Hacker News Reader

Comment by tromp | Hacker News Reader

> Problems start very easy on day 1 (sometimes as easy as just asking for a program that sums all numbers in the input) and progress towards more difficult ones, but they never get very hard: a CS graduate should be able to solve all problems, except maybe 1 or 2, in a couple of hours each.

I think this wildly overestimates the programming skills of the average CS graduate. My estimate of the fraction of CS graduates able to do that is closer to 1%.

View on HN

I ran a private leaderboard for conference tickets this year and then surveyed participants afterwards.

Earning a silver star took less than 1 hour each day for almost everybody on average. The gold star was tougher, between 2-5 on average.

https://blog.carolina.codes/p/advent-of-code-ticket-winners

Based on the leaderboard screenshot, only 7 people even solved up to day 12. The later days tend to be significantly harder than the early days. Based on top 100 leaderboard times, ~2x as hard for silver star.

Yeah this…the conclusion I would have drawn is not that the gold star took between 2-5 on average, but rather, 85% of people who self-select as enjoying a programming challenge where not able to finish more than half the tasks. Not that that's bad — seems slightly above par for the leaderboards I was involved in — just that the conclusion is really misleading.

You’re free to draw your own conclusions of course. I can only tell you about the feedback I’ve gotten directly.

Even people who didn’t complete any of the challenges have told me they plan to work them over the course of the year.

We’re going to make this an annual thing and announce it at the next conference in August, so hopefully we will have more participants next year.

Nothing will change the fact that the people participating in this challenge are not a representative sample of computer science graduates, hence an evaluation of their performance will not really tell much about what kind of performance one should expect from the average computer science graduate.

I will say that one of the guys who tied to win the competition is a 2021 Computer Science grad from Clemson. He's not a brand new grad, but that's the closest sample that I have.

The claim was that all computer science graduates should be able to solve those problems in a few hours each, the counter claim was that only one percent of them is capable to do so. The existence of one computer science graduates capable of winning the contest does not really tell us anything and I think nobody doubted that there are computer science graduates that are easily capable of solving such problems. But the question was or is, are they the exception or the norm?

The problems are essentially an elaborate IQ test (follow instructions, pay attention to the actual input, do some googling to get some polygon, graph algorithms/libraries). No deep knowledge is required.

You don't need a CS degree for that.

~5% of those who solved the first 2 problems solved all 49 of them. Given that it is a significant time commitment to spend several hours per day every day for 25 days straight, more people could have done it, given the time. These are not some elite self-selected geniuses. 1/4 of those who solved 1st problem, hasn't solved the 2nd one despite its simplicity https://adventofcode.com/2023/stats

This would also have been my opinion, computer science graduates should definitely be able to solve those problems in a few hours. The longest time to first solution this is year was something like 15 minutes, an hour or two for 100 correct solutions, so they are certainly not really hard problems.

But personal opinion aside nobody provided any real evidence that this indeed the case and only that is what I wanted to point out in my comment.

There's still a huge selection effect. He could have been solving programming puzzles as a hobby for the past fifteen years for all we know, and his degree might have little to do with his AoC skills.

Yea, it gets tougher for sure.

Everybody who participated seemed to enjoy solving the problems. Many of them have said they plan to solve the rest over the next couple of months, just not on the contest timeline.

The average CS graduate doesn't attend conferences though. Enthusiasts do. And even bigger enthusiasts decide to compete for tickets.

Unlike Project Euler etc. where one really needs to be good at algorithms/math etc to make the most efficient code, else it will just run for days to brute force, most of advent of code can be solved with terrible algo as the input data is very small.

I think most CS grad students can solve Advent of Code. Some people, probably don't finish it not because it is hard, but probably because they lose interest.

In my experience solving several years of Advent of Code, this is only true for part 1 of most days. A lot of part 2 solutions rely on heuristics to be solved in reasonable amounts of time.

There are some from 2023 that aren't. Day 12 part 2 at least for me had one input that ended up being 95% or more of all permutations and would have taken at least weeks.

I failed to brute force day 4 part 2

That's because most CS graduates can't code. At all. So yea...

When I started my CS degree and one of the first classes was how to use Microsoft Word I lost my shit.

No idea what school you went to but the first class is usually intro to programming, nothing to do with ms word

This was University of West of England.

Reply on news.ycombinator.com