Problem 44248. Mastermind V: Optimal Solver - average number of guesses

In the test script, what does "randperm(end)" do? Is that even proper MATLAB code?

@yurenchu.
Yes. The END here serves as the last index of P.

Either 7296 iterations is too much, or the test script causes an infinite loop somehow. I've tried several approaches, ones that worked smoothly in Richard Zapor's Problems, but I always get "the server encountered an error caused by long running MATLAB code". Are you sure "P(randperm(end))" does what it's supposed to do? Why not use the less ambiguous "P(randperm(numel(P)))" instead?

@yurenchu.
Cody has an approx 50 seconds time limit.
You could verify the following code:
>> a = 4:6; a(randperm(end))

I don't have MATLAB installed on a computer, so I verified your example using one of the beginner problems ("Solution 1225050"); and I'm surprised but you're right, this peculiar piece of code indeed works. So it seems the trick is in writing a function that can play the game 7300 times (= approx. 33,000 guesses) in less than 50 seconds.

@yurenchu.The challenge of the problem is to optimize average number of guesses.
For speed challenge, 7300 is not enough. I've test some solutions without speed optimization passed these tests in 10 secs.

@LY_CAO: Thanks, but I still don't understand. Your own Solution 1215381 from Zapor's "Mastermind II: Solve in 8 or less" managed to solve 1296 games with a total of 6508 guesses in only 3.02 seconds. Yet when I copy and submit the very same solution here (where I expect at most six times as many needed guesses, so a total of <20 seconds), I receive the "error due to long running MATLAB code" abort alert (see Solution 1225064). How is that possible? The only possible factor (as far as I can see) is the difference in test script.

@yurenchu.
It seems that the typo "ismemb2" error was suspressed by TRY, so the output was always [1 1 2 2].

@LY_CAO: Thanks for notifying me of my typo, which I hadn't noticed! I've finally managed to submit a working solution; but it still seems to be a challenge to keep the solution under the time limit without using ismembc2 or while applying a "minimal guess"-algorithm.

@yurenchu.
ISMEMBC2 generates the second output of ISMEMBER, so I guess it should work flawless even if 'q = ismembc2(...)' is replaced by '[~,q] = ismember(...)'.

Thanks, LY_CAO. What I meant was that before considering "ismembc2()"/"ismember()" with matrix indexing, I instead used relational operation combined with matrix multiplication, which calculates the same thing but which is apparently too time-consuming.