Monday, March 21, 2011

ACPT: The Numbers (2011)

This year I downloaded the full-version, extra-deluxe, double-the-cost version of the scores spreadsheet (thanks Doug Heller for making it available!) which has for everybody not just the points they scored but how they got there (wrong letters, wrong words, minutes on the clock). Granted they will still be making corrections (one of my errors was inexplicably un-highlighted, so I know there's at least one correction to make), but I can have some more confidence in my formulas. Maybe I'll make some pictures. (If you have some requests for pictures, then let me know.)

655 people were there for at least one puzzle, so my current 59th (after making above correction) will still put me in the top 10%, so at least I didn't reach my anti-goal. I missed all of my actual goals though (11500, B finals, and/or no mistakes, which when it comes down to it are pretty much the same goal at this point). The puzzles themselves definitely had a broader difficulty range, but skewing harder rather than easier (much unlike last year). I felt there was only one real clunker (#6; sorry Maura, but don't lie to me in the blurb: those puns were neither names nor from tomorrowland).

Puzzle #1: median score 1130 (corresponding to finishing correctly with 8:xx on the clock); 84% solved this correctly. I finished with 12 on the clock and 1230 points, which was conveniently enough also the top score in the room for this puzzle.
Puzzle #2: median score 870 (the puzzle proper, with no bonuses, is 890 points, so this is equivalent to one blank square with no time on the clock); 38% solved this correctly. I finished with 17 on the clock and 1465 points; top score was 20 on the clock and 1540 points.
Puzzle #3: median score 1390 (corresponding to finishing correctly with 4:xx on the clock); 50% solved this correctly. I finished with 19 on the clock and 1765 points; top score was 25 on the clock and 1915 points.
Puzzle #4: median score 1135 (corresponding to finishing correctly with 9:xx on the clock); 73% solved this correctly. I finished with 15 on the clock and 1285 points; top score was 17 on the clock and 1335 points.
Puzzle #5: median score 380 (corresponding to getting 38 correct words during the time limit); 12% solved this correctly. I finished with 13 on the clock, but 2 errors, and 1155 points; top score was 21 on the clock and 1595 points.
Puzzle #6: median score 1715 (which is not a score you can get to by finishing correctly, but still requires finishing with plenty of time -- one error and 19:xx on the clock is the easiest way to get there); 55% solved this correctly. I finished with 22 on the clock, but 3(!) errors for only 1695 points; top score was 25 on the clock and 2035 points.
Puzzle #7: median score 1670 (again requires an error -- two errors (in one word) and 12:xx on the clock will do it); 32% solved this correctly. I finished with 32 on the clock, but an error, for 2155 points; top score was 37 on the clock and 2475 points.

Looking at this, that means the median solver (well, the median ACPT solver, at any rate) finished five puzzles (but made errors on two of those) and couldn't get through #2 or #5. In fact, going through gives the following almost-but-not-quite "completion" statistics (presumably this doesn't count people who finish in the last minute, as they will have a 0 for their time-remaining just like people who didn't finish. Arguably this is a flaw in the scoring but I doubt it's at the top of anybody's priority list):
Puzzle #1: 632/653 97%
Puzzle #2: 327/653 50%
Puzzle #3: 436/652 67%
Puzzle #4: 575/651 88%
Puzzle #5: 113/651 17%
Puzzle #6: 617/651 95%
Puzzle #7: 424/616 69%

As you can see puzzles 1-4 all had no more than a 15% "oops!" rate, in terms of the difference between percentage finishing and percentage being all correct. It was rather closer for puzzle 5, probably because of so few finishers (113 finishers, and 81 people correct). In contrast, the "oops!" rate for puzzle 6 is 40%, and for puzzle 7 37%. How much of that is fatigue, and how much to the quality/trickiness of the puzzles is a little hard to get at. (For comparison purposes, last year's "oops!" rates:
Puzzle #1: 13%
Puzzle #2: 19%
Puzzle #3: 16%
Puzzle #4: 18%
Puzzle #5: 4%
Puzzle #6: 24%
Puzzle #7: 24%
so a little larger for the beginning puzzles, and smaller for the last two.)

1 comment:

Howard B said...

Good to see you again this year. Sorry to get in late.
That's quite an analysis, and is consistent wit my extremely-unofficial polling of people across experience levels and divisions.

It also makes me feel better that my response to new competitors asking about the dreaded puzzle 5 was that only about 10-15% of the room would finish it correctly if at all, so if they ran into trouble they would not be alone :).

Turns out I was 1 minute faster on puzzle 5 than I thought. I should check the scans to see if any errors were missed though, haven't done that at all.

Anyway, good year, and let's hope next time that we finish it clean. Congrats especially on that 1st puzzle. That's some speed!