Hello everyone!
acm.sgu.ru was created ~16 years ago by programming contests participants from Saratov university: MikeMirzayanov and AndrewLazarev. Today its codebase and infrastructure are completely outdated, and supporting it is unreasonably hard. Nonetheless, Codeforces team is regularly asked to revive acmsguru, since there is still a lot of interest in the problemset.
Here is the new functionality in Codeforces — additional section "acmsguru" in Problemset. All the problems from the old system are moved there. There are still a few things to do here: correctly port interactive problems, improve looks of statements, support the new section in Codeforces API, support adding these problems to mashups.
Anyway, you can solve the problems already, and view progress on separate standings page for acmsguru.
To save the spirit of acmsguru, viewing sources of other participants is disabled.
UPD: In some old problems checkers are compiled into 16-bit executables, so testing of these problems is temporarily not working.
Why can't I view others submissions?
To save the spirit of acmsguru, viewing sources of other participants is disabled.I wish if one problem is already accepted, then we can read the better code.
why so many "Denial of judgement"? I wanna test the code that I submit in sgu, and I got a "Denial of Judgement". Maybe it is not allowed?
In some cases it's WA 1, not Denial of judgement (37557189 CF == 1765512 SGU)
Whish to see testcases after getting wrong anwser:D
Oh shit, these tasks are hard :D So what, are we racing who will be the first to solve them all?
if they are hard for you doesnt mean their hard for every one :D
Actually that's exactly what it means... broadly speaking.
Can you please add tags to these problems as well, its a great feature to have while practicing certain topics
WOOOOW, good job
Thanks!
great update
thank u codeforces team ;)
An old friend has returned :)
Wow! That's awesome news!
Yeah i'm back
Is there any way to view/download your own solutions on acm.sgu.ru? I couldn't figure it out.
Some links seem to be broken. For example, clicking on problems' names at submissions tab doesn't work.
We will fix it, thanks.
That's amazing! Thanks a lot for this!
It was my favorite online judge when I was actively participating in ACM ICPC. Glad that now more people will be able to enjoy these problems.
By the way, is it possible to somehow get the archive of my solutions from acm.sgu.ru (given id and password)?
Could you revive TTB?
Some problems have TL = 0.25 sec, and the execution time is doubled. It's impossible to solve such problems in Java (e.g. 229 which I have accepted on acm.sgu.ru)
Problem 214 seems broken. Both xiaowuc1 and I got "Denial of judgement". I got AC on SGU and on ASC 03 in CF gym.
Thanks. We will look into it.
Could you please make it listing out the problems in increasing order of problem ID just like the old website ? Thanks.
Sorry, I read it wrong
Is it possible by now to add these problems to mashups?
a few moments later...
formatting is terrbile for few problems (Example: 113. Nearly prime numbers)
MikeMirzayanov, Can community help somehow?
Any news on adding sgu problems to mashups? :D
Is it possible to view test cases of these problems?
what is acmsguru
It's one kind of problemset.You can develop your skill very much,b coz you cannot see others solution here,and tutorial is banned for acmsguru.To develop yourself,you should practice the problems from problemset,acmsguru and contest page
easiest contest problem:[problem:[problem:4A.Watermelon]],and easiest acmsguru problem[problem:100.A+B]](A+B)
An old friend just came back.
If you want the editorial, solve problems from the main problemset.
Any news on adding sgu problems to mashups? :(
Problem 285 seems to be broken, my submission got Judgement failed. MikeMirzayanov, could you please look into this?
Fixed. Thanks for pointing out the issue.
What actual memory limits for acmsguru problems? I've got AC for problem 122 with submission which used 6200KB memory, although in statement said that memory limit is 4096 KB.
Is it not better to show testcases? Sometimes it's usefull.
Would it be possible to restore old acm.sgu.ru accounts' list of solved problems? I did some problems in the old site but I don't know which ones are left now.
I have been solving the problems in the last few months, and noticed that problem 145 might be broken. Particularly, there are no passing solutions and the issue seems to be that the checker rejects paths other than the suggested one (although the problem statement says "if there exist many solutions output any of them"). Can we please fix this MikeMirzayanov / kuviman?
Hopefully soon this will be one of the few left for me, and I don't want to break my daily streak. Can we please fix it :))
Only one problem left, can we please fix it :))
I started solving a problem from these every day maybe a bit over a year ago, and certainly would recommend trying them. Even though there were some that were purely implementation heavy, there were lots with cool ideas. I also liked the ones that required some random theory I wouldn't be otherwise aware of, so I also certainly learned a few things. Thank you for originally gathering them, and then moving them to CF!
Some things I found interesting after going through all of them was that LLMs were surprisingly bad at the harder ones in terms of figuring the idea out, but they have become absolutely great as editing tools — say using claude code / codex for implementing a function or an outline based on an idea you provide. You still have to actually know what to do, but it seems a lot of the implementation aspects have been "solved". After solving a lot of these, I have experimented a fair bit with trying to get GPT 5.5 / Claude / Grok to get to a solution without hints and it was interesting to see that unless given access to an internet search tool, they were mostly failing — the common outcome was an almost infinite "thinking" loop. And I found this surprising given the reported benchmarks and results. One theory could be that you actually need lots of compute to get there, while another is that the publicly available models aren't as good at it. It's also possible that some of these SGU problems are out of distribution because a lot of the solutions are missing from the internet (and so the training data). Either way, this is mostly a ramble on my experience from playing with LLMs on this.
While solving them, I have been keeping an archive of the solutions with something resembling an editorial after the first maybe ~150 problems, which might be helpful for people if they are stuck at some of these. At some point I made an LLM written editorial based on the solution idea I gave it + my code, and made minor edits afterwards. You can find some of these here too. I haven't pushed an update since around 2 months ago, but I will pickup the rest of the problems at some point this week.