SquareLOS experimental branch on CBRO

[Pollen_Golem]

All that said, I have yet to address the elephant in the room that is ranged combat. I frankly do believe squarelos is a bad thing for ranged combat, because it makes it intrinsically stronger and crawl's quality as a game is inversely proportional to how relevant the ability to damage enemies at range is.

All core components of crawl rely on positioning and LoS manipulation as a cornerstone, ranged combat is in itself heavily detrimental to both of those things

the reasoning for the change is not acceptable because it assumes having a feature as important as LoS with inconsistencies such as the ones circlelos presents has actually led crawl to have worse gameplay than it would have under squarelos.

You should bold these parts if you want people to read anything you wrote.

[Sivar]

This doesn't actually fix any of the problems with Crawl geometry. Because Crawl takes place on a grid, you cannot have Euclidean geometry in Crawl. Period. Consistent geometries possible on a 2D grid include Chebyshev and Taxicab geometry. Crawl currently attempts Chebyshev in some places (movement, etc), and has a weird, terrible Fake Euclidean geometry in other places (LOS, ranges, explosions, etc)

Changing the diagonal movement cost to 1.4 solves absolutely nothing, and in fact actually makes the problem worse - instead of the mix of Chebyshev and Fake Euclidean that the game currently has, you'd have a mix of Chebyshev, Fake Euclidean, and Geometry Where Pi Is Equal To 3.1016. In addition, and more importantly for gameplay, Crawl actions take place on a "grid" just like Crawl geometry: performing five 14 aut actions is not equivalent to performing seven 10 aut actions, because monsters, the environment, etc. are able to act during those actions, while the player character is not. The result is the kind of disgustingly bizarre behaviour outlined in this post. I guarantee if this were implemented in Crawl I'd play either old versions or not at all until it got un-implemented, because tactical positioning - the entire reason I like Crawl - would become utterly nonsensical.

If you want to make Crawl geometry consistent, you have to either get rid of the grid or use a geometry that actually works on the grid. My preference would be to switch to pure Chebyshev; square LOS, square ranges, explosions, tornadoes, etc. Unfortunately, problems still arise because Crawl also tries to have continuous angles which are not really meaningful on a grid and result in aliasing; I gave a very short explanation of the problem with that here. You cannot eliminate angle aliasing on a grid as far as I know (many games such as NetHack attempt to hide this by only allowing you to shoot in 8 directions, but that just moves the problem from targeting to positioning). Here the only option to eliminate it is to get rid of the grid entirely, which is an obviously impractical change for Crawl. You could greatly reduce aliasing by switching from a square grid to a hexagon grid, but that's still a huge change to the game. So I support just switching to Chebyshev geometry, which fixes the inconsistencies even if it does leave the aliasing.

So maybe I'm missing something big, but I didn't really understand most of the so-called "disgustingly bizarre behavior" in that post the first time I read it and I still don't now. Why is it unexpected under pseudo-Euclidean geometry that the most efficient path to take when running away from a monster is the one directly away from it along the line described by your two positions?

The example of entering melee range with the yaktaur is weirder at first glance, but it makes sense when you consider that "melee range" is really 1.4 (since you can attack diagonally), and you're trying to reach the nearest point on the circle of radius 1.4 around the target from a few discrete choices. It's difficult to visualize on a grid, especially when you're playing console and the tiles do not appear square, but it's not nonsensical. Melee range being 1.4 also solves the "inconsistency" of being able to attack all 8 squares around you, by the way.

This isn't really meant as an apologia of 1.4x-movement diagonals, since my preference is really purely aesthetic and I don't mind much either way, but many of the arguments against it seem specious, and most people who tried the experimental version with it at the end of that thread said they were fine with it. I think the worst problem with it is actually that new players might not look at the aut counter and notice the discrepancy between orthogonal movement and diagonal movement speed.

I wonder how many people would be happy with taxicab geometry, like Advance Wars or Fire Emblem, which gives both aesthetically pleasing "circles" (well, diamonds) and consistent geometry. (Of course, this wouldn't really be feasible to implement given that map generation and other things assume diagonal movement is possible, but it's a fun thought experiment.)

[and into]

All that said, I have yet to address the elephant in the room that is ranged combat. I frankly do believe squarelos is a bad thing for ranged combat, because it makes it intrinsically stronger and crawl's quality as a game is inversely proportional to how relevant the ability to damage enemies at range is.

All core components of crawl rely on positioning and LoS manipulation as a cornerstone, ranged combat is in itself heavily detrimental to both of those things

the reasoning for the change is not acceptable because it assumes having a feature as important as LoS with inconsistencies such as the ones circlelos presents has actually led crawl to have worse gameplay than it would have under squarelos.

You should bold these parts if you want people to read anything you wrote.

I read it, and aside from agreeing or disagreeing with the argument, I wish more posts in GDD were like it! (Well, I wouldn't want *all* posts in GDD to be that long, but I would want all of them to be as reasonable in their mode of argumentation.)

The TL;DR version, for those in a hurry:

I think dck built a pretty good case that squarelos, though more consistent in certain respects, would substantially buff things that do not need buffing (most notably player's ranged combat) while nerfing things that probably do not need nerfing, and making enemy encounters—and thus game play—more homogenous overall.

[duvessa]

The example of entering melee range with the yaktaur is weirder at first glance, but it makes sense when you consider that "melee range" is really 1.4 (since you can attack diagonally), and you're trying to reach the nearest point on the circle of radius 1.4 around the target from a few discrete choices. It's difficult to visualize on a grid, especially when you're playing console and the tiles do not appear square, but it's not nonsensical. Melee range being 1.4 also solves the "inconsistency" of being able to attack all 8 squares around you, by the way.

What about when monsters are able to act? What about when you make more than one diagonal melee attack? Here is an even simpler example than the ones in evilmike's post, from another thread by what appeared to be a new player (so can we stop using that "argument" now please?):

to illustrate that the matter is grounded heavily in taste.

That's very true. Our current system has drawbacks, and it would be solved by either making LOS/range square, or making diagonal cost 1.4. My personal preference goes to the latter.

No it would not. Lets take a very simple example. You are fighting an elephant in the lair. You are wielding a whip of reaching. The elephant left and up from your current position. You hit him and he panics and runs straight away from you on the diagonal. If you try to use your whip to "reach" him you fail. It does not matter whether it is a movement cost of 1 or 1.4. This is a bug. The expectation is that it does not matter which way the elephant flees 1 square I should be able to use the reaching ability to get one last shot. As it is you have to follow the elephant and wait for him to make a less efficient move so you can have a shot with your reach weapon. This is bad no doubt, but the 1.4 movement cost has no effect on the situation. If he moves anything other then diagonal you still get that parting shot. Not only that what is even worse is that if when he makes that first move on the diagonal you instead of directly following him you step either horizontally or vertically you now get your parting shot right then and there. In other words you make a move that should bring you further away from your foe, but you actually get closer do to the f'ed up way your system works. You want to talk about inconsistencies. That is a major one. It is completely unintuitive, and as such spoily which crawl is supposed to be against. This tactic is easier to understand and exploit with a range 2 attack, but it is still present at other ranges if a bit less consistent. Basically if you move orthogonally when you opponent leaves or is about to enter your range you often get an extra shot. Is that what you are shooting for?

I think dck built a pretty good case that squarelos, though more consistent in certain respects, would substantially buff things that do not need buffing (most notably player's ranged combat)

Reducing player ranged combat's range from 8 to 7 is a substantial buff? That's news to me. Even if you keep the range at 8, this argument still strikes me as bizarre - player ranged attacks with square LOS and square ranges are surely weaker than ranged attacks with "circular" LOS and square ranges, which Crawl had for years. Square explosions and areas are new, sure, but there really aren't very many with radius 2+ in the game, and even fewer that are actually good (sanctuary and ball lightning are the only good ones I can think of at the moment).

[Sivar]

The example of entering melee range with the yaktaur is weirder at first glance, but it makes sense when you consider that "melee range" is really 1.4 (since you can attack diagonally), and you're trying to reach the nearest point on the circle of radius 1.4 around the target from a few discrete choices. It's difficult to visualize on a grid, especially when you're playing console and the tiles do not appear square, but it's not nonsensical. Melee range being 1.4 also solves the "inconsistency" of being able to attack all 8 squares around you, by the way.

What about when monsters are able to act? What about when you make more than one diagonal melee attack? Here is an even simpler example than the ones in evilmike's post, from another thread by what appeared to be a new player (so can we stop using that "argument" now please?):

<quote removed>

No it would not. Lets take a very simple example. You are fighting an elephant in the lair. You are wielding a whip of reaching. The elephant left and up from your current position. You hit him and he panics and runs straight away from you on the diagonal. If you try to use your whip to "reach" him you fail. It does not matter whether it is a movement cost of 1 or 1.4. This is a bug. The expectation is that it does not matter which way the elephant flees 1 square I should be able to use the reaching ability to get one last shot. As it is you have to follow the elephant and wait for him to make a less efficient move so you can have a shot with your reach weapon. This is bad no doubt, but the 1.4 movement cost has no effect on the situation. If he moves anything other then diagonal you still get that parting shot. Not only that what is even worse is that if when he makes that first move on the diagonal you instead of directly following him you step either horizontally or vertically you now get your parting shot right then and there. In other words you make a move that should bring you further away from your foe, but you actually get closer do to the f'ed up way your system works. You want to talk about inconsistencies. That is a major one. It is completely unintuitive, and as such spoily which crawl is supposed to be against. This tactic is easier to understand and exploit with a range 2 attack, but it is still present at other ranges if a bit less consistent. Basically if you move orthogonally when you opponent leaves or is about to enter your range you often get an extra shot. Is that what you are shooting for?

I can't get this example to work.

If the situation is like this...

  Code:

....
.Y..
..@.
....

Y...
....
..@.
....


...then either you still get your parting shot, assuming reaching weapons have ~2.8 range and not just 2, or you don't because they do just have 2 range, but that's expected because the elephant was farther away from you to begin with, so really this just encourages the positional play of trying to stay orthogonal to opponents when you have a reaching weapon. It's basically a nerf to reaching weapons relative to normal ones, since they have ~30% more range instead of 100%. I don't have an opinion on whether that property is desirable. Also, in this latter case, moving orthogonally still does not place you within 2 range of the monster, so it is in fact optimal to follow it exactly. It is true that there are values between 2 and 2.8 that you could choose for the range of reaching weapons that would allow the degenerate case where the orthogonal movement would let you get an extra hit in, but that's neither here nor there. Please let me know if I'm modeling this incorrectly.

edit: I do agree that the fact that the player only gets to act at discrete points does make things messy, and there are probably other corner cases besides this one. If I get some free time this week I'll see if I can get galehar's implementation running and find some of them.

[duvessa]

It is true that there are values between 2 and 2.8 that you could choose for the range of reaching weapons that would allow the degenerate case where the orthogonal movement would let you get an extra hit in, but that's neither here nor there.

On the contrary, it's exactly what happens if reaching has a range of 2, which it did at the time of that post. Now it has a "square" range, which would introduce the opposite problem in a 1.4 move cost system: you'd get more attacks against a diagonally approaching or fleeing monster than against an orthogonally approaching or fleeing monster. Feel free to replace reaching with some range 2 spell like low power flame tongue to still get the exact effect acvar was talking about.

[Pollen_Golem]

duvessa's logic here is ironclad

[Sivar]

It is true that there are values between 2 and 2.8 that you could choose for the range of reaching weapons that would allow the degenerate case where the orthogonal movement would let you get an extra hit in, but that's neither here nor there.

On the contrary, it's exactly what happens if reaching has a range of 2, which it did at the time of that post. Now it has a "square" range, which would introduce the opposite problem in a 1.4 move cost system: you'd get more attacks against a diagonally approaching or fleeing monster than against an orthogonally approaching or fleeing monster. Feel free to replace reaching with some range 2 spell like low power flame tongue to still get the exact effect acvar was talking about.

Right; that's the effect of the extra .8 range, which you can't take advantage of along orthogonal angles because the discrete values distance can take along those angles are 2 and 3. Come to think of it, you can construct the same example with 1.4 range if monsters are forced to take a diagonal movement towards you instead of an orthogonal one, which could happen if something blocked the orthogonal approach.

Anyway, I think we agree fundamentally on the actual gameplay effects. It's just a question of how bad the corner cases are in practice, how easy it is to mitigate them, and what the net effect is on the tactical nature of the game. I'm persuaded that the downsides may be worse than the upsides in this case, although I think I'll still give the Euclidean diagonals build a shot.

Actually, I'm now wondering what Crawl would be like with FFT-style CT bars instead of every action pretty much being 1 aut. But that's not for this thread.

[adozu]

regardless, as noted the change wouldn't alter crawl geometry (i don't think the sqr(2) diagonal movements are considered any more seriously than hex crawl), just the fact that you currently "see" further in some direction than another.

quick recap: circular los favors diagonal movement to get shorter approach routes, square los favors orthogonal approach routes to reveal less terrain, i don't think the issue with crawl having "magic directions" is quite solved.

[duvessa]

It's the "magic directions" that are inherent to Chebyshev, though. The closest system I know of that fixes that issue is a hex grid, probably followed by Taxicab. Both still have aliasing. (Taxicab is really bad with all but 4 angles, it's actually a much bigger change than a hex grid despite being visually similar to the current situation).
"Circular" los already reveals more squares on diagonal moves, by the way. It's less noticeable than squarelos obviously but it still exists.

[Pollen_Golem]

Seriously, if we want SquareLOS why do we have to go in all or nothing?
LOS/targeting/AoE are arbitrary shapes.
It's possible to slowly flatten circles over subsequent releases, and players won't think on it for more than a few minutes.

a)Square
b)Square with tapered corners
c)Octagon
d)Orthogonally tapered circle
e)Other square-circle hybrid
f)Circle

Why are only A and F points of discussion?

[Siegurt]

Seriously, if we want SquareLOS why do we have to go in all or nothing?
LOS/targeting/AoE are arbitrary shapes.
It's possible to slowly flatten circles over subsequent releases, and players won't think on it for more than a few minutes.

a)Square
b)Square with tapered corners
c)Octagon
d)Orthogonally tapered circle
e)Other square-circle hybrid
f)Circle

Why are only A and F points of discussion?

Because the issue that is addressed (consistency of movement and range) by squareLOS is only addressed by going to a full square, anything less than that retains (or exacerbates) said issue, and hence doesn't actually fix anything (So why bother making a change at all, if said change doesn't actually fix anything)

The point isn't "Make LOS more square" it's "Make movement and range play by the same rules", CircleLOS is inconsistent, but it's a *known* inconsistency, with plenty of code that works already written to support it, other weird shapes have less, not more consistent rules, and so writing additional code to support weird geometries that don't really do anything to fix the perceived problem isn't a very appealing tactic.

[tedric]

"Make movement and range play by the same rules"

This. As I said in my OP, the immediate noticeable effect of SquareLOS is that beam-targeting becomes *much* more intuitive. To be honest, I would be happy if beam-targeting ranges became square, regardless of the shape of max-LOS or any other changes.

[Siegurt]

"Make movement and range play by the same rules"

This. As I said in my OP, the immediate noticeable effect of SquareLOS is that beam-targeting becomes *much* more intuitive. To be honest, I would be happy if beam-targeting ranges became square, regardless of the shape of max-LOS or any other changes.

Hm, now I'm imagining a CircleLOS at range 8-5 (current LOS) with a targetting that maxed at 5 square diagonal/horizontal/vertical, god, that that seems weird and foreign.

[PleasingFungus]

First of all since the game does not take place in a vacuum most if not all situations in which a sleeping monster is spotted will not be immediately followed by engaging said monster, certainly not if the circumstances actually make the encounter in any way relevant. At the very least there will be repositioning to see what else is in the neighborhood of the monster in question and to gain information on the terrain surrounding it; this is assuming that the player came to notice the monster from the perfect angle to engage it from, relative to his escape or fight routes that are explored.

...To shorten that somewhat, no proper gameplay situations that aren't of planetary alignment rarity will see themselves affected at all by this. This is because the intended impact of the change is that the player no longer has to reposition when spotting a sleeping monster in situations where attacking is the chosen route; but there are tens of features that go vehemently against that very notion, such as the ability to wake monsters up while they are out of your LoS, wandering monsters that may be on the level or why not levelgen itself not being particularly (thankfully) fond of huge empty spaces everywhere.

It takes very few turns to reposition, and spaces don't have to be particularly large for you to be able to reposition. When playing stabbers, I encounter situations in which it's clearly optimal to reposition very frequently; from comments I've seen others make (e.g. elliptic), my impression is that they do as well.

And this brings me to my next point which is autoexplore...

...crawl is not so brutal that one improper move will shatter all chances of winning? Should we not also realize that crawl has so much is going on that having sqlos allow to engage without repositioning in the rare chance autoexploring into a monster goes perfectly and having the monster stay asleep changes next to nothing of the actual game?

That doesn't even remotely follow. Most people use a feature that makes the game slightly harder, with the benefit of removing vast amounts of tedium; that's a crude patch over a fundamental design flaw in crawl, but that's neither here nor there. Because of this, we should keep in a feature that encourages weird, fiddly repositioning before attacking enemies, because the game offers you enough wiggle room to not bother...? That's not an argument!

Not only this, in fact, but stabbing as a whole will be a lot less interesting without circlelos allowing for diagonal shenanigans, because sqlos' diagonal moving makes the dangers of approaching at an angle several orders of magnitude greater and one cannot give any sort of stealth benefit to make up for the sheer amount of terrain being revealed or he'd replicating the perceived problems circlelos with stabbing at a different scale!

It's true that sqlos makes stealth worse than the 'optimal case' of diagonal-stabbing in circlelos. (Though, of course, if it's viable to stab by charging orthogonally at enemies in circlelos, it's even more viable to stab from the one-tile-closer edge of squarelos. And it orthogonal stabbing must be viable in circlelos, otherwise your claim that there's a reasonable decision between doing so and between backing off to stab diagonally would make no sense, right?)

But... how would adding a player stealth buff 'replicate the percieved problems [of] circlelos... stabbing'?


I'd also like to make an appeal for you to employ your inner editor and try to make your points in a more concise fashion. More people will read and engage with you if you do. You restated each of your points several times in your post, which bloated it remarkably; the missing punctuation, misused vocabulary, and run-on sentences made the situation worse. When I review my posts before hitting 'submit', I often find myself guilty of some of those sins; sometimes I even fix them . It would make me a happier person if you made an effort toward that in future.

[tedric]

"Make movement and range play by the same rules"

This. As I said in my OP, the immediate noticeable effect of SquareLOS is that beam-targeting becomes *much* more intuitive. To be honest, I would be happy if beam-targeting ranges became square, regardless of the shape of max-LOS or any other changes.

Hm, now I'm imagining a CircleLOS at range 8-5 (current LOS) with a targetting that maxed at 5 square diagonal/horizontal/vertical, god, that that seems weird and foreign.

I was imagining that targeting would expand as a square up to full LOS, even though the corners of the square will get cut off when range is more than 5. Current targeting really only bugs me for short-range spells, not those at or near full-LOS.

Obviously I'd rather have squares all the way down, but in the meantime this would solve the issue that most bugs me turn-to-turn during normal play.

[Pollen_Golem]

sqlos' diagonal moving makes the dangers of approaching at an angle several orders of magnitude greater

This literally means stabbing at an angle is as least 100x more dangerous in sqlos than O-los, and that's stretching the definition of several.

Someone fetch dck an eyepatch.

For those concerned about revealing too much terrain (ie monsters) while exploring diagonally in open areas with sqlos, remember that you can just press 868686 instead of 999 and you actually reveal less terrain per turn in sqlos than in O-los.

We need a smiley like but with square eyeballs.

[Wahaha]

I played a melee character until D:15 and the game played the same way in 99% of cases. Obviously stabbers and characters with ranged attacks are affected much more by this change, but I just want to mention that for melee characters square or circle los does not matter, from my experience.

[dpeg]

It should go without saying that if squarelos testing suggests further changes, that'll happen. I can imagine changes to stealth (which can be buffed in a number of ways) or spell ranges.

What I get from this discussion is that the principal aim of squarelos is achieved, and it should move from a branch to trunk.

[dck]

@PF:
The whole first half of the post is referring to the logic stated in the commit as one of the big reasons, namely repositioning prior to approaching asleep monsters. The entire point was showcasing how repositioning is almost always ideal regardless of LoS being a circle, a square or a triangle with pink dots. And how the "almost" goes completely out of the window the minute you factor in autoexplore and the way it makes the player meet enemies.

RE: stabbing, you cannot give incentives that make stabbing at a diagonal in sqlos (in a situation where LoS type actually matters, so open ground) any less of a horrible idea. I say cannot because if you do and the chances of getting the stab are roughly equal to those of an orthogonal approach there is still a huge incentive to approach orthogonally, since the amount of revealed tiles is vastly inferior.
Similarly, if you give diagonal stabbing any sort of boon that makes diagonal stabbing have a noticeably better chance at succeeding than an orthogonal attempt, then you are literally replicating the current situation with stabbing and circlelos. Situation which is seemingly not desirable or you wouldn't refer to the behaviour as "weird and fiddly", which are just the buzzwords du jour to make a given feature look bad.

Regardless I can only qualify your last part of your post as eye-opening, and can only hope you find your happiness elsewhere.

[njvack]

Mod note: removed 100% noise post.

Also, I realized I just moderated a thread I've commented on, which is a no-no. Apologies.

[advil]

So I was curious and checked out this branch, and more or less accidentally won a 3 rune game (I really intended to just try it out for 20 minutes or so). This was a loud melee char (qazlal in fact), so not necessarily the best test case for the branch. Some observations:

• This was the easiest game of crawl I've ever played. I don't think this was particularly due to squarelos, it was at largely that the rng was just giving me insane stuff at a fairly constant rate (and I haven't played a Mi, let alone a MiFi since like the 0.15 tournament). It sort of had the feel of a meleebug character (like, are y'all sure that something about randart generation or item quality in general isn't messed up in 0.17??). Unfortunately this may make my impressions less useful.
• Aside from the easiness, this just didn't "feel" particularly different from the 0.16 LOS. So from the perspective of melee I think squarelos is not a big thing. I should also say that I went into this expecting to hate it. But of course, I used no ranged, and was doing the opposite of sneaking up on things.
• It was _way_ more noticeable visually on webtiles than in console, because the color contrast isn't nearly as sharp in console, at least with my settings. I hadn't even been aware of this until someone in my group who was spectating pointed out how different from normal it looks in tiles. This makes me think (because I'm a console-only player) that I might not be as attuned to the intricacies of LOS as I should be (and I say this as someone who's played a whole bunch of stabbers, and is in the middle of a VpNe in 0.16). For a while I was playing in console while spectating myself in tiles, I would recommend this for testing for any console players.
• I do think that qazlal may have been more powerful in squarelos (I have a couple of earlier qaz wins to compare to). Overall it seemed harder for things to be in my los and not be hit by my clouds, and disaster area seemed very effective. I also feel like it might have a subtle but pervasive effect on qazlal to not have as many enemies in los even though they can hear you, especially in later parts of the game. But because of the aforementioned rng issues it's hard for me to be sure, so this is a pretty subjective evaluation.

Finally, I think actually trying out this branch might help inform some of the more theoretical discussion in this thread.

[njvack]

For what it's worth: this is in trunk now.

[celem]

Yeah I was just about to say... I didnt even notice for 5 levels of D as a stabber En. Its probably not gonna be a big deal...

[duvessa]

Is overall difficulty really that big of a concern here? I would think that if squarelos is making the game easier overall (which I still don't think is true at all), there are a lot of things that could be done to correct that. For example, 0.9 increased all characters' HP by 3 midway through development. You could make a similar change in the opposite direction if the game needs to become harder.
If a change requires additional changes, that can certainly make it a bit less attractive to implement, certainly, and if it requires too many additional changes it can be a deal breaker (ask dpeg why he hasn't removed HP regeneration yet), but at worst squarelos makes melee, including stabbing, a bit worse compared to ranged attacks, and makes turncount a bit lower. The former doesn't seem too hard to compensate for (you can always reduce ranged damage or improve stealth) and I don't see how the latter is actually a downside in the first place.

[dpeg]

duvessa: Very well spoken. I don't think that overall difficulty has anything to do with the LOS change. As everyone knows who's been around for a while: difficulty fluctuates noticeably from version to version. For example, Lair subbranch cuts and zombie expiration were two straight-up nerfs (without compensation, too).

[njvack]

Mod note: split artefact discussion to its own thread

[dowan]

duvessa: Very well spoken. I don't think that overall difficulty has anything to do with the LOS change. As everyone knows who's been around for a while: difficulty fluctuates noticeably from version to version. For example, Lair subbranch cuts and zombie expiration were two straight-up nerfs (without compensation, too).

Yeah, that's right! You devs owe us some compensation! So I suggest you bring back axe cleaving while confused, which was a nice little bonus to using axes that I think they could use. But don't bring back the weird old cleave mechanics where walls would mess up your cleaving.

Square LOS is nice, but the reduction in overall LOS does make it feel kinda squashed in open areas. I do really like not having to dance around as much to get enemies off the diagonal so I can shoot a spell at them.

[wheals]

duvessa: Very well spoken. I don't think that overall difficulty has anything to do with the LOS change. As everyone knows who's been around for a while: difficulty fluctuates noticeably from version to version. For example, Lair subbranch cuts and zombie expiration were two straight-up nerfs (without compensation, too).

Yeah, that's right! You devs owe us some compensation! So I suggest you bring back axe cleaving while confused, which was a nice little bonus to using axes that I think they could use.

this literally never existed

edit: oops, it did for 2 days

[dowan]

I thought that was always how it worked prior to version 14 or so. I could be wrong, I had a period where I really liked axe wielding orcs, and it just so happened in that time that while confused you still did a cleaving attack, meaning it was really easy to hit things adjacent to you even while confused.

[advil]

Is overall difficulty really that big of a concern here? I would think that if squarelos is making the game easier overall (which I still don't think is true at all), there are a lot of things that could be done to correct that. For example, 0.9 increased all characters' HP by 3 midway through development. You could make a similar change in the opposite direction if the game needs to become harder.
If a change requires additional changes, that can certainly make it a bit less attractive to implement, certainly, and if it requires too many additional changes it can be a deal breaker (ask dpeg why he hasn't removed HP regeneration yet), but at worst squarelos makes melee, including stabbing, a bit worse compared to ranged attacks, and makes turncount a bit lower. The former doesn't seem too hard to compensate for (you can always reduce ranged damage or improve stealth) and I don't see how the latter is actually a downside in the first place.

Just to be clear, I reported on overall difficulty for my game because it was unavoidable to discuss given the rng I had, but I don't think it had much to do with squarelos. I tried to assess any impact on qazlal difficulty just because I assumed that this might be something the devs would be tweaking, not as an assessment of squarelos. I completely agree that the difficulty of squarelos isn't really that interesting (and probably isn't possible to assess directly). And after playing a full game, I'm completely on the side of squarelos (this after going in hating the idea in theory).

[celem]

I've played all the way through a 3-rune stabber in squareLOS now and ascended him, so i'll give some thoughts.

I didnt even notice until 5-6 levels into D. Once I did notice I started paying attention to see what was behaving differently and how.
I didnt mind at all about the whole 'longer approach on a diagonal'. Despite playing a number of stabber characters previously it appears I never appreciated that circleLOS gave us magic directions. So since I wasnt using them before, I didnt mind that they were gone.

I quite liked being able to throw stones and tomahawks out into the corners, ranged combat felt less stressful as I wasnt getting swarmed quite as fast from these directions (as I say, I never appreciated the differences before, but did notice them now its changed). Likewise aiming rods and wands was something I did a bit more often, since I wasnt being 'flanked' from the diagonals and switching to melee early.

It turns out that the only aspect of circleLOS that I had learned and took for granted was its shape. Trying to come up with sneaksy approach paths that got me close to a sleeper required a bit of thought, I really did notice the increased number of tiles uncovered when moving diagonally here. Likewise trying to pull back and lose LOS on something needed a little bit of thought. I regularly reposition when something wanders into LOS but hasnt yet seen me. Turns out I automatically pick diagonal retreats in this situation, probably due to the old circleLOS where that was the fastest path to LOSedge. One I became aware of this it was as simple as pausing to think for a moment before hitting a key.

Didnt mind in the least.