Starting from today, the code size in these two formats will take into account the number of bytes, instead of the number of characters. One of the main reasons behind this change was to avoid some unfair methods, the most famous one being the use of minifiers in Python (e.g. GitHub - clemg/pythongolfer: Code golfer and minifier for Python https://clemg.github.io/pythongolfer/), so it should not impact how most of you play these games
This change was also done to prepare for another big change: we will add a lot of new Code Golf puzzles in the following days! Keep a close eye on the code golf page
Does this affect the code size limit for normal puzzles / multiplayers? A lot of the neural network solutions rely on this being a character count to fit the network weights in.
I don’t think there is some unfair methods in golf :p, as everyone play in the same environment. It’s a bit of shame to have just remove char scoring (at least in puzzle section) because it’s also an interesting metric. As you said there is the famous python trick but there is a lot more, it’s use in multiplayer and you kept it. Some impressive solution are now just gone. Perhaps a double leaderboard could be interesting ?
Speaking about “unfair”, there is also random and/or hardcoded solutions. At least temperature, thor and don’t panic are missing test cases and a lot of top solution could be qualified as “unfair”. Because you plan to add new golf puzzles, please pay close attention to test cases. It could be interesting to replay solutions sometimes, it seems to me that Thibaud was doing it.
it’s still the same environment for everybody, that environment has just changed. python is already shorter than most other languages usually so i don’t have a problem with eliminating this particular trick
True for The Fall and Vox Codei 2, they seem very hard to complete at all in the first place, let alone try to golf them.
The good thing is that it will motivate us to complete them - I wanted to to them for a while but never had enough motivation.
As there are a lot of puzzles in Code Golf now, I suggest rating the best place for a puzzle in any language instead rating all places for a puzzle in all languages.
e.g: for a Thor puzzle I’m 10/1000 in python, and 11/2000 in js, then I should get points only for js.
Currently it keeps your 5 best scores per puzzle, which is fine IMO, you don’t have to do all languages, just find 5 you can be competitive with.
So the max you can get per puzzle is 1000 CP, it seems good, not too much but enough to motivate you to ‘invest’ time on a puzzle.
Keeping only the best score would mean 200 CP per puzzle so a grand total of only 5400 CP for 27 puzzles, it doesn’t seem very rewarding.
There is another pattern used on the top of the leaderboards: “just find 1 you can be competitive with, and use it with system calls in all the other languages”. But this makes the per-language leaderboards rather pointless unfortunatelly.
Fully agree with you on that, the real issue was not the chars vs bytes thing but the system calls.
There are languages like C that I love golfing with but I don’t because system is more efficient.
Of course you could say " if you don’t like system, don’t use it" but then you cannot compete at top level - it was my policy for a while but I was stuck at some point by refusing to use it.
Any big guy to dm me about the sweet concepts of “Hardcode” and “System calls” ? Just to teach myself those tips. I’m always curious to learn something else
(like pardouin I didn’t hardcode any new puzzle).