Thursday, June 10, 2021

Hexagony Results (and Post-Mortem)

Hexagony, the puzzle contest constructed by Eric Fox and me, has now concluded. Before getting into the results, I want to take a bit of time to talk about the conception process of this contest. If you're not interested, then scroll down a bit for the results and stats. At some point last year, I had the idea to have a contest where half of the puzzles had hexagonal grids as a surprise twist. This was the entire conceit, and the entire point, and I brainstormed about a dozen possible genres that could have the same wording of the rules work for both a standard grid and a hexagonal one. I thought some savvy solvers might guess the gimmick in advance, and PuzzleCraig was actually half right shortly after the first draft of the instructions booklet- instead guessing everything was hexagonal. Kudos to him! I also briefly considered Honey Islands as an inversion- with a surprise square grid- but thought that would ruin part of the surprise and I also really didn't want to make them. This was originally going to be a Puzzle Ramayan style contest- 22 puzzles, 90 minutes, and if it wasn't for the surprise twist being potentially unfair, I would have asked Rakesh if LMI could host again because it went well for Puzzle Zodiac. The other mission statement that I picked up partway through was to make a low-stakes, fun contest that even newcomers could feel comfortable giving a try: given how many new names I saw in the results, I'd say this was a complete success- and without sacrificing the experience for the rest of the solvers, too. One of the biggest successes that I hope to see other contests emulate is the casual entry option- there's a lot of people who just want to solve the puzzles used in competitions without having to sign up and get a 0, or wait until the period is over. I had 34 different people take this option, compared to 95 competitors. But that's getting ahead of myself, a bit.

It took until March of 2021 to decide to really get this into motion, and I reached out to Eric Fox with an offer to help with test-solving and other checking the back-end parts of the process. While I was typing the message, I realized that a collaborative construction process could be a great benefit as our construction styles and strengths differ somewhat, and we quickly realized we had more than enough ideas to extend my original idea of 11 genres out to 15. The construction and testing process went smoother than any contest I've worked on before- Eric was a wonderful partner, and having ever puzzle tested shortly after it was made really helped compared to separating the making and testing processes, and we were done constructing by the start of May. There were only a few ideas that needed to be scrapped- I made a hexagonal Heyawake that was broken from the break-in and didn't realize, Eric had a failed attempt at a hexagonal Star Battle before the successful one that I caught the same day, and I also made a second Spiral Galaxies (the starting image for this post, in fact) for a potential lead-off puzzle, before Eric suggested going with the other one for being more logically interesting. It was obvious in hindsight, but it's hard to be objective about your creations. We then brought in calica, TostCronch, A Sickly Silver Moon, and shyeheya for checking the rules and getting test timings. Of these four I'd especially like to thank calica for continued feedback on administrative decisions and always being there to help check over things that needed checking.

Alright, onto the actual contest results! As I said I would do, some answers were manually given credit despite the written key being incorrect in some small way, if that small way was both obviously incorrect and close to the correct answer. I'll call out all the changes I made below the results table, and include some unadjusted scores if you're against the idea of manual corrections. Additionally, the rules I wrote didn't adequately cover what to do for late submissions- for the future, I'll incorporate a TVC-style penalty for submissions up to 5 minutes late, and anybody more than 5 minutes late is just out of luck. For this contest, I elected to let it slide up to 30 seconds late after the first late submission, and for any later submissions, I would reach out to the solver to determine what was solved last (and thus, late) and not score that puzzle. Again, I'll call out where I did this, but I consider the "Score" column below the true results.

*this submission was more than 30 seconds late, anything solved after this time was not scored, but I do not have an earlier submission time to use

SolverScoreElapsed TimeCorrectAttemptedUnadjusted
2.Freddie Hand5651:56:122830510
7.Prasanna Seshadri4352:13:08*2424325
10.Wei-Hwa Huang/onigame3851:52:532020385
11.Ben Cosman3701:58:252121370
17.Nikola Zivanovic3201:55:381819320
17.Joseph Howard3201:56:002021295
25.Marco W.2901:45:021818290
25.James McGowan / kiwijam2901:54:301415290
25.Aubin Danzo2901:57:302123290
28.Gavriel Hirsch2851:57:221719260
32.Rajesh Kumar2551:57:321414220
33.Jack Bross2451:51:531515245
40.Adam Bissett2101:42:491313210
44.Jaipal Reddy1851:57:281011150
47.Tom Coward / Polman Poppins1702:01:10*1012155
49.Nicholas Allendorf/Shogia1651:33:1899165
51.Gennady Korotkevich (tourist)1501:58:201111150
54.stefano forcolin1401:26:461010140
54.Bob Zwetsloot1401:53:4888140
57.Blecon François1351:58:22911135
57.Michael A. (randomperson1729/Some Guy)1351:58:3399135
61.Sébastien Tainon1202:03:35*890
63.Nikhil Khetan1151:47:5488115
63.Veronika Kolvekova1151:52:5599115
66.Pontier Garance/ Rubia1101:52:4478110
71.Paul Revenant951:59:328895
74.Swagatam Islam Sarkar851:38:506685
74.Christian Romon852:00:135585
78.Sean Day701:47:257870
79.Gaurav Kumar Jain651:39:215755
83.Puzlifouk (Nicolas Sigler)501:40:153350
84.Chandrachud Nanduri451:43:064445
87.Markku Lahti301:35:222430
87.Paul Smith301:51:443330
91.Rakesh Rai00:00:00000
91.Alexander Kirlitsa00:00:00000

Congratulations to Ken Endo for winning Hexagony and being the only clean finisher of the contest! Also congrats to Freddie Hand and Walker Anderson for scoring 500+, and rob, EctoPlasma and Tiralmo for a three-way tie for 4th at 445. I'm blown away by the fact that this had 95 entrants in total, with another 34 casual participants. Going in, I was anticipating about 25, maybe 50 on the upper end. Thank you to everyone who participated- you all made the extra effort I put in to run this as a contest instead of a stand-alone set of puzzles completely worth it- and I put in so, so much extra effort. The back-end stuff was mostly a solo effort with no established infrastructure, and I knew that going in. Even still, it took more than I expected.
Before getting into why I made the scoring adjustments I made, here's some stats on the puzzles themselves.
1Spiral Galaxies3131100.00%
2Spiral Galaxies343694.44%
7Bosnian Road404490.91%
8Bosnian Road353794.59%
13Star Battle575996.61%
14Star Battle475094.00%
23Double Choco4848100.00%
24Double Choco282996.55%
27Ripple Effect222491.67%
28Ripple Effect66100.00%
29Password Path555894.83%
30Password Path4242100.00%
The most submitted puzzle was Akari 1- I expected this would be the case as one of the easiest puzzles in the entire contest, and so I placed it early on to give people something manageable to work on at the start. The least submitted was Ripple Effect 2, to no surprise as the highest point value puzzle on the contest. Freddie Hand, Tiralmo, Swaroop, Ken Endo, sinnedk, and Wei-Hwa Huang were the only solvers who got through this very tough challenge.
10 puzzles had no incorrectly graded submissions, with Cave 1 having the most submissions of these 10. Conversely, Cave 2 had the highest error rate of all the puzzles- from experience when constructing, miscounting a clue is surprisingly easy here. Overall I think the point values were largely representative of the difficulty, unfamiliarity, and time to solve, with only a few like YK2 being a bit off, but always in the high direction.

In ranking order, any changes I made to the score from just tallying up the correct keys:
  • Many people: Akari 2 and Star Battle 2, marking "column" numbering as correct. I thought "index" numbering would be more intuitive as it was for me, I thought wrong. Both of these were marked correct.
  • Ken Endo: Completely clean, no manual adjustments necessary other than calculating a 50 point bonus for finishing 8 minutes and 20 seconds early. Again, well done!
  • Freddie Hand: Spiral Galaxies 1 and Tapa 1 both have "111" instead of "1111" in the codes. Awarded because a checksum (such as "the key must have a sum of 10") would catch and prevent this.
  • Walker: Slitherlink 1 has the answer keys in the wrong order: both given keys are correct.
  • rob: Aqre 1 is missing the second key. I deliberated on this for almost a week before deciding to award full points. The first key both is last to resolve and contains the critical connections, and splitting the entry field into multiple fields as most contests would do would prevent this from happening.
  • Prasanna Seshadri: Prasanna submitted at 1:35, 2:13, and 2:38 from misunderstanding the time limit. I reached out in order to determine what puzzles were definitely completed on time. Specifically, the 2:13 submission was after being stalled on Ripple Effect 2 for "a while" and working mostly in order. It's possible that Ripple Effect 1 was completed on time, but certain that everything else submitted was.
  • Anderson: Ripple Effect 1 first key is missing its final digit, however that digit is specified by the remainder of the keys.
  • Wei-Hwa Huang: An additional submission was made 1 minute late, containing a solve of Kakuro 2. This submission was not scored, but it was correct. I'm noting these instances for transparency, and so that if you believe I should have applied a "reverse bonus" for late submissions, you may do so yourself.
  • apiad: Four solution keys across three puzzles contained an extra 1: Spiral Galaxies 1, Kurotto 2, and Bosnian Road 2. All of these were part of a string of multiple 1s, and would be caught by a checksum.
  • tckmn: Over the course of multiple submissions (not edited submissions!) the key for Bosnian Road 1 was submitted over Bosnian Road 2. As both keys were correctly submitted and this wouldn't have happened on editing submissions, points for both were maintained.
  • Joseph Howard: Yajisan-Kazusan 2 begins "314..." instead of "315...". The second key is correct and would force 315, and a checksum would catch this.
  • qw014052: Answer keys for Double Choco 1 were reversed.
  • pleiades: Aqre 2 2nd key is missing a number (21211 instead of 212211), but the 1st, more error prone key is correct and a checksum would catch this error.
  • ksun: Cave 1 has a checksum error- first key sums to 10 and not 11, clearly a miscount or typo. Kakuro 1 has a transposition that I felt confident was a typo, with the first key ending "17" instead of "71". The former would place both digits into sums where they immediately break, and so I awared the points.
  • not_coal: A correct submission on Aqre 2 was made 33 seconds late. With "reverse bonus" scoring, this would be worth 316.7 points. I feel really bad about this one being on the bubble of how lenient I was being, but changing the rules based on the circumstances that arise was a slippery slope I wasn't going to go down.
  • James McGowan: An additional submission was made 1 minute late, adding a solve of Star Battle 1. This was not scored.
  • Gavriel Hirsch: Cave 2 answer keys were input in the wrong order. This was marked incorrect for the first posting of these results, and has since been corrected to match treatment of other key transpositions.
  • Swaroop: Swaroop made multiple submissions: the last was 12 seconds (-1.2 points, if I were to apply the reverse bonus) late, adding only the first Kakuro (15 points). This was scored.
  • sinnedk: Bosnian Road 2 second key begins "2321" instead of "233": this would result in having a shaded clue, so it's likely counting the "2" clue as its own entry by mistake.
  • Rajesh Kumar: Spiral Galaxies 1 has a miscount (151 instead of 141) at the end of the 2nd key, a checksum would catch this.
  • Whitehill9: Multiple submissions were made, the last (17 seconds late) added only Password Path 2. As with the rest of the submissions up to 30 seconds late, no adjustment was made.
  • Qhaces: Slitherlink 1 is missing half of the second key: I decided to award the points because of how the keys intersect, and the order of resolution in the puzzle. Password Path 1 (5 points) was added in an extra submission 25 seconds late, which as mentioned I decided to score as if it was on time.
  • ft029: ft's only submission came 30 seconds late: I had decided to not penalize submissions up to 30 seconds late before this submission came in as the first late submission as only submission. The alternative would be a "reverse bonus" for going over time, which would cost 3 points here. I didn't think this would be fair to apply to everyone (it would reward breaking the submission time, and penalize those who stop a bit early) nor did I think it would be fair to apply unevenly.
  • Roxis: Akari 2 key begins "46004" instead of "46005". Awarded because placing a bulb at (5, 4) would immediately break the central 2- this is either a typo or a miscount. Additionally, a submission was made 1 minute late, adding Password Path 1, which was not scored.
  • Jaipal Reddy: Fillomino 1 1st key was submitted as "77544431444", which is the beginning of the correct row followed by the remainder of row 1. This is too many characters, and the 2nd key is the more critical one, and so I elected to award the points. Star Battle 2 used column indexing, except ending "41586" instead of "43586". Row 8 does not contain a cell in "column 1" and so it's clear that this was autopiloted.
  • Polman Poppins: Polman Poppins submitted only once, 1 minute late: as the standard for > 30 seconds late submissions, I would remove anything solved after that time, which in this case was the 2nd Aqre (and it was wrong anyway!) Additionally, Fillomino 1 first key was missing a given 6.
  • stefano forcolin: An additional submission was made 1 minute late, adding Tapa 2. This was not scored.
  • Sébastien Tainon: Only submission was 3:35 late, and I tried to get in contact to determine the last puzzle solved but could not. I removed the highest scoring puzzle submission (Kakuro 2, which was correct) and scored the rest. If I were to apply "reverse bonus" scoring to this submission, it would be worth 153.5 points. Along with rob's missing key, these were the two hardest scoring decisions I had to make.
  • dragoon: Slitherlink 1 was the last puzzle worked on, it was not scored for being over time. Akari 2 contains an index miscount ("9179" and not "8179" for column indexing- two 9s there would have two bulbs seeing each other), and Fillomino 1 has the second key begin "33,333" instead of "331333". The 1 is a given digit, and so I could determine it was correct.
  • fu-nantoka-san: Password Path 2 answer keys were input in the wrong order, but both were correct.
  • jamesa7171: The erroneous submission is literally "something,7413256" on Kakuro 1. While technically correct, I can't actually award the points here. This wasn't an adjustment but I did think it was funny.
  • Christian Romon: Submission was within 30 seconds over, allowed without adjustment.
  • 嘉和逆天: A submission with all puzzles was made at 2:50. If I were to apply "reverse bonus" scoring, the score would be 575 (Cave 2 incorrect) - (50*6) = 275 points. As no contact information was given, I scored only the last intermediate submission, 20 minutes in.
  • Gaurav Kumar Jain: The solution to Akari 1 was placed in the box for Akari 2.
That's everything. Whew, I had a lot write on this. I'll post my puzzles with comments on each set (and possibly solution walkthroughs? If you're reading this far, let me know if you want them or not!) in the near future, but for now? I'm going to relax.

Thanks again everyone!


  1. Awesome puzzles. Thanks for this contest.

    1. Thank you! I'm very happy with how it was received on the whole :)

  2. Replies
    1. I linked the solutions booklet in the landing page post- if it's not accessible, let me know.

  3. Thanks for the test, great puzzles ! I hadn't seen that hexagonal thing coming, but it was a nice refreshing surprise.
    If you are willing to do them, yes I would definitely appreciate a lot solution walkthroughs :)

    1. I was slightly worried about how the surprise hexagons would go down, but it seems to be universally a positive response.
      If I wasn't willing to do solution walkthroughs, I wouldn't have mentioned it as a possibility. I'll probably start on these over the next few days.