Discuss ideas.

You are a leprechaun. With gold. You must hide your gold from other players by constructing an obstacle course. The game has two modes of play: design and play. The levels themselves will be Mario-like and will feature several mechanisms for obstacles (ropes, traps, enemies, etc). At the end of the level is the pot of gold that must be stolen by other players. When you design a level, you can make it as convoluted as possible. However, you must play and beat your level before you can upload it to the server. Once you upload your level to the server, other people can play the level and try to steal your gold. There will be scoreboards for the best gold thieves, hardest levels, etc. This solves a lot of problems: the multiplayer aspect is extremely limited, but still is the focus of the game and adds richness (pun, haha). No other users have to be logged in to play and you can still interact with their levels without them being online. We can seed the server with a dozen or so levels so that people can see what sort of things the engine is capable of. Additionally, the fact that there will be an editor in addition to normal gameplay will not necessarily add any extra stress for development since every PyWeek, I generally create a level editor anwyay (except in C#). Also I pulled the leprechaun/gold theme out of my ass. This could literally be anything. Also, this mechanism could be highly adaptable to One Way Trip. 

The game idea is fine, but I don't think the scoreboard will work very well. Qualifying the "best gold thieves" and "hardest levels" is going to be tough, especially quantifying level difficulty. Using "number of solves" is not enough. The amount of time it takes to solve a puzzle will have to be compared relative to other solves, and I don't think we can put together enough samples within judging weeks. Likewise, the best gold thief probably solved the most difficult courses, and that too needs a statistical formula.

Hmm...Attempts vs Success? Could average that on a per-user basis and per-level basis. How you fall in the distribution of other users who also tried to solve the level could give you a rating. Doesn't have to be entirely perfect. Alternatively, the amount of gold wagered on each level could just be 1 and go up by 1 each time a unique person fails. Once someone succeeds, it falls back to 1. Or something.

This idea is ikanreed's.

You have two pools of stuff (possibly more for more players) at opposite corners.  At the bottom of the screen, you have a collection of tetris-like pieces (somewhere around 3-5 of them, randomly chosen) in your color, as well as one bomb that you can use.  The pieces have pipes in them, and the goal is to get a pipe to the other player's stuff and take it

On your turn, you can place a piece adjacent to your pool or to your other pieces.  There must be a complete path from your pool to the piece you are placing.  You can also place a bomb.  The bomb must be within a few tiles of one of your pieces, and blows up everything in some radius.  It can open up a way through, but it can also put you at risk as well.

This is my idea.  It's part building and defense, part exploration, part infiltrating the other player's areas to take their stuff.

The world is procedurally generated and everyone builds in the same world.  You have a bag of stuff you are trying to protect.  You can move it around, which might be a good idea if you find a more defensible position, but you can't take someone else's stuff while carrying it, and you can't sit it down too close to someone else's stuff.  I'm imagining the setting and technology to be somewhat futuristic, but that's probably flexible.

You will be able to build various defensive structures.  Walls, traps, turrets, robotic guards, stuff like that.  You will not be allowed to build a wall that fully encloses your stuff and makes it impossible to reach.  If you destroy something, you get back a portion of the resources used to build it (but you cannot get money back).  You will have resources like wood, stone, and metal for building particular things, as well as money which is the most powerful and can be used to buy any other resources or items.

Items are things that are generally one-use or limited use.  Items might include healing kits, trap detectors (multi-use, limited charge, highlights traps within a range), recharging energy shields (can be broken), random teleport (teleports you to a random space in the world (but not too close to someone else's stuff, there's gotta be some challenge), useful for finding people's stuff far away, one use).

In addition to building, a big part of the game will be exploring.  You can find resources for building out in the world, items, and most importantly, other people's stuff.

If you find someone's stuff, you get to pillage it.  When you pillage someone else's stuff, you gain some money as well as some points.  Points are just so there can be a high score board.  Taking someone's stuff gets you points, having your stuff taken removes points.  Your score can be negative if your stuff isn't well defended.  

If someone has a low score, their stuff is worth less (therefore if they are pillaged repeatedly, it is worth next to nothing).  If someone has a very high score (and therefore should have more resources to build better defenses), then their stuff is worth more.  This should help balance things and allow people to catch up.

You will have certain abilities you can use.  Disabling traps (if you are aware of their presence), dodging things shooting at you (all about timing, you don't control which way you dodge, so be careful of traps), and shooting.  You have limited health, and when it's gone you are teleported back to your home.  You can also teleport home whenever you want, in case you get lost.

When breaking into someone's place, you can shoot and take out their robots and turrets.  They rebuild themselves after awhile though.  This is to keep someone from getting pillaged repeatedly after their stuff gets broken.