Log in

30 January 2008 @ 12:57 am
How I finally find youtube...  
Just before the Mystery Hunt, I mentioned a filming project I was involved in related to sudoku. The School of Medicine here at Stanford put together an interview piece with me talking about my life in science and my life in sudoku which came together really well. If you want to hear me for 8 or so minutes, or just laugh at how expressive I can be with my hands while talking very fast, you can feel free to click here.

While the interview segment is the most complete and interesting (I feel) of the content you can find about me online, the most intriguing part of the project will probably be a continuing series in which you can watch me solve a puzzle and then talk through how I solved that puzzle, giving tips and such. The first in this series is an "S.U." Doku I wrote specifically for a Stanford magazine of moderate difficulty. You can try the puzzle by going here first.

My solution on camera was in 2'11". I'm not completely clean in my solve here (consider how I fill in row 6 ..., or don't despite my seeing the obvious placement when I resolved in the demo) but its representative of my style.

If you have doubts that my solving a puzzle I'd written a week or so before is not fair, the rest of the series are puzzles written specifically by Wei-Hwa Huang that I had not seen before the documentary crew presented them to me. There will be a near speed record on one of them, and then a gradual increase in difficulty so you can track how you'd do compared to the US champion. It will be a fun eight week series leading up to the world championship and the release of Battleship Sudoku.

I also go back through the puzzle and give a description of how I went about it. As the first in the series, we covered more how to get started and specifically described my "either/or" notetaking which is all I use to this day, and so sometime midway through I go way too fast and the tips stop. There will be more videos and more tips to come.

I look forward to any feedback you have on this project, and feel free to link to the youtube videos as we are trying to spread word of this really nice video piece. It certainly is a much more fitting youtube sudoku speed solve than the phony or slow videos you can already find there.
XSG - 1 G 2 Manyxsg on January 30th, 2008 03:15 pm (UTC)
Well, now I know how you mark up your sudoku, and if you're not too set in your ways I think I can provide you with a method to speed you up a few seconds per puzzle...

When you place a mark in a cell (not on a border), you write the entire number. You don't need to... instead, you can break the cell up into nine respective cells (3x3) and mentally number them any way you see fit (but, of course, always number them the same). Then, instead of writing the number in the cell, just place a dot in the number's location. I usually break my cells up like this:

| 1 | 2 | 3 |
| 4 | 5 | 6 |
| 7 | 8 | 9 |

so if, for example, a cell can contain either a 2 or a 6, it'll be marked up like this:
|     .     |
|           |
|         . |
|           |
|           |

At your level, this suggestion might be too much of a switch and might cost you a bit of time initially, so use at your own risk. Still, I'd be interested in hearing if/how much time this help you out in the long run, though technically a long-run study might be influenced by other things...
motrismotris on January 30th, 2008 05:13 pm (UTC)
I certainly have read about the dotting system in many places and it is a reasonable suggestion. It should still be fine for identifying locked pairs and getting basic eliminations and for easy to medium puzzles would save some seconds.

However, one of the uses I find for numbers is that once I have a fair number of notes in a cell, I can go back to searching for naked singles by looking at the set of numbers I do not have written anywhere in that box/row/column. If I used the dots, I'm worried that I would lose the instant recognition of missing characters as I'm good at seeing 1-9, but not yet well trained to see a combination of both numbers and dots to do that. So, on hard to very hard puzzles, I think I'd lose more time than I'd save on simpler ones.
XSG - 1 G 2 Manyxsg on January 30th, 2008 06:06 pm (UTC)
The dots help me even for the more challenging puzzles because I find I use a different part of my brain for identifying dot patterns as opposed to numerical patterns... and my dot-pattern recognition is faster than my number-pattern recognition. I seem to be able to do the dot-conversion very quickly as well, as there's no discernable lag for finding missing characters. I understand it would take you time to re-learn a new way and it's probably not worth it for you, and since I'm nowhere near your level of sudoku performance, my sudoku improvement upon switching to dots is probably not a fair comparison for you.

I have a proposal, however...

The same method works very well in kakro. Try practicing difficult kakro using dots in order to train your brain to do the dot-number conversion quickly; this oughtn't affect your sudoku performance until you're ready to completely switch over...
zundevilzundevil on January 31st, 2008 01:58 am (UTC)
Here come dots
I've seen this applied in practice and am hesitant to commit to it. For one thing -- I sometimes will use the pencil as less of a writing implement and more of a pointer. xsq, have you ever incidentally made a mark that you didn't mean to, and thereby thrown things off?

I also don't know how quickly I would go from numbers -> dots -> numbers in a given puzzle. Say what you will about possible inefficiencies from scrawling tiny digits in the upper corners of a cell, but at least you've got numbers on the brain.

How about numbers written on lines between cells? I guess it's all in the mind, but a 3 written on the wall between two cells has a much more obvious interpretation than two "3-dots" in the adjacent cells.

Having thought more about it here -- this is intriguing, and I may end up giving it a look one of these days. Saving these milliseconds probably shouldn't be the #1 priority for my solving tho...

Does anyone reading this recall the notation scheme of one of the finalists in Philadelphia that caused Will to remark that he had never seen any notations like those before? Perhaps the individual behind the notating reads this very blog? :)
motrismotris on January 31st, 2008 02:14 am (UTC)
Re: Here come dots
Tammy's notation was not terribly different than mine in seemingly recording doubles as primary information.

However, she would write one number in a box and draw a line connecting it to the other box the pair was present in which was the novel part as it would connect boxes in row 1/3 more effectively perhaps. I don't recall anything else standing out but then I didn't have the chance to specifically watch her solving during the finals.
Adam R. Wood: butasanzotmeister on January 31st, 2008 08:15 pm (UTC)
Re: Here come dots
I used to use positional dots myself, before I tried positional digits, but I actually switched to numerals for much the same reason Tom cites: I'm mentally dealing with only one type of symbol instead of two, with no translation necessary.

Actually, I try to avoid notation entirely, especially given that I'm typically anal enough to have to erase them when I find the proper value of the cell. I only resort to markup if I get quite stuck. It also helps keep me from overcomplicating matters, forcing me to not delve too deeply in the wrong part of the puzzle and instead try to find the easiest move no matter where it is. As an aside, I would note that this is pretty much restricted to Sudoku; I have experimentally determined time and again that at, for example, Slither Link, I'm MUCH better off marking all the 'x's rather than drawing in just the loop, no matter how much time I spend crossing off those spaces between the dots. - ZM
Adam R. Wood: butasanzotmeister on January 31st, 2008 08:18 pm (UTC)
EIGHT WEEKS!? GAAAAH! Want... book... now.... - ZM
jdyer on February 1st, 2008 12:28 am (UTC)
ivoryvampire on January 31st, 2008 10:51 pm (UTC)
It looks like I use the same method as you for solving sudokus. The only difference is that you are much faster. Of course, not falling asleep while I'm trying to solve them on the everyday train back home might speed up a bit my technic.... XP

Thanks for sharing it! :)
decand on February 5th, 2008 01:32 am (UTC)
Well, at least I learned something. I've been mispronouncing Nikoli for a long time. That'll be a hard habit to break.

Looking forward to the rest of the series, though. Who knows what else I've been doing wrong?