tain.totalcodex.net

I've been messing with the TFL gameplay switch in Myth II levels and I have noticed that something feels a little off. As far as I can tell the only difference between how the two units handle is that TFL's seem to be a bit clunkier and slower, but does anyone know specifically what was changed between the two games?

I am not affiliated with magma so wasn't involved with coding the TFL-gameplay option, but in vTFL archers and duff bottles are way less accurate, duff bottles bounce a lot more often, duffs can (and do) throw bottles straight up. i am sure there are other differences, those are the ones that jump out at me whenever i play vTFL after a while playing only SB.

Adding on to what VR said, the pathfinding is also a bit different, units can't clump as closely together and units turn in curves instead of turning in place. Then there are various minor details like different scoring on KOTH, lightning affecting objects a bit further away, some tweaked projectile detonation probabilities and more really minor stuff.

Because of those differences, your strategy will need to be more balanced between melee and missile/artillery with vTFL. Melee will not clump in tight groups making an easy target for arcs and dorfs, neither will allowing them to auto-target give you victory over an opponent's melee who've been clicked to attack, as it does in straight M2.

A long line formation works best for melee battles, with the player directing troops stuck behind the lines around the flanks or to weak points in the line. At the same time dorfs can flank & hit enemy melee from behind, as can ghols with pus etc. The saying "a wall of thrall" originated in TFL, as it was possible to have a solid phalanx of thrall form a tough obstacle to reaching your flag etc.

Which is why I always liked M2's AI better. I was always annoyed at TFL's braindead AI when fighting an enemy where instead of killing the thing right in front of their face, they would instead run around the group till they found the single target they wanted to kill.

The AI is actually no better in M2 at all, in fact its pathfinding is worse in some ways. In TFL, if a unit was blocked by enemies, it would choose a direction and keep walking that way until it found a way around. In M2 it'll go 3 metres one way, turn, go 3 metres the other, turn, etc etc. The behaviour you described was caused by targeting code, which probably assigned that unit an enemy on the other side of the fight, and that happens in M2 as well which is why auto-attack is better than click-to-attack with melee units.

What also makes M2's appear better is the clumping, which allows multiple units to get so close they're swinging their swords through each other to chop the enemy in front, plus allows units to slip past easier in general movement.

iron wrote:The AI is actually no better in M2 at all, in fact its pathfinding is worse in some ways. In TFL, if a unit was blocked by enemies, it would choose a direction and keep walking that way until it found a way around.

Right, and perhaps I'm strange, but I find that behaviour infuriating. The pathfinding in either game is simply far too sucky to trust to do *anything*, so I much prefer units to "give up" if they can't find a relatively direct path to their targets in a fight. I will readjust/micro them if need be, but taking off to run half way across the map to get "around" units is not what I want.

Similarly given the relatively poor pathfinding (and relatively slow gameplay), I vastly prefer the M2 unit spacings. Again, the algorithm is just too poor to handle all the requirements of "wider" units, so it's better to just allow stuff to go closer together rather than to reveal the flaws in the underlying system IMHO. Furthermore it usually is desirable to clump up melee fairly tightly before fighting, unless you have arty to consider... but that's what makes for good micro and clutch arty shots! In TFL it's mostly better to just spread out as much as the map will allow, which is far less interesting from a micro point of view.

We did some testing with the TFL vs M2 pathfinding code recently, particularly related to the "trow polka" that you describe and even though it was fairly easy to get whatever behavior we wanted in terms of going around vs. trying to sneak through units, in the end it was more desirable to stick with the controllable system, even if it means some conveniences like walking "around" slow units is lost. So sure units will occasionally moonwalk while trying to sneak through a hole or go back and forth a few times, but they won't take off in a direction that is completely different than what you asked them to do because their instantaneous decision about pathfinding tells them they need to go trek across the map to accomplish it...

To each his own - and I'm glad that both options are available - but I personally think M2 strikes a better balance of controllability vs. trying to be clever and mostly failing.

personally, I'd prefer some kind of pathfinding-computation-distance-time-limit so a unit would only go a certain distance looking for a way around instead of going halfway across the map, and likewise it would only moonwalk back and forth a couple of times before trying to find a way around.

ex. trow goes up to tightly packed melee can't reach target, walks back and forth a couple of times looking for access if it can't find a way in it starts walking around but if it gets x distance away from the target it goes back to target and resumes moonwalking, repeat. if this happens more than y times it just attacks nearest enemy unit. (which could be exploited by having trail of thrall to lure trow away, but we can't have a perfect system now can we?)

i know. it's easy to design the way things "should" work when you don't personally have to code and test them.

What I'd like to see is when you order a group of melee units to attack a group of enemy units, the ones in the front of your group attack the enemy units in the front of the enemy group instead of trying to wade through them to attack the ones in the rear.

are you clicking on enemy units or click-dragging in front of them?

Yeah that's the problem with clicking to attack. In theory the code _should_ assign targets sensibly, but in practice it often ends up doing what Khadrelt's described, which is why click-dragging to let them auto is a superior option - they all try to attack the closest enemy, which quickly dies, then the next closest, the next etc etc, and if your opponent has clicked to attack then his units will be moonwalking trying to get to an unreachable target.

Bear in mind that if you tell 20 warriors to attack 20 thrall, the algorithm matches them up so that each thrall has a warrior assigned to it. For example let's assume each group is in a 3-deep line, and you can easily see how 2/3rds of the warriors will try to reach thrall at the back of the enemy group. If the thrall auto-attack then they'll all concentrate on the front row of warriors & thanks to clumping those in the back 2 rows will also get hits in, ensuring the front row of warriors is killed quickly.

Punk, I understand what you're saying and I'm not saying TFL pathfinding was perfect at all. There's plenty of instances of it doing dumb things, that's for sure, but on other occasions it worked beautifully in ways that M2 could not. I remember seeing crowds of TFL melee on the charge, moving fluidly through enemy thrall in ways that seem natural & quite spectacular to watch.

With the keep-walking-around method TFL employed, a click-to-attack worked because the units stuck in the back rows would seek a way around the flanks to find the enemy, whereas with M2's method they'd all hopelessly moonwalk back & forth. Really its only clumping that allowed M2 to work melee-wise with its pathfinding scheme.

I think Vinyl's idea is a good one, but implementing it in existing bungie code would be quite a challenge.

iron wrote: Punk, I understand what you're saying and I'm not saying TFL pathfinding was perfect at all. There's plenty of instances of it doing dumb things, that's for sure, but on other occasions it worked beautifully in ways that M2 could not. I remember seeing crowds of TFL melee on the charge, moving fluidly through enemy thrall in ways that seem natural & quite spectacular to watch.

Sure, in terms of watching TFL looked nice with fewer clumps and longer lines of melee fighting it out one on one... I just don't like it in terms of gameplay Same with the units rotating/walking arcs: neat to watch but annoying to play.

iron wrote: I think Vinyl's idea is a good one, but implementing it in existing bungie code would be quite a challenge.

Yeah more importantly I'm disinclined to touch gameplay code at that level unless something is clearly a bug. It definitely changes the gameplay, and at this stage that sort of change is more suitable for a new game rather than a myth patch.

My thinking too - its not something you'd remotely consider doing with Myth. Hmm ... new game. Where have I heard that before?

No need for a new game, you could just add a checkbox next to the AntiClump option, you could call it 'VRPathfinding'.

Seems to me that an "attack group" command shouldn't be permanently pairing up targets with attackers when you click, it should be more like "move towards the enemy and autotarget as you go".

Of course, even if a new system could be implemented well, it's not clear that'd be desirable to do. Enhanced melee AI commands might reduce the opportunities for micro and take some of the skill out of it, which is bad for competitive play. If it were optional (a checkbox), it could be divisive - which setting should be used for tourneys or ranked play?