Hacker News Reader

Comment by bhrgunatha | Hacker News Reader

There are some plots [1] of the time taken for the first 100 solutions to part 1 (silver dot) and part 2 (gold dot). They're a good visual indication of the relative difficulty of each question and how the difficulty of the second part increases.

[1] http://www.maurits.vdschee.nl/scatterplot/

View on HN

Those plots are really comforting for me, as on the days where a part was difficult (such as day 20 part 2 last year) I didn't find a solution. Thanks, I'll be keeping an eye on them this year as things progress (or not!)

And that's for the top 100. I would assume that the differences would be even larger for the average developers.

Ooh, that's really interesting. And given how many folks here talk about dropping out as it gets harder, that suggests the later problems look easier than they are.

The weekday problems tend to be easier so you get a ramp up, but with drops in difficulty. A lot of people give up after one or two of the hard days and don't continue, but there's no actual requirement to do them in order. If they'd skip the hard ones they'd probably find several more solvable (for them) problems remaining in the year.

These plots only include the first 100 people. Those dropping out as it gets harder were unlikely to be be represented in the top 100 for later problems either way.

I like that the data clearly shows the server issues on day 1.

That's crazy. I have no idea how anybody can solve this problem in 258 seconds:

https://adventofcode.com/2020/day/20

Did they hope the corner tiles had unique edges and look for those? (which would be unlikely under normal circumstances, considering there are more than a hundred tiles and only 256 different possible edges...)

From the problem statement:

> Tiles at the edge of the image also have this border, but the outermost edges won't line up with any other tiles.

So they didn't need to hope; the problem guaranteed it for them.

Ah, sorry, I missed that. That does make it easy.

While I agree four minutes is very quick, if you take into account that no side fits with more than one other side, and that the sides (#, .) may just as well be interpreted as binary numbers, it becomes an issue of parsing, mostly, and then matching numbers.

I remember being flabbergasted at first, and then realizing the above. My solution is here:

https://github.com/linus/aoc/blob/main/2020/js/day-20/app.js

So it turns out I missed those parts of the problem statement. That makes it fairly easy. I probably still can't write working code for that in four minutes, but it does sound humanly possible, at least.

Also, that's really nicely written code :)

Funny you mention that, the person who solved part 1 in 258 seconds recorded themself solving that one live: https://www.youtube.com/watch?v=ODUkgEeF7sc

Wow, it took me longer than 258 seconds just to understand the problem statement. I had to read it several times before I understood what "border" meant, as it does not match common usage. I noticed that you referred to them as "edges" instead which would have immediately made sense to me.

It's not just measuring your problem solving abilities but also your famiarity with the author's writing style.

Just curious, what is the difference is between "Border" and "Edge" to you? To me, they are nearly synonyms.

Apple's Dictionary (Oxford?)

> 2 the edge or boundary of something, or the part near it: the northern border of their distribution area | figurative : the unknown regions at the borders of physics and electronics.

https://www.merriam-webster.com/dictionary/border

> 1 : an outer part or edge

https://www.dictionary.com/browse/border

> the part or edge of a surface or area that forms its outer boundary.

Is this really saying for 2021, the first solution to the first problem came in 30 seconds after being posted? That's insane!

People are incredibly quick at both skimming/parsing the problem, coupled with a language like python where you can generally express stuff incredibly concisely and not have to worry too much about types, it makes it very quick.

I'm in a really bad time zone for AoC right now, but here was my solution in R for day 1 problem 1.

    library(tidyverse) # pre-typed
    lines = readLines("data/day1.txt") %>% as.numeric() # Read data
    diff(lines) %>% .[. > 0] %>% length() # Part 1

This didn't involve me being a sophisticated or good programmer, and if I was in a time zone to be competitive I also would have written a function to submit my answer from within my IDE (which can also be pre-written). I could absolutely buy people wanting to be competitive and being able to do this in 30 seconds. I actually could have pre-written the second line of three, in point of fact.

whoa I didn’t even consider automating my answer submission, suppose it’s just a curl away

...is the y axis showing DD:HH:MM or HH:MM:SS?

HH:MM:SS, the first 100 solvers are usually done within minutes, hours at most, and never days. None of the problems are that hard.

wow 0:

Reply on news.ycombinator.com