In previous years, he was the MC at the closing ceremony. And guess what, he was annoying af...

So answer is like 1 or 2 ACs maybe?

Thanks, fixed the typo!

I agree that clarification requests about this issue should have received responses (though for several reasons I'm not convinced that it makes sense to respond to clarifications about things that are clearly specified in the problem statement).

Thanks so much for the thoughtful reply--I really appreciate your willingness to engage with my comments. A few responses:

I would also love to hear your feedback on the harder end of the set (E, G, H, K)

I haven't looked at these problems closely and probably won't until I'm more emotionally removed from the contest, but happy to send you a message if I end up spending time with them in the future :)

I find your classification of "the eight problems approachable for all but the strongest teams" somewhat confusing.

I didn't phrase this especially well--what I meant to say is that these were the only eight problems solved by a large number of non-gold medalist teams, not that every team could approach these eight problems. (My main point is that these eight problems were the main ones that affected most teams' experience at the contest, which is why I felt okay giving feedback without full information on the four hard problems.) I agree that J/L are in a different category from the other six.

If we did this right, the implementation cost of L is not horrible, and the core of the problem becomes "did you think through your solution well enough to notice the edge case" (and maybe then the problem does actually belong with A and D).

I think I agree with this--I'd still feel like L contributes to the implementation-heavy nature of the problemset, but in isolation I wouldn't mind it too much if every team knew about the testing tool and if there were more other problems that involved more thinking time/less implementation time. (The main other thing I don't love about it is that "did you notice the edge case" ended up being a significant determinant of the medals, and stylistically I prefer for awards to be determined by "did you come up with the substantive idea for a hard problem". I think WF22 R, WF23 B, and WF24 J are all problems I like better as medal deciders.)

In this case, the average time to get those six problems was over 3 hours, which I agree is too much.

Agreed. I do think, though, that even if the easy problems went faster than this, it'd still be nice to have more novel ideas since the weaker teams will spend most/all of their contest on these problems (and even stronger teams can enjoy easy problems if they're nice). WF22 Q and U were good examples of this--they were solvable by most teams, but the ideas were novel and they're still interesting to solve for very strong teams.

Another thing you say is that the "hard" part of the contest contained two problems which you see as "kinda easy" (J and L), and not super-interesting. I'll comment a bit more below.

I don't think I'd characterize J this way; it felt like an appropriate difficulty for where it was in the set. My only issue with J is that data structures are naturally a little repetitive/implementation-heavy, which is fine if they're some of the most implementation-heavy problems in the set (which is usually the case). I just didn't love another standard-ish problem in the context of the other, more standard problems in this set. (In isolation, I think J is on the better end of DS problems and I think it'd be a good problem in most ICPC sets.) Re: L, I don't think I'd characterize it as easy in general, I just think the vast majority of the difficulty comes from the implementation and noticing the edge case and not from the general idea.

Another thing that surprised me is how well J went.

I think a lot of the stronger teams, especially those with former OI contestants, are pretty well-trained on problems where you take a segtree and basically use it to encode a DP state. (Nile from this year's IOI is spiritually similar, as a particularly recent example.)

I wonder how much would your opinion of the set change if we removed some two problems from the medium-easy end (say, I and F, or A and F), and made sure that the testing tool for L was clearly marked

I think these would both be good changes (especially clearly indicating the testing tool for L). I think if we knew about the testing tool, my feelings after the contest would mostly have been disappointment in my own skill issue on J (to be clear, I think J was a pretty good problem, I'm just mad at myself for messing up on it) rather than frustration at missing L for reasons that felt a little bit out of my control.

I do think -F and -some other easier problem might have left some less experienced teams without enough approachable problems to do, so ideally some other easy problems would be added in their place. (That said, my team finished the easy problems in a little under two hours, so they weren't as much of a bottleneck for us as they were for a lot of teams. Thus I wasn't upset with them keeping us from the harder problems, it was more that I was hoping to see some cool new ideas in the early half of the contest.)

I wonder which of the problems in the medium-easy part of the Luxor contests you liked

WF22 Q, U and WF23 H were all especially good easy/medium problems (and I didn't especially mind any of the easy problems in those WFs except maybe WF23 I, which felt a little silly). Turning Red was a little standard, but other than that I felt like the easy-medium problems all involved original ideas (and I didn't mind Turning Red because it was the most standard thing in the set).

Some feedback on the medium/hard end from Luxor: I really liked Three Kinds of Dice--"hidden convex hull problem" is always a satisfying thing to figure out and it doesn't feel like it's been overdone at this point; it really isn't obvious from the problem statement that there will end up being a geometry component. WF23K (Alea Iacta Est) is one of my all-time favorite ICPC problems, although I'm disappointed that it could be solved using iterative methods and I'm disappointed in every team at the actual contest for solving it this way :) I solved it using the intended solution, which I think is a quite nice way to understand the problem, and figuring out how to represent the states in a useful way is a satisfying puzzle. This kind of graph search-style approach was also not something I had seen before, so the idea really felt novel and exciting.

Also, while I generally don't like implementation deciding medals at WF, I think WF22X (Quartets) was on the better end of implementation-heavy problems, because you actually had to think some about how to represent the information effectively in order to make the implementation work. (Amusingly, when I VCed WF22, this was the third problem I attempted, after the two easiest problems. I think I got midway through implementing it, but when I realized I'd have to do the matching step I decided there was no way it was actually on the easy end of the set, and I ended up coming back to it later.) While R and X were collectively a little implementation-heavy, I think WF22R/V/X worked well at collectively deciding medals, and I really liked WF23B/K as medal deciders (if not for the fact that WF23K basically ended up being a test of "do you try to implement the iterative approach" rather than of whether you actually come up with the intended idea).

Thanks for all your work on these contests!

I think one thing is to remove problems that have been seen too many times (like A and I) and try to balance the ratio between implementing and thinking by simply removing L.

here is my feedback, I will be talking of the problems ABCDFI. I did not have time to try any other problems unfortunately.

Problem B : This is a great first problem imo. Simple and short. Only issue is the statement being somewhat informal leading to it being misintrepretable. More on that later.

problem F : Its just meh. From the onset, it is pretty obvious what to do. Each step is obvious and standard from the previous. Still, i think it's fine to have one such problem. Something I would like is : removing the recovery part of the problem. Maybe it only saves 5-10mins, but I think such small things of "caring about the contestant" adds up.

Problem C : Please kill implement the statement problems....they definitely should not exist. They are boring and only time killers, and you have to hope you don't get WA and lose even more time. There were other issues with this too, such as not being clear about whether brute force passes (it did) and the big negative years issue.

Problem I : I have seen this at least 2 — 3 times before, and so have most experienced contestants. I don't understand how WF jury thought it is fine to use this despite it being so repeated already. Surely most members from the WF jury have also encountered the exact same problem before? So, the only explanation I can think of is that they think its fine because no "internet access".

From a first time solving, O(n^2) is somewhat ok, while O(n) is just tedious and boring. (Fun fact : I was trying a puzzle by huawei 2 days before the contest, and it reminded me of this very problem. I even tried to remember its source and look it up then, but i was unsuccessful)

Problem D : Personally, I made a lot of mistakes in this problem which i found after stress testing, and that cost me a lot of time. Nevertheless, I think the problem was completely fine and decent.

Problem A : I have seen the idea before, and all that was left for me to do is actually implement the lines thing, which is very tedious personally for me. unfortunately, I ran out of time in WF while debugging this (I had a division by 0 error for a long time). Here are some variations I would have liked more :

• Lower constraints to allow $$$O(n^2 \cdot log(n))$$$ solutions : This in itself would be a good improvement I think

• Remove the linear function section altogether : Either, make the function simpler (like ranges of constants, instead of ranges of linears) or change it to something like "functions are hidden, but you can query the integral in a range".

• For the "you are allowed to query integral of a range", There exists a cool solution in nlog^2 using Divide and Conquer. I want the problems to be more focused on ideas like these, rather than pointless wasting time thinking how to handle several linear functions within the time limit. (Funfact : Infact, i was told how to solve this very problem by our IOI leader while travelling to IOI 2023)

Summary

Within the 6 problems I mentioned, I think only B, D and F are fit to be used as is. A could have been made a lot better, while I and C just need to go.

I would very much enjoy if ICPC WF wasn't so implementation oriented, where the difficulty lies in implementing the idea rather than coming up with the solutions. I was implementing for maybe 3 — 4 hours in contest. Somebody rightly mentioned "ICPC gives you 15 hours of thinking time, and 5 hours of implementing time, while in reality you need the reverse".

Strong contestants are held back from trying hard problems. Weaker contestants are left frustated as they keep failing to implement correctly. I don't think it helps anyone.

Somewhat related to this is "Try to make the contestant's lives easier". Maybe you found a cool problem in $$$O(n^2)$$$, but then it is tedious (and also much harder) to solve in $$$O(n)$$$. I would request authors to not use the harder version. Anytime you want to use a harder version, I think you should always argue if it adds much to the problem compared to the amount of effort needed. For example, problem A linear functions and problem I $$$O(n)$$$ doesn't add much to the problem, while they both take considerable amount of effort.

Statements : There were multiple places where I had ambiguity in understanding statements. Personally, I prefer Atcoder style statements, but I can understand having stories. What I think is really really necessary, and why this issue even arises in the first place is not being mathematically formal in the important parts. Write your story, but please also give me an accurate definition. The best example can be gotten in geothermal's suggestion to B.