Solution: Learning from Our Old Mistakes

Answer: MISSISSIPPI ADMITTED TO UNION

Written by Jonah Ostroff, Justin Melvin

When most teams first open the puzzle, they are confronted with a list of "errata" for other puzzles they've seen in the hunt. The errata are all of the form "[time stamp] [puzzle name]: The clue [clue] should instead say [clue]," usually with multiple such errata listed for each puzzle.

These clues don't actually appear in the puzzles, but we can solve them anyway. They always yield a pair of words that are one letter apart. Reading the "corrected" letters spell "SEE WRONG ANSWERS; FIND SHIFT FOR EACH LETTER; THEN SHIFT TITLE OF THIS", possibly repeating many times.

The time stamps and puzzle names are exactly the instances where the team submitted a wrong answer to a puzzle earlier in the hunt. Furthermore, the number of corrections for each puzzle corresponds to the number of letters in each wrong answer.

The title of the puzzle is 26 letters long, so we need to decide how to shift those 26 letters. By comparing the letters in the wrong answers to the letter changes in the clues, solvers can determine that each letter in the alphabet corresponds to a specific shift. For example, every erratum corresponding to an "A" has a shift of 1, while every erratum corresponding to a "B" has a shift of 4.

Here are all 26 shifts:

Shifting the title by the 26 shift amounts in order yields the answer: MISSISSIPPI ADMITTED TO UNION

Author’s Notes

Jonah: Some recent hunts have experimented with puzzles whose content is somehow customized to each team. As soon as we decided to have a History department, I wanted a puzzle based on things your team had done earlier in the hunt. Wrong answers seemed like a fun data source, especially since there’s an easy way to generate more of them. This was a weird enough idea that we were willing to write 478 crossword clues for it.

Justin: One of the difficulties with this particular puzzle on the tech side was the number of weird interactions with other puzzles from the hunt that had special considerations: the CS round (what exactly constitutes the team's wrong answer when there is an input and an output componenet?), English (how do we handle wrong answers with multiple answers?), and Hydra (since there are really only a hundred or so 'real' puzzles in the system which wrap around, wrong submissions might be misleading). Late in one of the betas, we found that we were getting server errors because we missed adding an entry for a particular Q bigram that only came up when one team had some particularly odd answer submissions. But overall the puzzle was a fun challenge to put together.

Appendix

For completeness, here are all the possible clues: