe even more off-topic, I think that in some cases a general shift away from bands could be good. It is usually much easier to escape from a band of weak/medium strength monsters, just because only one or a few of them can be very close in chasing you at one time. It's also pretty much trivial for any character to wipe out a band by killing one or two of them at a time and retreating. This may be why many players feel mid to late game can be a bit of a grind. What if, after a certain point, all yaktaur bands were replaced with a quicksilver dragon or a lich? Ugly thing bands with a deep elf blademaster, etc.? Would this overall be better or worse?

[evilmike]

Yeah for crawl the best approach to AI isn't making stuff as smart as possible, it's using it to give certain monsters unique, well defined behaviours. This is usually done with spells, abilities, etc, but AI is another tool for accomplishing this. This applies to all monsters including uniques. It's far more interesting than an approach like "just make them smart as possible", which would be rather ham fisted in terms of game design, even though it would be considerably more complex under the hood.

I would also suggest that "how should we make [monster] smarter?" is the wrong question to ask. The right question is "how can we increase the amount of tactical depth when fighting [monster]?". The answer to that may very well be "give it better AI", but it doesn't have to be. In other words, I think improved AI can be a legitimate means to achieve a certain design goal, but it should never be an end in itself.

When it comes to monster bands, I don't think it makes sense to have them all behave super intelligently, with flanking and controlled retreating and so on. What's better is to think of one or two interesting behaviours, and try to apply that to a specific monster band. Again, this doesn't have to be done with AI, often it's enough to think of spells/abilities that can work. I think it can be more fun to fight a small-medium sized group of enemies that use their abilities to compliment each other, rather than a large horde of identical enemies which don't achieve anything special by being in a group (other than "there's a lot of them"). The best use for AI here would be to try making it so monsters use buff-other spells more intelligently, and have it so "support" enemies stay at the back of the monster band (this can be done by having them swap places with the front liners... that's how geryon and his hell beasts work).

In a sense, this would make it so monster bands act as more of an organic unit, rather than a horde of individual enemies. Take ogre bands, for example. Right now they're just an ogre mage (scary enemy which works well by itself too), with a bunch of normal and 2headed ogres (which again work fine by themselves, at the appropriate depths). As a band they don't have any real synergy, it's just a lot of easy ogres with a scary one thrown in. They aren't badly designed monsters, but as a band there's nothing interesting. A better design would let the ogres interact and support each other in specific ways.

I don't have a concrete proposal for that, though. I just know that "make the ogre band super smart" isn't going to cut it.

[Fergy]

Flinching:

Every monster given two variables.

• Flinch Rate: (Double) Determines how often a monster will flinch when hit. Most monsters have a flinch rate of 1. Monsters that rarely flinch will have a rate close to 0. Skittish monsters would have a rate higher than 1.
  
• Flinch Time: (Int) Determines for how long a monster will flee. Most monsters will have a flinch time of 50.
Flinch Chance: Every time a monster is hit:

if 1d(max HP) >= Damage taken * Flinch Rate, then Monster flinches (flees).
Flee time = 1d(Flinch Time) / 10

So what does this do?

When surrounded by monsters (or fighting in a corridor), this will make it so the monsters will naturally rotate, so you will be more likely to face a high hp monster. Monsters that have little HP will flee a greater distance, putting them at the end of the queue.

edit: simplified proposal

[rebthor]

We had lots of monsters that fled at low HP. It sucked.

[Fergy]

We had lots of monsters that fled at low HP. It sucked.

Yes it did, but flinching wouldn't make them run across the map (unless the flee time was set really high).

Considering the suggested average of 50 flee time, and an average speed of 10. This is a Maximum of 5 tiles when the monster is on very low health. Most flinches would only flee for 1-2 tiles.

[Fergy]

Swapping mostly does this, but it becomes nearly static once the monsters have ordered themselves. Consider the possibility of facing 3 gnolls in a hallway, the first has a spear, and the others have clubs. Tactically it would be best for a clubman to go in front of the spearman, but this will not happen. What will happen is you will kill the spearman, another gnoll will move forward and wield the spear. This creates the same inefficient tactical position as before.

The proposal is about improving the swapping behavior. What if it only triggered if there was a friendly monster in the fleeing direction? This would prevent the annoyance when fighting single monsters. Call it 'Tactical Swapping' instead of flinching.

Note: This proposal will also cause non-tactical things to happen, such as a clubman switching places with a spearman. However, this shouldn't be too bad because it would spread the damage to both of the monsters instead of just the clubman.

[mikee]

What's so fun and interesting about running around corners so you can take hits on the lead gnoll before the second one gains vision on you? Is there something wrong with fighting gnolls the regular way? Are gnolls too easy?