I know, but I was trying to be ambitious.



Think it's a good idea to cater to all people who compete. It brings a different challenge for the top solvers, somewhat comparable with what the short rounds in a WPC bring.

I agree with Bram that it is good to test solvers at different things (sprints versus large puzzles, "classics" versus variants) as well as occasionally having tests with easy/easier puzzles. My one worry is having 40+ minutes on the clock when I hit submit and having one tiny error to lose 80+ points. I've been pretty clean recently, so let's hope I can race to the finish and do that 36 minute target justice.

Sounds cool. My plan is to sprint through the first 63 minutes to see how far I get, and only then enter the answers. Or, perhaps, finish before 63 minutes? Or, perhaps, not write the test again. :(

why is there more of puzzle tests and less of sudoku tests?

There is 1 puzzle test and 1 sudoku test every month. July had 5 weekends, so there was an additional test. If the additional test was a sudoku test, someone would've asked "why is there more of sudoku tests and less of puzzle tests?" :-)

Between May'2010 and Jul'2011, we have had 17 monthly puzzle tests and 16 monthly sudokus tests, so I think that is a fair ratio.

We would like to continue in this mode (1 sudoku and 1 puzzle test per month) for as long as we can, but if ever we have to reduce the frequency, we might reduce the frequency of sudoku tests. Firstly because we get more puzzle sets from authors than sudokus, and also to quote a former World Sudoku Champion, sudoku captures just 10% of all interesting logic puzzle space

The bonus system will most probably be changed. Deb and I are not happy with the current one, but we didn't have much time to think about it before. We definitely want to avoid the case of a player solving everything very quickly and losing a huge amount of points because of a small mistake.

If the player has a lot of time left, the player "can" spend time (as much time as the player wants) in checking the answers before hitting on "claim bonus". If someone still submits early with a mistake (with a "lot" of time left), he/she should be penalized to some extent.

Of course he should. The question is "to what extent ?". I agree that players can check their answers before hitting the "claim bonus" button. But on the other hand, we ask players to go as fast as possible; expecting them to spend 10' checking carefully every puzzle and re-reading all their answers is somewhat contradictory.
Mistakes will be penalized; but the idea is to avoid players restraining themselves to go fast - I do not see puzzle solving as russian roulette.

Anyway, no decision has yet been taken.

What about this. A player finishes, he's (or she's) happy with the solution and clicks on "CLAIM BONUS". then he/she notices there is a mistake. There should be a second button like "SUBMIT NEW/DIFFERENT ANSWERS or AMEND ANSWERS", the player clicks on that, can amend the answers, click on "CLAIM BONUS" again. In this case the player will be awarded a bonus reduced by 50% (calculated on the basis of the time he clicks "CLAIM BONUS" the second time). facility to be used once only, if there are other mistakes he/she has'nt found the first time, they will stay.

I think too many buttons will not be good. It is understood that once you click on "claim bonus" the test is over for you.

One possible solution that I don't know if you are thinking about is to not necessarily deduct the value of time bonus, but to deduct some fixed number of minutes from the total count of time bonus (similar to a WPC playoff where a mistake is a one minute penalty, but immediate grading isn't used on this test). The penalty would still only be used in the case of just one wrong, and the solver would still lose the puzzle points, so it would be a 20-30 point mistake most likely. One thing I prefer about penalizing time and not point rate is that it applies equally to all solvers, and seems to let you fairly run a test for long enough that many many solvers get a chance to finish.

Edited by motris 2011-08-29 9:24 PM

Here's an idea that combines what you just said above with some of your earlier ideas about instant grading (which I said back then I was skeptical of). When you hit Claim Bonus, if there are any wrong answers,
- you lose some amount of time: maybe a fixed 5 minutes, maybe 3 minutes per wrong puzzle, maybe something else.
- you are told which puzzles are wrong.
- the timer continues, if it didn't get knocked down to 0 by the penalty, and you may correct your puzzles and hit Claim Bonus later with no other penalties beyond the subtracted time.

Awhile ago it was brought up the author/organizer has to set a very thorough list of alternate acceptable answers for this instant grading to work. The above system has the same concern of course.

I don't want to take-over this discussion thread with implementation discussion that doesn't feel appropriate to this test which won't have any interface changes. But the one thing in your suggestion that I do not like is to dynamically change the solver's clock during the test. It feels safer to log potential penalty data, and apply scoring changes later (where some penalties may be removed if a solver's submission is indeed considered right for whatever reason after inspection), but make sure every solver gets the same total time for initial submission.

A rather different implementation, similar to the end of the screen test, would be to have a 2 minute "review" process after the test finishes for any solver, either when their clock runs out, or they hit claim bonus. In that time "mistakes" can be corrected, but only for puzzles that already had submissions, and the test grader will certainly see the strings that were amended to determine if a solver was fairly using the time just to check and not to do more solving. I'm still trying to decide on the right match for my future LMI test, but I think we'll really address these questions when that comes up. For Bastien's test I'd be happy with something rather simple but fair, as I do look forward to a test that perhaps 30 or more people might finish.

Thanks for all the suggestions. Indeed, for this test we wanted to avoid bonus systems which would have needed to change the interface; so we have decided to stay on something quite simple but, I hope, more fair than the previous system. It will be made of two parts :

1) a player who finishes early, either with 14 or 15 correct puzzles, will get 3 bonus points per minute.
2) additionally, if all the 15 puzzles are correct and whatever the remaining time is, the player will get a fixed 25 points bonus.

This way, nobody should loose more than 25 points (the puzzle points excepted) on a single typo.

The points distribution is close from what we could get with Thomas' suggestion of deducing a fixed number of minutes ; the main difference being that this way, there remains a significative difference between two players finishing with a few minutes saved, one with 15 puzzles and the other one with 14.

An updated version of the instruction booklet should probably be available in some time.

An updated version of the instruction booklet IS available.

A small note about Akari. It is only necessary to enter the column position of left most light bulb for each row.

For the following grid, answer key is BABDACB. Same rule for Star Battle.

The submission page was showing target time of 60 minutes for all players with rating < 500. This was a typo and has been fixed. Thanks Fred76 for reporting.

Note that target time is just for display purpose. For everyone, submissions will be allowed until 75 minutes, irrespective of the target time.

Puzzle Booklet uploaded. It has 8 pages. Each page has 2 puzzles, except the last page. There is no header page.

Flash submission enabled too.

So, now my target time is 75 minutes, but what about my target points

In your case, your target points are exactly equal to your number of posts on LMI forum.

Now if you are wise, you will not answer this.

you know me: you know that I am wise

Score page link - http://logicmastersindia.com/M201109P/score.asp

As usual, please use the score page to settle all claims

For Fortress, the Answer Key instructions advise: "Write the length of separate white cell blocks in the marked rows and/or columns (write 0 if there is none). The answer for the example would be 1, 1."

Frankly, its a bad noninstructive example. If there were 2 separate white cells of length 1 in the first row, should the answer have been 11, 1 (a one for each separate white cell block) or 2, 1 (providing the aggregate "length of the separate white cell blocks")

Thanks.

For Galaxies, the Answer Key instructions advise: "Write the length of distinct regions parts in the marked rows and/or columns."

Just as in Heyawacky, symmetrical "galaxies" can double back on themselves and have arms or legs separated by other galaxies. In such instances should the answer reflect each portion of a multi-part single galaxy by the length of its aggregate distinct parts, or each portion separately?

Thanks.

Edited by thesubro 2011-09-03 8:02 AM