I mean.. we are comparing students abilities here, and doing stuff fast is one of those abilities. Even potato peelers in a restaurant are valued more if they're faster, why not programmers too? Or DMV workers?
It is actually very informative when one person can
Like anything i had to do in a test when i was taking my CS degree is maybe 5% if not less of the portion of my real job tasks. Even if i was triple as fast at taking those tests, i think that would be a neglibile increase in on the job speed.
I've never seen that come down to processing speed. Even as a programmer -- I can program probably 10x faster than most of my peers in straight programming contest style programs. But in terms of actual real work -- I'm probably slightly faster. But my value is really I spend a lot of time really understanding the ask and impact of the work I'm doing -- asking good questions, articulating what I'm delivering, etc...
That is, my faster processing speed results in very little added benefit. That is, time to deliver results can matter. Processing speed typically is a very small percentage of that time. And for these tests processing speed is often the main distinction. It's not like they're distinguishing one kid who can't solve this equation and another kid who can. It's generally more likely one kid can finish all 25 questions in 32 minutes and the other would take 38 minutes so they only finish 23 of them in the allotted 32. I don't think that ends up mattering in any real way.
For example, get 90% on a test, that's applauded and earns a distinction. In a job context, 90% gets you fired. I don't want a worker who produces "90% well soldered boards". I don't want software that runs on "90% of our customers computers". Or a bug in every 10 lines of released code.
A test puts an arbitrary time limit on a task. In the real world time is seldom the goal. Correctness is more important. (Well, the mechanic was going to put all the wheel nuts on, but he ran out of time.)
College tests are largely a test of memory, not knowledge or understanding. "List the 7 layers of OSI in order." In the real world you can just Google it. Testing understanding is much harder to mark though, Testing memory is easy to set, easy to mark.
Some courses are moving away from timed tests, and more towards assignments through the year. That's a better measure (but alas also easier to cheat. )
I'm also surprised at how common it is for people to openly discuss how irrelevant leetcode is to the actual work on the job but how it is still the status quo. On one hand we like to claim that an academic education is not beneficial but in the other hand use it as the main testing method.
I think why I'm most surprised is we, more than most other jobs, have a publicly visible "proof of competence." Most of us have git repos that are publicly available! I can totally understand that this isn't universal, but in very few industries is there such a publicly visible record of work. Who else has that? Artists? I'm not sure why this isn't more heavily weighted than these weird code tests that we've developed a secondary market to help people optimize for. It feels like a huge waste of money and time.
/s