Racing Consistency

[faytleingod499]

While I managed to make a drone that consistently finishes the race track in around 11.7-11.9 seconds, sometimes, it will just suicide against the wall on that famous hairpin turn. It will go like 10-15 races just fine, then suddenly it decides to just not turn. It's very aggravating that there is so much what appears to be randomness in the performance of your drone.

[Hex7]

I always viewed the randomness as sort of a mixed blessing. While it does occasionally screw you over, it also occasionally falls in your favor and pulls a win out of the jaws of near certain defeat.

I think I'd hate it if there were only one possible outcome to a given race.

-----

R.I.P. walled drone. It's racing with RNGesus now.

[faytleingod499]

I see what you mean, but it also kind of eliminates fine tuning. When you're trying to squeeze every last bit of efficiency out of your drone, randomness makes it:

A: Tell if what you did makes it better or worse and

B: More or less eliminates any good side effects of doing said fine tuning.

Sure, it can eliminate the bad side effects too. But having a large degree of control removed in favor of random chance during a competitive event kinda of makes me not want to even bother. I can face off against an obviously inferior drone but still lose because my drone decides this match to do things slightly differently? Seems kinda broken 😕

[Markus]

I see what you mean. I think randomness can be a fun thing as long as randomness is a smaller part than skill and does not screw you badly/regularly.

Ideally we would have a deterministic physics engine (which unity does not provide currently) and a physics engine which is optimized for such complex rigidbodies as nimbatus drones are.

Unfortunately the unity physics engine does not provide any of the above, so we have to work around those limitations and design game modes which are still fun even with all those drawbacks.

If you're micro-optimizing in the milli-second region, I agree that the randomness in Unity physics is a big factor and limits you. Unfortunately that's the technical limitation we have now  

[faytleingod499]

I see, well if it's the limitations of the engine, that's understandable. I do very much enjoy the mode regardless.

[corona_wind]

22 hours ago, faytleingod499 said:

I see what you mean, but it also kind of eliminates fine tuning.

The racetrack is the worst place to fine-tune, can't zoom in, can't slow it down, just have to throw things at the wall monte carlo and see what happens.  Sometimes you have to mock up a craft with a lot of LEDs and control it with cursor to see how it behaves to get what really happens.

I think the game would be a lot duller if it weren't the slightest bit random.  You have to make a bot that can respond to a variety of situations, not just a bouncy puck with nineteen VERY carefully aligned engines.  And a three-star arena can't block you forever.  Probably.  Unless its champion is REALLY good.

Also...  Some of my most fun designs come from seeing a drone hit the wall, lose half its parts, and keep going in a new way.  Have to stop and think, "How did it do that!?  ..can I do that on purpose?"

[faytleingod499]

I think the game overall would be duller without a little randomness too, but I think racing specifically demands precision. I don't think that it would detract from the fun of it at all. You see a good design and you try everything you can to beat it. It forces you to get more creative with your design and in the process, figure out how to manipulate the physics just a little bit better than your opponent. Instead though, good just means you make something that hopefully will randomly break itself less than your opponent. I've beaten many opponents that were technically faster because they got unlucky and blew up (usually on that tight turn). 

[corona_wind]

Death zones are very capricious and the one on that corner catches everybody.  Make it deadlier and racing will be way more deterministic.  Dodge or die.

Besides that, most "random" breaking ought to get cleared up once we have proper slow-motion in testing.  It took me hours to realize my back engines were hitting my front ones, for example - it wasn't random, but the only testing method I had, trial-and-error, made it look so.

What's left are butterfly problems -- hitting that one-in-a-thousand logic hole, riding the edge of ramjet shutdown, knocking that one crucial gate off your nose, etc.  I don't think these can be avoided without 100% pixel-perfect determinism, which would suck.  But all these are design problems, avoidable.

[garr890354839]

You get around the randomness with trial and error. And broken logic can be solved with a detachable logic chip with all your logic gates and stuff on it.