This is about hitting the high notes, as Joel Spolsky put it in his [post].
A lot of people are good software engineers and can wrote good software. To get Jeff Deans that write revolutionary software, we don't yet have a better method than training people in these abstract, very hard problems, making them compete and hoping for the best.
As much as we software engineers / computer scientists believe we are special snowflakes, this exact same method is applied to most STEM PhD to professor tracks, mathematics, classic music instrument players, [athletes], and what have you.
And, of course, being good at coding competitions doesn't mean that you're necessarily a good software engineer. The Jeff Deans are rare. "Hoping for the best" is pretty accurate.
I suspect there is a real relationship. General intelligence is a real phenomenon. Being smart at one task indicates, though does not guarantee, that you'll be smart at another task.
This is of course a different question than whether this test provides maximum predictiveness of future success. But it seems obvious to me that the test provides some information. It's really hard for a variable to be completely non-informative.
Lastly, I think it's important for us to not be too defensive. I sense, perhaps incorrectly, that sometimes we don't want to give value to tests we are bad at, because doing so would imply we aren't as good as feel we are.
I've interviewed and rejected many TopCoders and IOI competitors. Not sure why but it seems like a totally different skillset than software engineering.
I wonder too. I think it's like asking whether a math olympiad champion would make a great engineer. Certainly they have a lot of potential because they show extraordinary abilities, but there is much more to being a good coder than just being able to solve very short and extremely specific problems.
I've been playing these games for a couple of years (initially, for interviews preparation), and it's been very instructive. The competition format makes it convenient and addictive. There's no doubt that I improved a lot on algorithms, data structures and on "small scale" coding. While this is useful, it is quite different than what you do when working on real projects.
Math Olympiad Champion::Great Engineer as Excellent Sentence Diagrammer::Great Writer.
Puzzles are fun, but it's best not to put too much predictive faith in them, because otherwise you're just affirming the consequent. Exceptional individuals are quite likely to be good at puzzles, but you cannot (in my experience) assume that because someone is good at puzzles, they have actual skill in solving real-world problems. I've worked with too many counterexamples.
When I was preparing for interviews I found that the skill that overlapped between interview puzzles and everyday work was the "head interpreter" as I call it. Typically you end up relying on the IDE and the compiler telling you if what you're writing makes sense or not. Static analysis by the IDE would only tell you if you've made a syntax error, you would need to run a test to tell if what you've written makes logical sense too.
So there's a couple of advantages if you can execute code in your head line by line while storing state as well. You can be slightly faster and more correct while coding yourself and while reviewing the code that others have written. Typically in the latter case you don't have the benefit of the compiler or running tests.
In my anecdotal experience I found an improvement in my "head interpreter" after practising interview questions for a while because typically you're not allowed to execute your code. I can't say if any of the algorithms I learnt and practised helped, but this, in my opinion, made me slightly better at my day job.
Thank you! After reading blog-post I had zero understanding of what the problem is about and how they solve it. Tried to read again - no luck. Then I opened codejam link and saw that it's basically about placing maximum amount of bishops using maximum matching. (Other parts of problem are trivial)
That's really sad that this pattern of bad explanations is so frequent in the internet.
This wasn't my intention. I hope the feeling subsides!
I did try to keep it short although some parts still (fortunately) shows how to come to those conclusion. That might give it the impression of being smart but the presentation is definitely selected among many idea and rearranged.
'You may use any programming language to solve a Code Jam problem. However, for any round after the qualification round, the compiler, interpreter, libraries, and external services you use must be available such that anyone else can use it for free without a time limitation and without violating any rights of any person or entity'
After you reduce it to maximum bipartite matching you just use whatever stock algorithm for bipartite matching you want.
Another comment links to the official writeup. The official writeup says that you can think about it as the "rook subproblem" and the "bishop subproblem" separately. This is the easiest way to think about it imo.
I was also confused about algorithmic skill and software engineering skills. But I got clarification after working in industry for a while.
1) Most of the software engineering work is product development, just wrapping already existing algorithm and optimized computer program into apis and test and maintain it.
2) There are very few software engineering work involving writing algorithm and optimized computer program.
One example, everyone here might have just used mysql database. But if you look into the code, you might see there are complex data structure, and algorithm are being used.
32 comments
[ 3.2 ms ] story [ 82.6 ms ] threadI don't even bother replying to their recruiters any more. The interview process tells me more than enough about the company.
A lot of people are good software engineers and can wrote good software. To get Jeff Deans that write revolutionary software, we don't yet have a better method than training people in these abstract, very hard problems, making them compete and hoping for the best.
As much as we software engineers / computer scientists believe we are special snowflakes, this exact same method is applied to most STEM PhD to professor tracks, mathematics, classic music instrument players, [athletes], and what have you.
[post] https://www.joelonsoftware.com/2005/07/25/hitting-the-high-n...
[athletes] http://rittersp.com/wp-content/uploads/2014/03/Chambliss-Mun...
[musicians] http://www.nytimes.com/images/blogs/freakonomics/pdf/Deliber...
This is of course a different question than whether this test provides maximum predictiveness of future success. But it seems obvious to me that the test provides some information. It's really hard for a variable to be completely non-informative.
Lastly, I think it's important for us to not be too defensive. I sense, perhaps incorrectly, that sometimes we don't want to give value to tests we are bad at, because doing so would imply we aren't as good as feel we are.
Or to put it more pessimistically: even worse than we feel we are. Impostor syndrome and all that.
The part where you make large software, working in teams, to serve real users, can only be learnt on a job.
These competitions are not looking for anything fancy. They recruit juniors, they'll be trained on the job.
I've been playing these games for a couple of years (initially, for interviews preparation), and it's been very instructive. The competition format makes it convenient and addictive. There's no doubt that I improved a lot on algorithms, data structures and on "small scale" coding. While this is useful, it is quite different than what you do when working on real projects.
Puzzles are fun, but it's best not to put too much predictive faith in them, because otherwise you're just affirming the consequent. Exceptional individuals are quite likely to be good at puzzles, but you cannot (in my experience) assume that because someone is good at puzzles, they have actual skill in solving real-world problems. I've worked with too many counterexamples.
So there's a couple of advantages if you can execute code in your head line by line while storing state as well. You can be slightly faster and more correct while coding yourself and while reviewing the code that others have written. Typically in the latter case you don't have the benefit of the compiler or running tests.
In my anecdotal experience I found an improvement in my "head interpreter" after practising interview questions for a while because typically you're not allowed to execute your code. I can't say if any of the algorithms I learnt and practised helped, but this, in my opinion, made me slightly better at my day job.
That's really sad that this pattern of bad explanations is so frequent in the internet.
I meant that bipartite matching was enough for getting the answer, not that the two are equivalent.
I did try to keep it short although some parts still (fortunately) shows how to come to those conclusion. That might give it the impression of being smart but the presentation is definitely selected among many idea and rearranged.
'You may use any programming language to solve a Code Jam problem. However, for any round after the qualification round, the compiler, interpreter, libraries, and external services you use must be available such that anyone else can use it for free without a time limitation and without violating any rights of any person or entity'
Another comment links to the official writeup. The official writeup says that you can think about it as the "rook subproblem" and the "bishop subproblem" separately. This is the easiest way to think about it imo.
P.S., thanks for answering some of the questions in here! Guess I should reformulate that last part.
1) Most of the software engineering work is product development, just wrapping already existing algorithm and optimized computer program into apis and test and maintain it.
2) There are very few software engineering work involving writing algorithm and optimized computer program.
One example, everyone here might have just used mysql database. But if you look into the code, you might see there are complex data structure, and algorithm are being used.
https://github.com/mysql/mysql-server
Its a good thing to solve these kind of algorithmic problems.