Proposal: Level Solving Competition

[Minim]

Hello everyone.

So basically what I'm proposing here is not a level design contest, but a level playing competition. So far for playing levels the competition always seems to be taking place in multiplayer, but what about single-player? Here are some rules as to how this contest is planned to work:

* The engines that can be used are NeoLemmix and Lix, although they can't be used together because of different mechanics. Both also have different recorded data however; for example NeoLemmix's time taken to Lix's Skill count.
* One level will be shown at the bottom of the opening post for users to download and play.
* The aim is simple: Just solve the level.
* The competitive part is to try and solve it within the fastest time in the case of NeoLemmix, or the fewest skills in the case of Lix. The total Lemmings saved factor could be prioritised depending on competitive viability. Would a maximum requirement reduce the number of viable solutions to the level?
* Both engines have an "auto-save level solutions" kind of button, so make sure you turn it on so you don't lose your replay!
* The process to submit results is similar to Family Feud. Upload your best replay to the host's personal messages box, not on the thread. Just to remind the guests that you need to be logged in to send PMs.
* The competition's playing phase will be open for let's say a week. After a week expires the user's Lemmings saved/fastest time/fewest skills etic. will be displayed on a leaderboard.
* The winner of the tournament gets to choose the next level for a competition. Any additional prizes such as the ones from the Level Design contest can also be suggested to be included here.

The poll above this post is a list of two engines. Cast a vote as for which one you think would be more viable to use for this competition. NeoLemmix is more popular but Lix allows more skills to be used in-game. If we want to use Lix then you'll need to download the latest version. The boards for other Lemmings Projects haven't been updated since November, so we'll leave SuperLemmini etc. out in the meantime.

So, I hope this new idea turns out to be a popular one. Here are some example levels off the top of my head which I think are not too hard but have enough variety in skillset to deserve competition. However as always, I'm open to any suggestions or questions.

NeoLemmix: Original Lemmings
* Fun 29: Worra lorra lemmings
* Tricky 10: There's a lot of them about
* Tricky 11: Lemmings in the attic

Lix: LemForum Pack
* Lovely 4: Diamond Dash
* Lovely 20: The Borderland
* Quirky 1: The Adventure Playground

[IchoTolot]

I don't think the fastest time would be a good criterium. This would result in people fiddling around their solution to cut off single frames. I think this would kill all fun, as it's more about fiddling around with the found solution to make it as efficient as possible instead of finding it and actual level solving.
This would result in this being more like a speedrun competition instead of a solving one. I definitely have no interest in trying to speedrun, sry. It's not my kind of fun.

The fewest skills would result in the competition being in need of a very specific kind of level. Most likely an x-of-everything level where searching for the route with the fewest skills is the key. The most efficient one being very hard to find.

I think the ruleset must be overthought as a result.

Also I think a bunch of new, unsolved, never before seen levels would be better. Otherwise it would be possible to look up speedrun/optimal solutions before the start.

Maybe even something similar to the live L2 races would be more fitting.

[Simon]

You can count skills in NL manually.

Fewest skills will produce many equally-ranked entries, you'll want to pick massive maps with 15 or more assignments.

Lix can display physics updates during play: Options -> Graphics -> Count physics and FPS. I didn't expect broad interested in optimizing solutions for physics updates, therefore the post-play stats never display physics updates used. I'm open to discuss stats displayed. Don't let either game's shortcomings dictate the contest's design; if stuff is important, we can work around.

TASing (minimizing the number of phusics updates) has niche appeal. I would be up for it. I also assume that some more will be interested, especially for a one-off event. But it will be fiddly.

I'm also interested in improving Lix's replay editing, e.g., move an assignment by one physics update sooner or later, while instantly seeing the result 200 physics updates later. I still lack insert assignment without cutting replay afterwards.

-- Simon

[Proxima]

NeoLemmix is more popular but Lix allows more skills.

NeoLemmix has more skills than Lix (17 to 15).

We have existing challenge topics for fewest skills and least time on the original levels, and fewest skills is already heavily optimised. New results are always possible, but I don't think a selection of three levels per week is likely to turn up very much.

I would definitely be interested in a challenge topic for fewest skills on the Lix community set (and other stable packs such as Rubix's), but I'd want to see whether other people are interested before starting one. It has the disadvantage that to an extent, the levels are still in flux (especially as regards order), so a bit of extra maintenance work would be required, but I don't mind doing that.

You mention prizes such as the ones from the level design contests. Are you prepared to fund those?

[Simon]

More ideas, some from Forestidia in voicechat.

People optimize a replay, then send this replay privately to the host as in Family Feud forum game. How would one even be able to cheat then? You can't copy other people's solutions before scoring.

The host can design a level in private, then release it when the round begins. This gives everybody an even start except for the host; the host knows their own map already. Still, the host should be allowed to enter their own contest for fun. Maybe preclude the host from winning.

If one prefers discussing ideas over working alone: Can several people enter as a team?

I'm willing to extend Lix to display more stats, ingame or postgame. Replay-editing features would be even better but I'm busy these weeks, can't guarantee the time to develop a large feature. Still, I'm happy to discuss anything.

-- Simon

[namida]

NeoLemmix has more skills than Lix (17 to 15).

I think he means how many different skills a single level can have. If I'm not mistaken, a single Lix level can contain almost all 15 skill types in a single level (aside from the two different bombers). NeoLemmix may have 17 to choose from, but any individual level is limited to using 8 of them. Although I do have to wonder if this may need to be expanded, especially with even more new skills on the horizon - not perhaps to the extent of no limit, but maybe expanding it to 10 or 12. Even prior to the introduction of the Fencer, I've hit the skill type limit from time to time when designing levels. Assuming not too much has changed since I was maintaining it, the biggest difficulty with this would be fitting it all onto the skill panel nicely, especially for the compact layout - outside of that, IIRC it's just arbitrary hardcoded limits.

[Crane]

It might be a good way to get a brand-new level field-tested for a level pack, but which doesn't fit the criteria of one of the design competitions.  I myself have posted a couple of levels for people to play, but they haven't generally gotten much attention except from Nessy and IchoTolot.  It is fun though, taking a brand-new level, never seen before, and trying to solve it.

But yeah, the 'shortest time' category is asking for trouble (have you seen Super Mario Bros speedruns?!), and 'fewest skills' kind of rules out levels that are designed to use all the skills (e.g. "No added colours or lemmings") unless they get backrouted.

I don't know what the competition categories would be in that case... "most enjoyable" or "most creative" are fairly nebulous categories.

[Minim]

I think the ruleset must be overthought as a result.

Also I think a bunch of new, unsolved, never before seen levels would be better. Otherwise it would be possible to look up speedrun/optimal solutions before the start.

Maybe even something similar to the live L2 races would be more fitting.

IchoTolot is right about the fact that I've overthought the ruleset. It is a proposal after all though, so I think it is important to put all that info out there for the community to discuss.

I've never seen an L2 live race. It is interesting! I'll see if I can find a video and see what I think of it.

You mention prizes such as the ones from the level design contests. Are you prepared to fund those?

Unfortunately, I don't know how the process of the other two works. That's why I thought if this idea is successful I could consider these sorts of prizes. I'll scratch that one for now. If there's enough demand then someone could PM me how they work.

People optimize a replay, then send this replay privately to the host as in Family Feud forum game. How would one even be able to cheat then? You can't copy other people's solutions before scoring.

The host can design a level in private, then release it when the round begins. This gives everybody an even start except for the host; the host knows their own map already. Still, the host should be allowed to enter their own contest for fun. Maybe preclude the host from winning.

If one prefers discussing ideas over working alone: Can several people enter as a team?

I did think about both of the first two points about an hour after I've posted it. I've realised myself that the Family Feud PM process is a fool-proof way of people being unable to cheat. The second point certainly makes the host feel disadvantaged because they can just pick out the best replay and submit it as theirs.

I don't understand the team prospect. So do you mean something like two teams should enter this competition and the winning team takes the shared prize?

[Simon]

I've never seen an L2 live race. It is interesting! I'll see if I can find a video and see what I think of it.

Last year, I streamed on twitch some races against geoo. I didn't highlight any of those, and thus twitch has deleted them all.

I've saved one clip though: Mud Slinger backroute in Quest From Kieran 2; after the stream, Kieran fixed this backroute.

I don't understand the team prospect. So do you mean something like two teams should enter this competition and the winning team takes the shared prize?

Not restricted to two teams. Like Family Feud, the basic idea is that each player should work alone, then submit the best replay that they could make. Many players can enter like this. This rule will stay.

In addition, can two people cooperate in private, then submit a single replay? This team of two will behave like a single player from the contest's point-of-view.

-- Simon

[Proxima]

So it looks like the main thing to decide is: are we going with Simon's idea of the host making a level for each round, or minim's original idea of using original / existing levels? I would vote for new levels, as I've already given my reasons why I don't think original levels would work. In that case, who would host the first round?

[MandelSoft]

Some proposals from a forum lurker...

Instead of taking just skills and time as a whole, why not take multiple factors into account and assign points to them?
For instance:

• Each second on the clock is 1 point
• Each skill you use is x points (possibly depending on the skill)
• Every lemming that doesn't make it to the exit is 10 points.
  

So, in this case, the goal is to minimise the number of points, and that would be deemed the most efficient solution. One has to fiddle with the weights of the points, though, but you could do a few test runs on a few levels to see what the result is. But the system would provide a nice metric to compare the results and still make it easy to dertermine which solution is better. You could even say in public you've found a "300 point solution" without revealing a single bit of your solution

As for the level choice: I'd either go for a completely new level, to level the playing field (no pun intended), or use an existing level with a twist, so that the most common solutions don't work any more. For instance, by changing the skill set or adding a one-way wall or steel where there used to be none.

Just a couple of thoughts...

[namida]

At one point, NeoLemmix calculated and gave a score based on those kind of factors (although the algorithm was slightly different). It wasn't a very popular feature.

[geoo]

I feel both blind race and skill optimization could work well as formats.

Blind race is what I did with Simon for most of Kieran's L2 levels, and I think that's a lot of fun. Someone just needs to release a new set of levels to play through for this, and ideally everyone participating should try to play at the same time, so logistically it's a bit trickier.

I like the idea of optimizing for skill usage. The previous contests just judged on whether you solved a level or not. Number of skills (with number of lemmings saved as tie-breaker) is much more fine-grained. I personally think even on existing levels (Adventure Playground for example seems quite suitable) it could work, I would guess for most levels non-one has put really serious effort into optimizing skill usage. A different idea would be different people submit a level each, and then the contest is played on these (number of skills added up over all levels). Coming up with such levels probably wouldn't have to be too much effort as you are not disadvantaged if your level is easy and open-ended, and the solutions people will find might make for interesting levels after applying the skill restrictions.

[Simon]

I'm fine with one stat (either physics updates used, or lemmings saved) as the main stat, and possibly one or two other stats as tiebreakers.

I discourage elaborate scoring methods that compound many stats. Those make the endeavor more complex than necessary. Or do you feel there's a problem with a simple stat to optimize?

-- Simon

[Minim]

So it looks like the main thing to decide is: are we going with Simon's idea of the host making a level for each round, or minim's original idea of using original / existing levels? I would vote for new levels, as I've already given my reasons why I don't think original levels would work. In that case, who would host the first round?

So far by looking at the poll, the majority are choosing NeoLemmix over Lix. Looks like I'm the favourite to be the host, as I've come up with an idea for a level which is still a WIP, while trying to take as much advantage of its capabilities as possible. Therefore I'm leaning towards Simon's idea of creating a new level, which I've disregarded by accident in the OP. Keep up the discussion though! Every post helps as I (or another designer) try to make the final product.

I realised that I can record Fewer skills in NeoLemmix; well, to make things easier for me I'd rather invert it and record the number of skills saved.

Lix can display physics updates during play: Options -> Graphics -> Count physics and FPS.

I just tried it and I find the numbers going up too much of a distraction! At least it stops when it pauses, and that I can record which frame the success sound plays.

A different idea would be different people submit a level each, and then the contest is played on these (number of skills added up over all levels). Coming up with such levels probably wouldn't have to be too much effort as you are not disadvantaged if your level is easy and open-ended, and the solutions people will find might make for interesting levels after applying the skill restrictions.

Because I want to do more than one contest, I disagree with the idea of compiling all the levels together for one contest because I fear that parts of the solution to the next level would turn out too similar to the previous. Because we're still in the early stages I think one large level would be enough for the moment. We'll see how the results turn up though. If they end up too similar then we could go ahead with your idea.