Monday, March 19, 2012

ACPT 2012 By The Numbers

Here comes another year's worth of score data (thanks due to some combination of Will S, Doug H, Matt G, and probably some other people for getting the full data on the web) to play with.  First the median scores (plus my scores because, well, it's my blog darnit):
Puzzle 1: median 1065, corresponding to finishing perfectly with 7:xx on the clock (me: 1165, 7th overall for the puzzle)
Puzzle 2: median 800, corresponding to 80 out of 100 words with 0:00 on the clock (me: 1385 -- 20:xx on the clock, but 3(!) errors in 4 words)
Puzzle 3: median 1160, corresponding to 116 out of 118 words with 0:00 on the clock (me: 1730 -- clean with 16:xx on the clock)
Puzzle 4: median 1255, corresponding to finishing perfectly with 13:xx on the clock (me: 1330)
Puzzle 5: median 680, corresponding to 68 out of 102 words with 0:00 on the clock (me: 1420 -- clean with 10:xx on the clock)
Puzzle 6: median 1575, corresponding to finishing with one error and 16:xx on the clock (me: 1700 -- one error with 21:xx on the clock)
Puzzle 7: median 1875, corresponding to finishing perfectly with 13:xx on the clock (me: 2450 (36:xx), 9th overall for the puzzle)

So the typical contestant finished four of the seven puzzles, albeit one incorrectly (down one finished-but-wrong puzzle from last year).  The completion/"oops" statistics are:
Puzzle 1: 92.4%/17.2%
Puzzle 2: 41.3%/35.7%
Puzzle 3: 51.0%/36.2%
Puzzle 4: 97.8%/13.9%
Puzzle 5: 26.7%/32.7%
Puzzle 6: 88.9%/64.2%
Puzzle 7: 76.4%/35.6%
In each of these, the first numbers are the percentage of solvers (who turned in a paper) who completed the puzzle (defined as either time on the clock or 0 errors+0 time), while the second is the "oops!" rate: the percentage of the first group who had an error (by def they are people who had time on the clock when they turned it in).  For comparison, last year's completion numbers were 97/50/67/88/17/95/69, so it looks like puzzle 3 got tougher and 5 got easier (although that's definitely a relative term).  Puzzle 6 had a huge error rate, and we all know which square that was.  The completion stat for puzzle 2 is a lot better than I thought it was just looking around the ballroom (that's my poor visual estimation skills again).  The error rates around 35% seem a bit high, but I have no historical data (I didn't quite extract the same data in previous years) to compare it with.

Other score notes:  I added a column to the spreadsheet calculating how many points each person "gave away" on errors.  My 460 points given away was not, this year, enough of a difference to get back on stage, which means I will need to pick up speed again.  And congratulations and/or commiserations to Benjamin Aisen who was the lowest-ranked scorer to make no errors, finishing 132nd (top of the D division!).

Maybe one year I'll make some graphs.  If you have a graph you'd like to see let me know.

No comments: