FMC Solution Checking

[FatBoyXPC]

Implement a service similar to @lgarron alg.cubing.net that we can use to check FMC solutions. This is to remove all human error when checking FMC results.

[jfly]

I definitely like this. @lgarron, @justinj, @kitclement,what are your thoughts on having the wca website require that fmc solutions be uploaded along with scrambles?

[jbmertens]

I wouldn't try to claim this will be less error-prone than just performing the solution by hand, as I'm perhaps more likely to make a typo than mis-scramble/solve.

I've wondered about storing (and processing?) images of the sheets though. I wonder how easy this would be nowdays.

[lgarron]

Welcome to 2008. :-P https://www.speedsolving.com/forum/showthread.php?8090-FCF-and-FMCTyper

The TNoodle FMC form was actually originally designed for easy digital transcription. ;-)

Some thoughts:

• We should definitely still make sure competitors submit solutions on paper.
• Now that we have smartphones, inputting pictures is much faster and easier.
• OCR might not actually be that difficult. Typing out moves should not be bad, though (I wish I had a screenshot of FMCTyper, but again, the current form was designed for this).
• If you make a handful of typos, the computer can detect that and tell you to check the particular move you typed wrong. ;-)
• I think FMC solutions should be public, but this is not explicitly stated in the Regulations. I have a stack of old Nationals FMC sheets I've been meaning to type up, publish, and let people analyze. :-P

[jbmertens]

Ah, that rings a faint bell! Snapping photos is probably feasible, and may even be efficient for larger competitions. I'd worry about the associated overhead for smaller competitions though (no way it is worth it without a seamless point->shoot->type ui).

The error detection bit is cool too.

[FatBoyXPC]

I feel like eating a bit of time to type all the solutions up would be an overall benefit. That said, a point and shoot UI with some fancy OCR would obviously be ideal.

I decided to stackmat typing images of the solutions on Lucas' website: 1: 21.61 2: 18.44 3: 15.69 (error, typed F2 instead of R2. Could have been due to misreading, though) 4: 16.47 5-1: 1:12.34 6: 29.97 7: 22.21

Scrambling then solving: 1) 21.78 2) 25.11 3) 18.94 4) 25.xy (first) / 59.68 (solved, scrambled, did it again) - I normally do it twice myself before asking a 2nd person to verify. Note: this solution did not work for me! 5-1) 2:18.31 - I didn't do this one twice like normal. Those cube rotations were brutal! 6) 1:04.76 - Accidentally stopped the timer when first run through did not solve the cube. 35/29 were the two times, and that wasn't counting the time to solve the cube in between. 7) 1:19.81 - This took 2 times as I messed up the first time as well (33/46 were the breakdowns of first solve then solve + scramble + check solution)

While my typing might be faster than some, I'd say it's pretty clear that my typing is faster than scramble checking. If we have that typo checker built in, then that makes things even better.

Side note: If we were to type these in, we should make it so that RFLBDUMES automatically get capitalized, and wxyz do not. That way we don't have to hit shift/caps lock.

[KitClement]

MES isn't even legal for FMC anyways, so no need to worry about that. We do need to worry about bracketed rotations, like [F], but we could just type z in this case instead.

I agree -- I think it would be a really nice feature to be able to look up solutions, and it seems that with the proper infrastructure, it would take just as much time to check solutions this way.

On Tue, Jun 9, 2015 at 2:59 AM James LaChance notifications@github.com wrote:

I feel like eating a bit of time to type all the solutions up would be an overall benefit. That said, a point and shoot UI with some fancy OCR would obviously be ideal.

I decided to stackmat typing images of the solutions on Lucas' website: 1: 21.61 2: 18.44 3: 15.69 (error, typed F2 instead of R2. Could have been due to misreading, though) 4: 16.47 5-1: 1:12.34 6: 29.97 7: 22.21

Scrambling then solving: 1) 21.78 2) 25.11 3) 18.94 4) 25.xy (first) / 59.68 (solved, scrambled, did it again) - I normally do it twice myself before asking a 2nd person to verify. Note: this solution did not work for me! 5-1) 2:18.31 - I didn't do this one twice like normal. Those cube rotations were brutal! 6) 1:04.76 - Accidentally stopped the timer when first run through did not solve the cube. 35/29 were the two times, and that wasn't counting the time to solve the cube in between. 7) 1:19.81 - This took 2 times as I messed up the first time as well (33/46 were the breakdowns of first solve then solve + scramble + check solution)

While my typing might be faster than some, I'd say it's pretty clear that my typing is faster than scramble checking. If we have that typo checker built in, then that makes things even better.

Side note: If we were to type these in, we should make it so that RFLBDUMES automatically get capitalized, and wxyz do not. That way we don't have to hit shift/caps lock.

— Reply to this email directly or view it on GitHub https://github.com/cubing/ccm/issues/249#issuecomment-110253482.

[larspetrus]

This could be useful, but I definitely think it should start as an option, and not be required until it's proven to be reliable and beloved.

A form where you check boxes rather than write letters could make the OCR easy and the solution unambiguous.

How to enter this on a keyboard in a fast and non error prone way is also an interesting topic.

[lgarron]

A form where you check boxes rather than write letters could make the OCR easy and the solution unambiguous.

You mean a form where competitors check boxes? Turns out to be a massive pain for competitors:

Conclusion: PLEASE don't make us use this

[larspetrus]

Great minds, etc :)

Do you remember the specific problems that made it so awful for Arnaud?

It would be a much cleaner form with only LRUDBF...