2 posts were merged into an existing topic: Shadows of the Knight - Episode 1 - Puzzle discussion

# Shadows of the Knight - Episode 2 - Puzzle discussion

**Thank you so much!!!** I finally could be able to solve this puzzle by reading your comment! First time trying, I could only got all the test but the last one (which is really frustrating anyway ) . Then after reading your statement, I could finally understanding what to do.

**Guys (or girls),** if you guys keep trying to mirror the coordinate through the bisector, it would make you jump back-and-forth around the edge if the the bomb is near the edge. And it would be reallyyyyyyyyyyyy time-consuming, considering that you have to visit the same old window multiple times. (And for me, the result is 5 rounds away from the actual number of rounds needed to solve the final case.)

**So,** my advice for you is that focus on the statement that Geropy has stated:

For anyone who could not wrap their head around this idea, I would give you a little hint:

*Try to approach the edge fast, really, (but not too fast ).*

By doing this, I got 3 spare rounds in the last case. And I really hope you could come up with something like this, or even better!!! ^^

And as always,

Happy Coding!!!

coz the bomb is at 999,999

5 ā¤ W constraint is not true in 3rd test, the tower one, since W = 1

Fixed, thanks.

I also did the binary search separately for both coordinates with one optimization:

Because you cannot jump off of the grid, sometimes your next jump will not give you any information at all, just adjusts the position of the bat for future jumps. It makes sense to make these zero information jumps for both coordinates with a single two dimensional jump. Which means you have to work on both coordinates āin parallelā.

Hi, Iām stuck on this problem for a while.

Iām searching for the X and Y coordinates separately and everythingās fine since I pass every test.

But here is my problem : when I submit my code, I get 87% cause I pass every test but the FIRST one.

Of course, there is no way to know why I donāt pass this test because of submission mode.

Did any of you have the same problem or any idea of what could be the reason (I know the data in sub mode are slightly different)?

Maybe this was the most messy task for me so far. No matter what Iāve tried it didnāt work, I had really tough time understanding why. At the end I got the logic with many many little tweaks and trials and errors and still everything is not clear to me why my solution works

I donāt understand, how could people solve this by searching for the X and Y coordinates separately. With two separate binary searches, IDE test cases 07 and 08 use too much jumps and fail for me, so Iām currently stuck at 75% and donāt know how to progress without hardcoding some jump-orders for specific testcases.

All I could say is searching separately for X and Y allows to reach 100% both during training and validation phase.

The sole tip I could give you is: if you know your next move wonāt learn you anything, then change your search axis.

yo i searching it seperately but still donāt have enough run (n)

do u have any advice for me pls?

thanks for reading man

For those who are doing binary search separately for each axis, some tips:

- Try to go the next move that mirrors the current move with respect to the lower and upper bounds for the axis. (this cuts the search area by half)
- If the next such move is out of bounds, then get back to within the lower and upper bounds (I went into the centre for my code)
- When mirroring, there is a
*tip*to reduce the number of runs, by mirroring with an odd distance. Hereās why:

a) When the result is āwarmerā or ācolderā, the centre row/column can be eliminated instantly

b) When the result is āsameā, the centre row/column is the row/column with the bomb.

Hence, an odd distance mirroring can significantly cut down the number of runs, with a potential for a jackpot āsameā result, revealing the row/column instantly.