I think you might somehow misread my comment. I agree the "simple" application of tricks is not cool.

Talk about the "Rainbow Bridge" problem. The task is fine, even nice. But it's just not of my taste. I agree that matroid is cool. When I firstly encountered matroid several years ago (like in 2010), I was thrilled. I even skimmed through a book titled "Matroid Theory" to learn more. I am doing theoretical cs research, though not in the subfield of combinatorial optimization. I can also recognize the importance of matroid. So in either way, I am not ignorant about or belittle the matroid itself.

It seems to have a tendency to prefer advanced problem. But it's of personal style, while I am talking in the general competitive programming context. I was trying to make a point that our community should emphasize the problem-solving skill, not laying the hardness of a problem in its sophisticated background.

In the meantime, I think introducing advanced stuff to our community is necessary, especially in training contests like Petrozavodsk. I personally like BM-algorithm introduced into our community in recent years. It's super nice. But if I find such problems are presented in a serious contest like World Finals, I feel frustrated. I also have to admit I have such feeling because team Dreadnought didn't solve a problem in WF 2016 without a good linear programming solver in their libraries and lost the competition.

I don't like problems about [obscure mathematical thing] because "can a contestant solve this problem?" very often reduces to "contestant knows [obscure mathematical thing] and how it's used to solve these kinds of problems". Any tricks specific for that problem will usually be trivial compared to being able to spot that it should use that mathematical thing, and if they aren't, the problem is most likely utter overkill.

You need to know some theory for everything, but it's a question of difficulty. If you don't know Dijkstra, you can figure out the algorithm by yourself in-contest, it's not so hard. If you don't know Grundy numbers, you can... print some patterns and figure out Sprague-Grundy theorem by yourself, I guess? Still, that takes a lot of luck and it turns out that a lot of problems on games are boring applications of that theory, sometimes with heuristics / pattern searching to speed up computing Grundy numbers. It's already skirting the line. If you don't know matroids, you can go outside.

Fun fact: I had Lewin propose a matroid problem to a contest I tested. I didn't understand the theory and he couldn't explain it in a way that I would understand (intuitively enough), so I couldn't honestly accept it as a contest problem.