Rating Assassins

Our weekly <a href="http://www.sudocue.net/weeklykiller.php">Killer Sudokus</a> should not be taken too lightly. Don't turn your back on them.
sudokuEd
Grandmaster
Grandmaster
Posts: 257
Joined: Mon Jun 19, 2006 11:06 am
Location: Sydney Australia

Post by sudokuEd »

Found a bug in SS which could affect many puzzles & their scores. Back later - hopefully not 5 months this time!!

Cheers
Ed
frank
Regular
Regular
Posts: 10
Joined: Sat Oct 07, 2006 2:28 pm
Location: Victoria, B.C., Canada

Post by frank »

Richard's SudokuSolver v3.0.0 rates this as 0.48 and also 0.60 depending on
how I open the file. However if a puzzle ever deserved a rating of 0 it is
this one. I have been very interested in SudokuSolver ratings and have
used it to rate many puzzles from different sites with fascinating results.
I applaud the effort to get a rating scheme is - it is well worthwhile.

But - let me know what happened to this one.


Image


Text: 3x3::k:1024:769:1794:259:516:1285:1542:2311:2056:521:2314:267:2060:1805:1550:1295:1040:785:1298:1555:2068:789:1046:2327:1816:281:538:2331:284:1309:1054:1567:800:545:2082:1827:1828:1061:1574:551:1320:2089:2346:811:300:2093:558:815:1840:2353:306:1075:1332:1589:822:1847:1080:2361:2106:571:316:1597:1342:1599:2112:577:1346:323:1092:837:1862:2375:328:1353:2378:1611:844:1869:2126:591:1104:
Andrew
Grandmaster
Grandmaster
Posts: 300
Joined: Fri Aug 11, 2006 4:48 am
Location: Lethbridge, Alberta

Post by Andrew »

Nice April Fool's Day post Frank! :wink:

Taking the point of rating this diagram, there are 81 Naked Singles so the score must be slightly greater than zero.
sudokuEd
Grandmaster
Grandmaster
Posts: 257
Joined: Mon Jun 19, 2006 11:06 am
Location: Sydney Australia

Post by sudokuEd »

frank wrote:Richard's SudokuSolver v3.0.0 rates this as 0.48 and also 0.60 depending on how I open the file. However if a puzzle ever deserved a rating of 0 it isthis one
Ha, nice one Frank. Richard loves bug-finders! He'll enjoy that one.

0.48 is the equivalent of a zero rating. We've always called this an Assassin killer scoring project - not killers in general. As you can tell, SS is not so good at the bottom end and tends to squash them in pretty tight. Ruud's software method that he was working on with SumoCue (unreleased) was much more aimed at the lower end and spread those out much more.
frank wrote:I have been very interested in SudokuSolver ratings and have used it to rate many puzzles from different sites with fascinating results
An update here on any generalizations/observations/results would be very interesting, especially as I'm a regular lurker elsewhere (sadly not Frank's for a long while :oops: ).
frank wrote:applaud the effort to get a rating scheme is - it is well worthwhile
Thanks! Richard has a long list of wish-list changes to SS(v3.0) (hidden killer triple, ALS cage block, killer quad etc) including many things for other formats, so the next V of scoring won't be for a good while.

Just to restate my last update - SSscore(v3) is only reasonably accurate up to about 1.40. It will usually err on the lower side. This is because "easier" puzzles that have a long solution set the maximum at about 1.40.

Thanks for your interest.

Cheers
Ed
Post Reply