Disclaimer : this blog is not attacking any problemsetters directly, its about general trend. And my English is horrible so...
So lets first analyze contests that a hosted nowadays.
- Codeforces : 2 hour contests with cf like problems
- Atcoder : 100 mins contests with cf like problems
- Codechef : 2 hour contests with cf like problems
You get the idea. Everyone is trying to copy cf. So what`s bad with that? CF like problems become more guess the idea, test it on samples, roll your face on a keyboard, and pray it pasts pretests. Its exaturation but still, I asked my expert friends when was the last time they used pen/paper on D2A, D2B or D2C. I got a lot of answers of : "Sometimes on C, but in general I dont use pen and paper on first tasks.". Same thing with proving your solution works. I never do that on ABC, and same with people whom I asked. So the thing is, a lot of people solve ABC only at div2, and their placement will depend heavily on how fast they can solve. And if you are trying to solve ABC using normal methods of pen/paper, you will get overtaken by a lot of guessers. And some guessers who were unlucky to get some WA, will be below you. IF you want more proof of that, look at peoples rating graphs, do they look normal to you, or they look like heartbeat chart? And even worse with performance graphs, when it almost looks random, with points generated randomly from 1000 to 2000.
And there are 5 hours contests that become more like 3 atcoders in a row. Subtasks. Do you see any difference between them and cf problems? You have to get all tests right and only then you can get points. Same as cf. So in this systems solutions that work fast enough only in 10% of cases when idk, n < 10 and a[i] < 100 can get the same emount of points as solutions that works in average in n log n, but n ** 2 in worst case scenario. Is this fair? I dont think so.
Also thats why people are stuck. You cant guess solutions to all the tasks, and to me guessing part ends after ABC.
