As each time, I've tried to think off and try out new concept for Ludum Dare (as I don't really see any other point in it anymore). This time I've went for building quite complex physics simulation and game around it, just for checking what feedback can I get for such game.
I'm actually quite glad for the feedback, as it was mostly positive and often praising technical side and innovative idea. Yet some pointed out that the game is "too simple", the original idea was making it more complex to control, yet look at the ratings, compared to previous years:

So, what are my findings (and why ratings are lower compared to previous times), plus what I've learned:
Complex Physics/Simulation driven games are getting torn down on Ludum Dare
From experience, only simple games are rated high on Ludum Dare, the more simple you go, the higher you get. This can be seen from TOP games each Ludum Dare so far - when games are catchy and easy to understand - players beat them and therefore give high score. Whether I will go more simple, or more complex next time is up to a question - depending on the theme, how fast I can get game idea and how much time will I have to actually work on game.
I won't ever do any WebGL build anymore in future Ludum Dares
This has 2 reasons actually, I won't tell you the technical one (not until the end of post) - but just the motivational one. We've got most votes in previous runs based on desktop one (for LD38 we released WebGL late, and I believe that if we didn't, we'd reach top 3 places in graphics). This time I released WebGL almost immediately - even though you specify people to use Desktop version, due to heavy problems with textures, shadows, etc. when running WebGL version (they are toned down to extreme level - to keep performance in browser acceptable) ... majority played the WebGL version only.
Next time I'm going to do some more though integration & checks in game
This time I ran for high scores. I could match some people in comments, and that made me quite happy... I appreciate any feedback I get, as on Ludum Dare one has a chance to impress various target players, which isn't that easy in reality, as only target gamer group will buy your product.
So why tighter checks and implementation of let's say achievments, high scores and maybe secrets? Because of 2 things - first of all it helps me match feedback to actual progress of player (assuming he gives out his nick), and possibly finding points in game which frustrated him or where he quit. And apart from that, people here tend to love high scores - so why shouldn't I bring that to next level?
So, am I going to participate next time?
Sure, why not. Also I'll do some small shameless self-promotion - next time I'm going to participate with custom game engine and not Unity, based around Direct3D 12 (Windows 10 only).
If you would like to see some progress from it (mostly graphical) - you can check me on twitter (although my activity isn't nearly as high as Donald's) - https://twitter.com/VilemOtte