Hi there.
While this is my first post here, I've accumulated a little design experience with other CRPG editors (mhmm... is it a thousand hours already? then again, do I want to know?). Thus, I already made most mistakes one can make - perhaps my experience can help you a little
Veterans, feel free to skip as appropriate...
1. - Know your resources
Time, energy, creativity. A well-made and complete five-level dungeon is worth infinitely more than an unfinished Moria, no matter how brilliant the vision behind it. Don't laugh about the creativity issue. I've known people who were great puzzle designers, but lousy storytellers, and vice versa.
When in doubt, seek help.
Second, of course, are the technical resources like wallsets, monsters, items, scripts - know what is possible (and what isn't). The reason for this is point 2.
2. - Do something unique (see Neikun, above)
The most important currency on the internet is attention, and you will need to promise players something that is worth their attention.
Pick an area where your creativity excels, then use the - excellent, I might add - technical resources available to support it. Even though Grimrock is about puzzles at its heart, I'd say people would forgive it if your puzzles aren't all that creative if you make up for it with a storyline or a secret that makes the players drive deeper into the dungeon because they just need to find out what's behind all this. (Don't disappoint them then, though.)
3. - Planning
Many aspiring dungeon designers only plan the rough outlines of their levels, reasoning that this way they will be "exploring" the dungeon as they create it, so the process will be more entertaining to them. That is not a bad thing, but you must realize that with this approach, things will look not as coherent as you might like them to be. You will need to leave a lot of loose ends which you will have to fit together in the end, instead of being able to lay out clues right from the beginning that lead to the great revelation in the end, which is obviously the more exciting experience for the player. Second, if you tend to create things out of thin air, you will regularly be stuck without an idea to harvest. Planning thus actually will save you a lot of time in the end.
4. - Puzzles
Balancing puzzles can be incredibly hard and will likely require testing, but here are some guidelines I've found to be true:
- Both the goal and the rules of the puzzle should be reasonably clear. Having lots of buttons and pressure plates affecting things the player can neither see nor hear is a great recipe for frustration.
- Give hints to the players. As said before, teaching them with easy puzzles of a similar type at the beginning is a fine way to do this.
- Making players rely on trial and error is only suitable if there's a limited number of options, and trying each option can be done without expending a lot of resources (and yes, playing time counts).
5. - Options
A good way to avoid having players stuck at one puzzle is making the area behind it optional. There is much design space in here and as of yet I've not seen a lot of it implemented in Grimrock and its mods, probably because it's against the spirit of the hardcore dungeon crawler. But what if, for instance, there was a crucial door with two locks and not two, but four of the appropriate keys placed in the dungeon? This way the challenges that seal them could be quite tough, as the player wouldn't necessarily have to bypass all of them, but instead only the two he would feel most apt solving.
6. - Player Resources
Not to be confused with 1.), but easily overlooked. Creating the evil vault for a super magic sword is fun, but you should make sure players have the resources to cope with it. This includes food, reagents, missiles and lesser weapons. You have a lot of pressure plate puzzles? Throw in an additional rock or five. And as said before, assume that players will not find each and every hidden cache.
7. - Immersion
Neikun has said "it should make sense". This is one of the parts of an immersing experience - you want players to get caught within your game. Interestingly, the whole dungeon crawler genre has a certain reasoning problem. As stated in "The Mine of Mayan Vael": "Why would a necromancer leave riddles in a mine?" - The original Dungeon Master didn't have that problem, as with its real time mode, the "living" monsters and other features it was actually one of the most immersive games at its time. Legend of Grimrock bypasses this issue very elegantly by establishing early on that the whole complex was built by alien minds for an alien purpose and thus is not to be measured in human terms of reason. While it has become commonplace in the genre to just ignore reasoning and include puzzles for puzzles' sake - which is fine - giving a reason for the setup of your place will give it a special quality.
In the same vein, consider the layout of your levels. If you create, say, a castle or a similar structure and its map actually looks like that of a castle and not like a brainmap of the Joker - more points to you.
Finally, details are your friend. You can use the "Frozen Temple" wallset and call your level "Frozen Temple". You might also call it "Temple of the Frost God Ymir", include a few notes with made-up religious lore and make the player's characters give off some appropriate statements when they walk the area. Yes, the latter one's more work. But not that much - the effort/effect-relation is very good on this part.
So much for today, this is already way too long. But maybe it helps.
