Log in

25 September 2011 @ 10:47 am
Croco-Puzzle Update  
Time for my 60 puzzle update on croco-puzzle, this time the "month" also covering my first extended time away from the site (and what a good time that was). I've had a pretty solid 20 days back, with a tagesrating above uvo -- save for one ugly fehlversuche yesterday and a much better score of 2662 before it -- as I (slowly) close the gap to first.

But I still don't view myself playing these puzzles long-term once I reach my 2700 target. I run into pretty familiar frustrations whenever I play because this is really a site for me about competition and not fun and the puzzles are rarely noteworthy given the computer-generation.

Foremost amongst the frustrations is the "crushing" feeling of losing a 1st in the last hours of the day. And the most crushing of these is when the steal is not from a top player, but from either (1) someone (anqu) I view as having cheated in the past, and therefore not worth giving full credit now, as happened on the Sternenhimmel on 20/9/11 or (2) a completely unknown player. In his first two days, gorogoro stole a first from kirarin on an insanely fast Masyu that I cannot replicate by trackball for either solver, and then on his second day stole a first from me with a very fast Bahnhofe where we were both 20% ahead of everyone else.

Now, we often run into the "new Japanese" solver effect, with a quick announcement of skill, as first happened when kirarin and then xevs arrived, and having a new crowd at the top of the Masyu and Heyawake is not unexpected. gorogoro's "bests" suggests familiarity with Japanese styles as is common with the "new Japanese" solver effect. But Bahnhofe is originally a Canadian puzzle type that isn't in as many places. And while I'm horrible at computer entry at loop puzzles, so I know the perfect me with adequate mouse control can shave off the seconds that gorogoro did, it's harder to accept such a performance on a player's second day when no one else, even good loop solvers, was close to me. His performances have fallen a little back to earth in his second week (and he's learning a new word - Fehlversuche!), but my introduction to this solver was a rough one to say the least as I spent a whole week fearing gorogoro whenever I topped a puzzle.

I should also note "jano87" just started in this month too. Based on his birthdate, and the first half of his nom, I'm confident this is Jan Mrozowski/janoslaw. The field gets tougher and tougher. If only the puzzles were better.

And this is my second frustration. I have absolutely no confidence in the puzzle generator for about 6 of these types. As a result, I often have to make a choice between thinking and guessing in the middle of too many solves and suffer from playing the wrong side of this coin at times when on human-constructed puzzles I would never jump to "guess". On the guessing puzzles, uvo and I and others most often lose time to the "lucky" solvers. On the thinking puzzles, we almost never do. I don't mind types (Schlange) that are almost always guessing, as I know what to do from the start and I think everyone agrees the results are a lottery at times. But I hate hate hate Thermometer where some of the grids have no logical path where others certainly do even if it's a bit harder to see. I've overthought my way to bad times, and overguessed my way to mistakes. I'd place Doppel Block in the same category where the generator almost always is too sparse in clues so that there is no "logical" way through, but you can't necessarily know. I managed a first this week in one where I saw the needed contradictions rather fast simply because they were finally intuitable. But it's still not a type I ever want to see on croco-puzzle.

The only nice change in the last month is the end of the ligasystem which I voted down in the German forum as well as about 2/3rds of the others. The voting system suffered in my opinion for two reasons: first having + and - votes is never the right way to conduct a ranking - either let people vote for or against something but don't give both powers - and second the vote was public and you could change your vote as often as you wanted up to the final day. So it seemed to at times resolve to a kind of internet "flame war". You'd see a puzzle (ie Japanese Summen) jump to the top of the rating system one month then get 8 instant no votes on day 1 the next month, to try to sink it before other votes came in. In full honesty, I was one of those 8 instant no votes when that situation arose, but generally because I think there are way too many almost identical arithmetic puzzles on the site and I use my 1st rate percentage by type to prioritize my voting. Here is the data I looked at to start my voting:

# of 3000 scores:
Hashi - 13 of 19 (68%)
Fillomino - 12 of 19 (63%)
Pillen - 6 of 13 (46%)
Schlange - 4 of 9 (44%)
Heyawake - 6 of 15 (40%)
Magisches Labyrinth - 5 of 14 (36%)
U-Bahn - 5 of 14 (36%)
ABC-Box - 5 of 19 (26%)
Rundweg - 4 of 18 (22%)
Gebietssummen - 4 of 18 (22%)
Sudoku - 3 of 14 (21%)
Masyu - 4 of 20 (20%)
Japanese Summen - 5 of 26 (19%)
Vergleichssudoku - 2 of 11 (18%)
Buchstabensalat - 3 of 17 (18%)
Killersudoku - 2 of 12 (17%)
Doppel Block - 2 of 12 (17%)
Hitori - 2 of 13 (15%)
Magnetplattenratsel - 3 of 20 (15%)
Hochhauser - 2 of 14 (14%)
Doppelstern - 5 of 37 (14%)
Tapa - 5 of 37 (14%)
Pyramidenratsel - 2 of 15 (13%)
Kropki - 2 of 17 (12%)
Thermometer - 2 of 19 (11%)
Domino - 1 of 10 (10%)
Zeitlager - 1 of 12 (8%)
Kakuro - 1 of 13 (8%)
Bahnhofe - 0 of 4 (0%)
*Sternenhimmel - 0 of 8 (0%)
ABCD-Ratsel - 0 of 10 (0%)
Slalom - 0 of 13 (0%)
Laser - 0 of 13 (0%)
Arukone - 0 of 15 (0%)

* This will equal 1 of 8 if anqu is expelled again.

Now, there are types where the new make-up of the field makes these percentages unreliable. Heyawake is now more often a sure second or third for me and never a first given kirarin, melon, and xevs (and gorogoro I guess). But my votes would almost always start with Arukone and Slalom and Laser down, Hashi and Fillomino and Pillen up, and I have to admit that almost all Hashi and Fillomino puzzles on croco-puzzle are boring. So I'm not voting to make the site a better place to play puzzles. I'm voting to make it a better place for me to have a high rating. It turns out most of my "best" puzzles are on list 2, and none are the list toppers that get 5+ puzzles a month. So I never really liked the psychology of the voting or the results. Now that it is gone, and random choice will take its place, I hope the "surprise" factor can return and some types we rarely see start popping up more consistently as I'm more interested in the less common types usually. And I'm curious to see what the "harder" puzzle 2 will consistently look like. Obviously just changing the grid size limits for that puzzle will mostly steer the difficulty in the right direction, but I find the 2-3 minute puzzles best at segregating the top solvers, provided they are thinking puzzles and not guessing puzzles.

Onto the stats (last "month"'s stats in parentheses):

Current Rating: 2646 (2623), 2nd of 758 (2nd of 728)
GT7D: -54 (-77)
GTU: -58 (-89)
200D: 2472; 4-Dan (2338; 3-Dan)
Total Puzzles/Solved: 540/540 (540/540)
Top Ranking: 111/540; 20.6% (101/480; 21.0%) - with 1 new double tops (8/9/11) for 12 total
Top 5 Ranking: 299/540; 55.4% (267/480; 55.6%)
Fehlversuche: 41 on 35 puzzles, 6.5% of puzzles (39 on 33, 6.9%)
Categories led: Hashi at 2828, Magisches Labyrinth at 2762, Rundweg at 2757 (Rundweg at 2771, Magisches Labyrinth at 2708, Hitori at 2641)
All-time records: Hashi at 2828 (none)
(Anonymous) on September 27th, 2011 08:27 am (UTC)
Re: Why?
You are very competitive and inspiring. No wonder you always win. I like u.

Mr Nobody
motrismotris on September 27th, 2011 11:19 pm (UTC)
Re: Why?
I actually find my hyper-competitive nature to be one of my worst traits. A little bit of competition can serve as a healthy motivating force towards self-improvement at something, but I think I take this to crazy extremes.

And on further reflection I'm finding almost nothing defensible about my immediate response to the earlier anonymous commenter on this thread. I wrote my first train of thought on seeing the message in my inbox, and it was an instinctual response. But looking back at it, I should have stopped after "I do not like solving domino puzzles" as I seem to have missed the point by arguing further when there isn't so much to rationalize. If something isn't enjoyable, and isn't something you have to be doing, you shouldn't be doing it.

I spent a bit of time after that domino puzzle, which was rather unfun to solve, trying to get to the answer logically. There is a unique 4-3 domino that is the only 1st principles thing I found (other things like a forced 1 in a 1-6 and a forced 6 in a 6-6 actually don't give any walls). There's nothing too easy on 2nd principles either. I couldn't find a way to solve it without eventually bifurcating. Somehow anqu did (or got unbelievably lucky again despite almost all past observed performances suggesting something else). And it's been bothering me all day that he seems to have suddenly become a better solver than me. I find myself being incredibly defensive of my own puzzle rating and the system in general and neither seems very healthy when it's much easier these days to go down than up. Perhaps beating uvo is my white whale and I really need to give that quest up. I'm not getting any better at solving puzzles by playing on croco, and I'm not enjoying the experience, so I think I see that commenter's point more clearly now. Without a good answer to "Why?", I should certainly stop.
(Anonymous) on September 28th, 2011 06:56 am (UTC)
Re: Why?

Mr Nobody
Teesside Snog Monster: puzzlejiggery_pokery on October 1st, 2011 09:47 am (UTC)
Re: Why?
I salute your self-knowledge. Well, specifically, I am jealous of your self-knowledge.

However, however, the croco-puzzle U-ratsel metagame activity is clearly great fun, even if some of the puzzles disappoint. Recognising that writing interesting puzzles is far from easy, I wonder if there could ever be some sort of alternate daily competition which uses croco-puzzle mechanics but uses puzzles that are more consistently of interest? Perhaps there might be some way to assemble a moderately large cadre of puzzle generators so that each individual generator might only have to face (and, presumably, skip) their own puzzles once every couple of weeks. This might well cause more problems than it all solves; at least, in the current fashion, nobody can say that there was any risk of anybody seeing the puzzles in advance. (Though at least one other type of cheating has been seen over the years.) Additionally, I suspect that I am not the only 1200-rated solver (though my 250-ish at LMI is probably more accurate...) who quite likes the fact that the puzzles offered are, generally, standards or only the very slightest of variants.

Or perhaps I am missing the point here?
motrismotris on October 1st, 2011 04:58 pm (UTC)
Re: Why?
I'd say Nikoli is very close to what I want, except they don't set up their system in the best way to log puzzle solving times or try to keep a global ranking or puzzle type ranking at all.

But they do store solving solutions, which is very interesting to review from time to time, and for 50% of their types they even have published times so that, regardless of other solvers on that day, you can compare how you solved with a real standard. You do have to pay to see the puzzles, but over a thousand people do, and I think it is worth the money. All the puzzles are standard types, approximately 12 styles (a new one adds every year or so), with a majority of these also on croco-puzzle (Slitherlink = Rundweg, Masyu, Hashi, Hitori, Sudoku, Kakuro, Heyawake, Numberlink = Arukone).
MellowMelonMellowMelon [wordpress.com] on October 2nd, 2011 04:18 am (UTC)
Re: Why?
While there are occasional glitches like people with bugged flash timers (NPMcH) or others who clearly peek at other solving histories before trying a puzzle (halca), one can obtain a mostly faithful leaderboard for each puzzle sorting by open count first and then time. You'll knock out both re-solvers and people who broke the puzzle, reopened it, and solved it much faster than they would have on a first attempt, and those should not be included. You will also knock out people who reopen a puzzle but don't solve a second time, which can easily include legitimate solvers that glitched out or something, but that's just kind of a necessary evil.

But it is true there is no global ranking system on nikoli.com. It sounds like Ziti tried something sort of like this a long time ago, although that was before I had signed up for nikoli.com (and thus before anyone had any inkling of who I was). I wonder if there would be a way to do some sort of unofficial "fanmade" ranking system, with the formulas resembling those used on crocopuzzle, and with some moderation to catch obviously illegimate times. But you'd want to automate the collection/analysis of leaderboard data for the maintainer to keep their sanity, and I wouldn't begin to know how to do that on a technical end.
motrismotris on October 2nd, 2011 05:17 am (UTC)
Re: Why?
The short-lived Nikoli League aimed to do this but with self-reporting of scores for about 8 players. The limitations were inconsistent players and self-reporting needing to be done. Making a leaderboard for more than just 8 players would be where this would become a meaningful thing, and for that I think some kind of automatic extraction is necessary. I know how to do that on croco-puzzle since everything is in a single page. I'd imagine on Nikoli you'd need more of a spider that can access multiple pages. Still, it would be nice to attempt sometime.