Lemmings 3D Level Design Strategies / Considerations

[namida]

Okay so, I've seen multiple people say that coming up with concepts for L3D levels is difficult. And in my own experience - having made close to 50 levels now - I definitely agree with that. It's a harder task than creating a 2D level, even just taking into account the "coming up with an idea" part.

Of course - given the work I put into creating an L3D editor, I want to encourage and help people to at least give level editing a go. And don't take that the wrong way - if L3D editing doesn't take off, it doesn't take off. The biggest thing for me was that I wanted an L3D level editor; other people using it is just a bonus, and the only true benefit I get from that is having some more levels to play. But it's still nice to encourage it.

But given that - I'd like to start a topic discussing strategies for designing levels.

A few things I've found:

Learn the keyboard controls - or at least the most important ones!
In-game, and in other editors, keyboard shortcuts are useful. They save a bit of time. In L3DEdit, they should be considered almost critical - while it's possible to work with mouse alone, it's extremely tedious. And don't get me wrong - keyboard controls alone are just as bad. Using both of them together is by far the fastest way to work - in my case, for example, I almost never change the locked axis; instead, I use the mouse to select blocks on the X / Z axes, while using the numpad controls 5 and 2 to move along the Y axis - this is my biggest timesaver, and is a huge part of how I'm able to create my levels relatively quickly.

It's less 3D, and more multi-plane 2D.
Lemmings don't walk at diagonal angles. They can only turn in 90 degree increments. You can never have a situation where two lemmings slowly converge or diverge from each other while walking in (relative to themselves) a straight line.

This means that Lemmings 3D is less a 3D game, and more a game that has multiple 2D planes. Turners or deflector blocks are essentially instant-teleporters between the same position on two different planes. The setup of which planes intersect which other planes at what points, is then represented by a 3D visual. Of course, it's a bit more complex than this. Bombers can affect the planes either side of themself, as one example. But this kind of thinking does help a lot when coming up with level designs.

The "make something, then add a solution later" approach does not work nearly as well as it does in 2D lemmings
There are two exceptions here. The first is if you're making a level where the gameplay is entirely in 2D (eg. Lemmtris), or at least where any 3Dness in it is purely or near-purely cosmetic (eg. This Is The Army). The second is if you're going for a general "anything that works is fine" Fun / Tame type level. For both of these types of levels, it works fairly well. Otherwise, you run into one of two issues: an interesting solution is nearly impossible to find, or you produce levels where turner-based backroutes are problematic to a ridiculous extent. You don't necesserially need to have the entire level planned out, but you need to at least have some vague ideas. At the very least, you need to think about what the general path through the level will be, if perhaps not quite the specific solution that will be used to achieve that path.

Turners are a huge source of backroutes
You thought builders or even stoners were bad? Turners take this to a whole new level. As a general rule of thumb, except perhaps in maze levels (I mean "maze" in terms of the layout, not the graphic set, here) or very controlled indoor / pseudo-indoor levels (ones that have outdoor visuals, but involve navigating mostly through narrow corridors similar to how an indoor level would); if your level has more than one or two turners, it has backroutes (at least until you spend some serious time de-backrouting it, of course). Remember that as long as it's not on a builder bridge, you can use two turners pointing the same direction to emulate a blocker, or two pointing towards each other to hold the crowd in one place - and by combining it with a few other skills, it's even possible to save one of the turners when releasing them while also guaranteeing the lemmings come out facing the desired direction. On the other hand though, builders are not as backroute-causing as in 2D lemmings. Don't get me wrong - they're still a significant source - but not nearly as much as they are in 2D.

When building complex structures, it's usually much easier to build them from the bottom upwards... and it's always MUCH easier to get them right the first time
I think this is self-explanatory. The bottom-up thing is mainly due to that L3DEdit cannot currently shift to a below-the-blocks perspective. This is a feature I want to add eventually, but attempts to do so so far have produced fairly glitchy results. Don't get me wrong - I will be taking another crack at it sooner or later - but for now, bottom-up makes things MUCH easier.

Avoiding some of the borderline-glitchy physics behaviours does more bad than good
Don't get me wrong. You should avoid stuff that's outright blatantly glitchy, such as the jump glitch. Any consideration of these glitches should be limited to "how can I avoid them occuring during the intended solution" and/or "how can I avoid backroutes that use them". But for example, bashers taking away 3/4 of a block height instead of 1/2 at certain vertical heights - that's very difficult to avoid, especially if you're wanting to have bashers freeing blockers (try it, it doesn't go as smoothly as you might initially think). I think some of these are cases we're going to have to get used to.

Consistency between levels is also somewhat difficult
The obvious example would be an indoor vs outdoor level in the same graphic set. The wall will generally be made of blocks that may, on other levels, be destructible - but you sure as hell don't want these walls to be destructible. To some extent, every L3D level has to be treated as having its own rules. In cases where these rules may be (purely based on my own feeling) contrary to one's expectations, or simply a case where there might be no solid expectation either way, I try to demonstrate what the specific level's rule will be very early in the level. For example, one level I've made quite clearly requires bombing away some ice near the start, then using a turner on that ice. This exists purely to show "on this level, ice is not indestructible / not only destructible where there's a red X" - because there's no way to even get started on the level if that weren't the case.

Another one is an indoor Pyramid level, in which destructible blocks have cracks on them. To make this completely clear, the first one the player encounters actually has the "Bash Here" animation from the practice levels next to it. Not far beyond this is another one, without the Bash Here graphic, but with clearly no other option but to bash through it. Beyond that is where the level really starts.

I suspect custom graphic sets will be able to deal with this a bit better, as they can be designed with these issues in mind. This will likely only be a problem with official graphics. Heck, maybe even over time, we'll establish a general consensus on how to handle certain matters, and it won't be an issue anymore.

...But many of the usual rules still apply
Test things. Keep replays or descriptions of your solutions. It might be a new game with some new mechanics, but it's ultimately still Lemmings. Oh yeah, speaking of replays...

Check your replays every time you modify the level. Even if the modifications are purely cosmetic.
I haven't narrowed down the cause yet, but I have found sometimes, a replay will suddenly not work correctly, despite no physics-related changes to the level. The common factor I've noticed is that, in all cases, this happens on levels where I've made adjustments that shouldn't interfere with the solution since creating the replay file - in one case the only change was moving the preview pivot point, which shouldn't matter at all once in-game. It isn't universal that this will break the replay - often it doesn't, but it has happened more than once.