ACPT By the Numbers 2016 (Pt. 2)

This post is going to look more at completion and error rates, comparing this year to historical (-ish) averages; part 3 will look at the “par score” concept (which doesn’t require the full stats, so I can do that for more years).

We have “full” stats (including # of minutes and # of errors) for all the tournaments from 2010 onwards, so for this post “history” will be 2010-2015.  This year the puzzle sizes changed, which may make part 3 a little weird (since that involves scores on individual puzzles), but this post will just involve percentages so we should be fine.  With all that said, here we go:

  1. Puzzle 1 historically has 94.9% complete and 14.2% Oops!; this year was 90.8% complete and 11.2% Oops!  This is the lowest complete rate in my records (previously 2010 was 92.4%) which corresponds with my opinion (which seemed pretty widely-held) that this was one of the hardest puzzle 1’s we’ve had.  (Looking ahead, only puzzles 3 and 5 had a lower completion rate, making this probably the first tournament ever where puzzle 1 was the third-hardest puzzle in the event.) Since Oops! is a strict subset of the complete rate, a lower complete rate will naturally knock down the error rate too, so that doesn’t seem unnaturally low (the lowest Oops rate for a puzzle 1 was 2013’s 7.7% out of 96.5% complete).
  2. Puzzle 2 historically has 65.7% complete and 19.5% Oops!; this year was 91.4% complete and 33.2% Oops!  This is, by far, the highest complete rate for a puzzle 2 (the previous high was 2013’s 80%)  This is the first year in the records, and probably the first year ever, that puzzle 2 played easier than puzzle 1.  I have, somewhat empirically/arbitrarily, decided that an Oops! rate that is more than one-third of the complete rate suggests there may have been a tricky crossing somewhere; by this guideline, four of the seven puzzle 2’s in the records have had such a thing, so that’s fairly typical.
  3. Puzzle 3 historically has 58.5% complete and 18.4% Oops!; this year was 79.1% complete and 31.1% Oops!  Again this is unusually high; the previous high was 2011’s 68.3%, so it seems unlikely that this was a re-purposed puzzle 5.  Again a rather high Oops! rate suggesting a tricky crossing; one of three puzzle 3’s with a high rate.  As a pure number this is also by far the highest Oops! rate, but again that’s tempered by the unusually high complete rate.
  4. Puzzle 4 historically has 91.5% complete and 20.3% Oops!; this year was 92.0% complete and 6.5% Oops!  This is the cleanest puzzle 4 we’ve ever had (the previous low Oops! rate was 11.2% in 2014), although there’s only been one puzzle in the records with a suggested tricky crossing (last year’s).
  5. Puzzle 5 historically has 22.4% complete and 5.8% Oops!; this year was 37.3% complete (highest in records) and a 13.3% Oops! rate (also highest in records).  In fact this Oops! rate is high enough that it trips my sensor for a tricky crossing, although I don’t remember much buzz about it (and I had to leave before getting copies of the puzzles, so I can’t yet check).  This is the only puzzle 5 with an error rate above that threshold.
  6. Puzzle 6 historically has 93.1% complete and 32.5% Oops!; this year was 93.9% complete and 20.2% Oops!  This year, this was the easiest (by completion rate) puzzle, which isn’t all that uncommon.  Puzzle 6 generally has a rather high error rate, even when the tricky-crossing sensor doesn’t go off (whether because of size, or it being late in the afternoon, or just as a reaction to puzzle 5), so this year was relatively clean.
  7. Puzzle 7 historically has 81.9% complete and 31.3% Oops!; this year was 93.7% complete (highest in records) and 23.2% Oops! rate (lowest in records, even without adjusting for higher solve rate).  Puzzle 7 also generally has a high error rate (the historical range 26-40%); a few meet the threshold for tricky crossings, but even the ones that don’t are fairly high; again I’m expecting that to be mostly based on size.

I had definitely heard buzz this year about puzzle 1 being harder-than-average and the puzzles being generally cleaner (in terms of fill) than usual, and both of those are borne out by the data (the latter looking at the lower-than-usual error rate).  That last, certainly, is a trend I think we’d all like to see continue.

ACPT By the Numbers 2016 (Pt 1)

I’m going to do some historical comparisons in part two, I think, and get this year’s stuff up now.

For each puzzle, I’ve got some numbers: the completion rate is the % of solvers who either turned in their puzzle early or had a completely correct puzzle, while the Oops! rate is the % of solvers who turned in their puzzle early and had an error.  (It might make some sense to report the Oops! percentage relative to only the completed puzzles, but I keep the bases the same so that the difference between them is exactly the % of correct solvers.)  I’ve also calculated the “par” scores for B, C, and D; that is, the 80th percentile (for B–top 20%), the 60th percentile (for C) and the 35th percentile (for D) on each puzzle, which should give you some idea of what you would need to achieve those skill levels.  I’ll also provide my numbers, because it’s my blog and you’ll just have to deal.  I have excluded any competitor who did not finish at least four puzzles (and any non-humanoid competitors), so my official competitor count is 573 this year.

Puzzle 1 (15×15, 15 mins): Completion rate: 90.8%, Oops! rate: 11.2%; B par score 1155 (correct with 9 mins on clock); C par score 1105 (correct with 7 mins on clock); D par score 1030 (correct with 4 mins on clock).  [Me: correct with 10 on clock = 1180]

Puzzle 2 (17×17, 25 mins): Completion rate 91.4%, Oops! rate 33.2%; B par score 1410 (correct with 16 mins on clock); C par score 1335 (correct with 13 mins on clock); D par score 1135 (correct with 5 mins on clock)  [Me: correct with 20 on clock = 1510]

Puzzle 3 (17×17, 25 mins): Completion rate 79.1%, Oops! rate 31.1%; B par score 1435 (correct with 13 mins on clock); C par score 1310 (correct with 8 mins on clock); D par score 1110 (correct puzzle with no time bonus)  [Me: one error with 18 on clock = 1365]

Puzzle 4 (15×15, 20 mins): Completion rate 92.0%, Oops! rate 6.5%; B par score 1255 (correct with 13 mins on clock); C par score 1180 (correct with 10 mins on clock); D par score 1105 (correct with 7 mins on clock)  [Me: correct with 16 on clock = 1330]

Puzzle 5 (19×19, 30 mins): Completion rate 37.3%, Oops! rate 13.3%; B par score 1435 (1 error with 12 mins on clock); C par score 1100 (110/118 words); D par score 790 (79/118 words)  [Me: correct with 15 on clock = 1705]

Puzzle 6 (19×19, 30 mins): Completion rate 93.9%, Oops! rate 20.2%; B par score 1725 (correct with 19 mins on clock); C par score 1650 (correct with 16 mins on clock); D par score 1500 (correct with 10 mins on clock)  [Me: correct with 23 mins on clock = 1825]

Puzzle 7 (21×21, 45 mins): Completion rate 93.7%, Oops! rate 23.2%; B par score 2325 (correct with 31 mins on clock); C par score 2200 (correct with 26 mins on clock); D par score 2025 (correct with 19 mins on clock)  [Me: correct with 36 mins on clock = 2450]

As a note: the B par scores add up to 10740, which would be 89th/573 (the actual cut score is 10525); the C par scores add up to 9880, which would be 206th/573 (the actual cut score is 9740), and the D par scores add up to 8695, which would be 371st/573 (the actual cut score is 8685), so on a per-puzzle basis the par scores for B and C seem to be a bit inflated, I expect because a number of solvers at these levels don’t hit par every time (for instance, a B par all the way across would be a perfect solve, and lots of B solvers have an error somewhere, so they’ll likely miss par on that puzzle — I didn’t even make B par on puzzle 3, and I’m not supposed to even be a B any longer, in theory).

As I said, I’ll do the full historical thing in a later post, but I expect that puzzle 1 will be harder than puzzle 1’s tend to be (in terms of completion).  I remember hearing some moaning about common error squares for puzzles 2 and 3, and not about 6 or 7, so that error rate being about 10 points higher probably leads into that.

While we’re here, the geographical breakdown:

NYC 122
Other Mid-Atlantic 86
Other New England 70
West 63
Westchester/Upstate NY 50
Midwest 48
New Jersey 42
Connecticut 36
Long Island 24
South 23
Non-US 8 [there were 6 listed in the program, which suggests that 25% of the “field” for this showed up on the day of, just because]

Part two to follow later this week.

The NYC Riichi Open: Thoughts From the Curmudgeon

(The following are based solely on my opinions and perceptions, so may be wildly inaccurate depending on how badly I perceive.)

I’m not sure when I played my first game of mahjong.  I think it was at the end of college (maybe the beginning of grad school), when I was on a kick of trying to learn basically every card (or in this case, card-like) game that existed.  (Bridge!  Skat!  Pinochle!  Tarot the card game!)  I was telling people at NYC Open that my first exposure was Tom Sloper’s Activision game, but upon further review I think I had found Arto Tenkanen’s Four Winds Mahjong first (and it’s the one I still have on my computer now).  That program had (and has) a large number of rulesets in it — starting out learning, I began with “Chinese Classical” which is more-or-less the “original” game as I understand it, but then moved on to the International Rules that were just being developed around this time (we’ve moved forward in the timeline to something like 2002 by now, as I think I had gotten my then-fiancé to play CC) and probably would have started playing online at Mahjong Time maybe a year or two thereafter — my account on MJT says I joined in April 2007, but I’m pretty sure I had been playing before that there (at the very least before I finished grad school, which was in 2005).  Based on what I had found, Riichi seemed to have a bit of a reputation as a confusing game, and the one or two hands I played against my computer didn’t change my mind too much, so I focused on International Rules (I have never played American Mahjong) exclusively, and ended up playing on the “American team” at the 2007 OEMC in Denmark.  Riichi came to my attention again there, since Jenn Barr (pro player in Japan) was there for moral support (and gonging the gong once, as I recall).  I took a look around again at Riichi Mahjong in the US, and as least as far as I could see, it was an offshoot of the Japanophile/anime scene. This isn’t per se a problem, as everyone who plays has to find out about the game somehow.  Unfortunately what I saw (which may have just been the people most active online, and not the core group who were more active inseat, I can’t say) were mostly people who wanted to “be” the anime (hence an obsession with huge hands, and crazy tactics, and those three-quarter-clear sets), and I decided that wasn’t where I wanted to be and stuck with International rules.  At some point a few years ago, I finally got a bit fed up with some of the weirdnesses of International Rules (such as triple points for self draw) and looked back around at some other variants and gave Riichi another try.  Whether I was just now ready (ha!) for it, or I just took it more seriously than I did originally, I managed to pick it up and come to enjoy it; and although I still play both online I probably play more Riichi than International Rules, and have even “medaled” in some of MJT’s online Riichi tournaments.

Which brings us to last weekend: the NYC Riichi Open where 40 people from all over (well, all over the Northern Hemisphere I guess) convened in Manhattan to play.  I was fairly nervous, as I hadn’t actually played in person since the 2007 OEMC eight long years ago, and I knew I was going to suck at actually handling tiles and keeping track of things on a real table,  All of that happened, but I want to give a lot of thanks to all the other participants who were pretty uniformly tolerant of my taking a little extra time to do all the little things that need to get done each hand.  I was really expecting to play a lot worse than I normally do online; fortunately I think only played a little worse — there were only two sessions (the 1st, where I was still adapting, and the 6th where for some reason I completely came apart and couldn’t even tell when two tiles were the same and inadvertently broke up several pairs) where I felt that my play was being seriously compromised.  All those other poor performances I earned, often by not playing very solid defense which is always a weakness of mine (my internal meter is apparently permanently biased towards earning points rather than not losing points), sometimes getting outdrawn, sometimes just not having much of a hand.  I had, in a ha-ha-only-serious way, been telling people that I was going to go up there and get last place, and my finish of 38th/40 is close enough to that.  The up side of that finish is that I was glad to see that we had some solid groups of people who were good players of the game and took it seriously (even if they all like their anime and maybe even have one of those three-quarter-clear sets, because after all we all have our guilty pleasures).  The unserious players we will always have with us — I heard a couple people complain that they would have many many sessions of teaching new people to play at anime cons or anime clubs or etc and most of them would never come back — but we have several clubs with enough serious players to be sustaining.

One issue I had that I had not expected (but probably should have) was language.  I had “come up” through the ranks of CC and International Rules, where all the hand names are in English; and since the only hand in Riichi that isn’t in International is, well, Riichi (and all the variants of riichi, and dora), that’s basically the only hand I know the Japanese name of.  However, everyone else was counting their hands in Japanese.  I could look at their hand and see three doubles, and they would count off three things, so I always assumed that those were the right three things, but I never actually bothered to try to match them up.  If I ever go to another in-person tournament (and I hope to) I will definitely have to brush on my hand names (and my tile skills, and my defense) first.  (I don’t know whether it will be good for the overall game to stick with Japanese names for the hands, or switch to English — having things be exotic-sounding is probably simultaneously a draw (for some people) and a repulsion (for some other people), and if everybody sticks with Japanese that would ease overall communication I suppose.)

I’ll finish with some thanks.  Many thanks to David Bresnick (& USPML, to the extent that they are different) for organizing this and keeping everything running, and to the JPML representatives (Messrs. Moriyama and Yamai) for taking time out of their schedule — and making quite a trip! — to come play with us and running strategy sessions (even if I was way too mentally exhausted to stay for any of it).  It was a lot of fun, even I have nothing to show for it, and am looking forward to losing again next year.

Blog Housekeeping and a Puzzle

Just spammed a bunch of comments, all from the same company (they apparently go through Marketing Directors like water, based on how many different names there were in the comments), so everything is all sparkly again on the backend of the site.

I’ve spent some time today typing in a lot of numbers under “Montgormery Blair” for PACE NSC, and in honor of their continued journey through the tournament I figured I’d post one of the puzzles I wrote for their most recent Puzzlepalooza contest they hold for their magnet students at the end of school (culminating again this year in an off-site “event”.)  They would have had to have done some cosmetic work on the puzzle before fitting it into their hunt, because I don’t know what sort of theme they had going on, and I don’t know if they did any editing, but this was the puzzle as submitted:

twaINlay

  • They sell more root beer than hamburgers these days
  • Top-level storage?
  • Write notes on your hand, perhaps
  • Gently admonish
  • Give way to
  • Be hugely larger than
  • Obliterate
  • Persian, these days
  • A bond that isn’t covalent
  • Elie Wiesel memoir
  • You get one in every box of Cracker Jack
  • FM receiver
  • Building manager
  • Marty from Madagascar

I think the students knew (from context, if not from a rule) that the answer was two words.  If you want to show your stuff, you can do so in the comments (but you should use ROT13 for actual answers/spoilers).

ACPT By the Numbers (pt 2): Puzzle breakdown

For the next installment, I’m going to go back over the past few years and see how this year compares to previous years; but for now, we’ll focus on this year.

For each puzzle, I have two measures of hardness: one is completion rate (how many people turn in either (a) a puzzle with at least 1:00 showing on the clock or (b) a completely correct puzzle (even if at the end of the time)), while the other is the “oops!” rate: of the people who turned in a complete puzzle, how many have an error on that puzzle.  Now admittedly there are non-puzzle reasons for people to have an error (moving too fast and misspelling a word that doesn’t get noticed, etc), but presumably those will be roughly even throughout all the puzzles, so I’m still expecting that a puzzle with a high “oops!” rate (yes, I’m going to write that with the quotes and the exclamation point every time, thanks for asking) has some sort of tricky thing going on (an obscure crossing, or an “alternate” answer for a clue that doesn’t actually fit the crossers, etc).

As with the previous post, I have removed all non-human competitors, as well as all human competitors who finished three or fewer puzzles.  So by completion rate, the puzzles are then (from easiest to hardest):

  1. Puzzle 1, 95.7% completion rate
  2. Puzzle 6, 95.7% completion rate (this is actually an exact tie)
  3. Puzzle 4, 92.7% completion rate
  4. Puzzle 7, 85.7% completion rate
  5. Puzzle 2, 61.7% completion rate
  6. Puzzle 3, 55.5% completion rate
  7. Puzzle 5, 19.9% completion rate

By “oops!” rate, the puzzles are (from easiest to hardest):

  1. Puzzle 1, 21.1% (114/540)
  2. Puzzle 6, 23.3% (126/540)
  3. Puzzle 7, 29.2% (138/472)
  4. Puzzle 5, 31.3% (35/112)
  5. Puzzle 3, 32.9% (103/313)
  6. Puzzle 4, 36.7% (192/523)
  7. Puzzle 2, 50.9% (177/348)

So puzzle 2 was the hidden landmine this year, while puzzle 5 was the dreadfully obvious landmine.  That 50% “oops!” rate seems awfully high—that’s one of the patterns I want to look for in the previous years’ data, of one high “oops!” rate puzzle.

I also have the median times/wrong squares for each puzzle (recall: median means that half the field has more than that time/wrong square count, and half has fewer):

  1. 8 minutes, 0 wrong squares
  2. 5 minutes, 2 wrong squares
  3. 2 minutes, 2 wrong squares
  4. 10 minutes, 0 wrong squares
  5. 0 minutes, 63 wrong squares
  6. 16 minutes, 0 wrong squares
  7. 20 minutes, 0 wrong squares

Which means that the “typical” ACPT solver turned in all the puzzles (but puzzle 5), with four of them clean.  If I can think of a nice way to do it, I may try to get histograms online for the solving times.

ACPT By the Numbers (Pt 1): ACPT Geography

Getting this series restarted this year, and beginning with a look at where the contestants came from.  The last three years have been fairly steady in contestants: 570 in 2013, 580 in 2014, and 564 in 2015 (I’m only counting human contestants who turned in at least four puzzles, so only showing up for one session wouldn’t count).  Unsurprisingly, Connecticut just about doubled, going from 5% in 2013 and 4% in 2014 to 8% in 2015.  Also unsurprisingly, New York City went down, from 24 and 23 percent while hosting to 18% in Stamford.  Other New England also went up a couple of points, and West and Foreign went down (I don’t know if I can actually say that Foreign entries went down “precipitously”, since they only started at 2% and finished at 1%).  Whether that’s extra travel woes on top of an already long trip, or other random commitments, I don’t know (but maybe we’ll find out next year).  Westchester/Upstate NY has been increasing steadily over the past few years, for whatever reason.

And of course, there is the fact that the “West” has four of the top ten and six of the top fifteen contestants in 2015 with 11% of the contestant body to pick from.

Tables and charts after the break.

Continue reading

Charity Puzzle Contest Update

So normally I aim for my Valentine’s Day-themed charity puzzle contest to come out, you know, around Valentine’s Day. This year, though, I was traveling a lot and reading a lot of quiz bowl in February (which would have made it difficult to deal with sending out the puzzles since weekends tend to be when I get most of the sign-ups) and, for that matter, I haven’t finished all the puzzles yet. Consequently I am aiming for the contest to run around Anti-Valentine’s Day (a/k/a the middle of August) which hopefully will allow me to both get everything polished in the puzzles and set aside some time to send out puzzles and deal with registrations.

So be on the lookout for that in a few months!

Has it been a year already?

Since the last post? I guess it has, meaning I never did my ACPT follow-on post this year either. Hmm.

Nevertheless, I come bearing a puzzle! This was a puzzle I wrote for this year’s Puzzlepalooza at Montgomery Blair High School, that got massaged a bit when used up there (ie the cryptic bit got taken out). But if you like cryptics, or ciphers, or maybe even both, then this puzzle is the puzzle for you. (There may be a teeny-tiny flaw in the cipher part, but it shouldn’t keep you from figuring out what’s going on.) As a bonus hint, your final answer is a person.

The sad part? I haven’t really looked at the puzzle since I wrote it (last summer), and I remember the parts that were important to the overall puzzle. The filler: not as much. Can I still solve my own puzzle? Would I get beaten in a time trial by Erik Agard on it? (Hint: yes.) I’ll post in the comments one way or another!

Look! Land! I mean, a puzzle!

I am posting a puzzle I wrote for the most recent edition of Montgomery Blair High School’s Puzzlepalooza, which finished a week or so ago. Note: The first part of the puzzle (i.e., the first page) leads to an instruction that (should have) made sense for the students, as it refers to a physical prop they had been given; it might make less sense to you (but you should at least recognize that it is English). Once they had successfully done the action, they got the rest of the puzzle (which for you is on page two).

You are looking for a one-word answer.

Thanks to the Puzzle Lords for allowing me to republish the puzzle.

Sports Widow wrap-up

The 2013 edition of my Valentine’s Day puzzle contest has ended.  You can go to this very basic page to access the puzzles and, perhaps more importantly, the answers.  (I have made a few comments along the way in the answers as well, so you may want to read them anyway.)

The important business: giving out money!  Fifty-two people sent me a correct answer, and I had random.org shuffle the names around.  The $50 went to Yossi Fendel, who took a gift certificate to ThinkGeek, and $25 to Steve Williams, who took a gift certificate to Amazon.  (These will be filled in with whatever info the winners are willing to share.)  (If you’re curious as to where you might have landed, you can email me and I’ll look you up.)  Also I just wrote a check (well, not actually, but that sounds better than “Paypalled”) for $955 (and a little bit of change) to Transitions.  Adding in the UK donations to Refuge, we are somewhere north of $1000 in money raised.  So THANK YOU very much — such generosity coming in is really humbling.  (And those of you who know me know how hard that is.)

Please use the comment section for questions or remarks or grousing — we’ll consider that the spoiler zone.  NOTE: I don’t get a lot of spam, but it’s still about as much as real comments, so your comment may have to get approved before posting.  I’ll be trying to keep up as much as I can.