I started thinking about this puzzle in early 2010. The main features of the final version were there from the start – a 10×10 grid with the PI IS 3 POINT in row 1 and TO 80 PLACES in row 10, with the 80 decimal places of pi in between. Well, almost – my original 80 digits ended 899 and it was quite a while (see below) before I realised that the rounding to 80 places meant they should actually end 900. I had in mind from the beginning that clues would produce clashing cell entries and that these would somehow decode to the ten digits to provide the final grid fill, with a further decoding required to produce the top and bottom row messages.
All that remained was to produce an initial grid fill that would lead to the above. I had a number of attempts over some months to find combinations of groups of letters which would produce enough valid across and down entries to fill the entire grid but I couldn’t make any of them work – I could get some of the way there but would then find chunks of space that just didn’t seem to provide any further means to progress. I abandoned the idea and worked on what seemed a more promising one, which produced System Analysts. When that one went off, I spent some more time producing Lawbreaker. With those both in the Listener queue, and more than a year after the original idea, I decided to take another look at the pi puzzle.
Looking at it afresh, I was immediately struck that I had never really thought about how to get from the groups of clashing letters to produce the required numbers. Obviously some sort of key was required and after a bit more consideration, using the title as a key seemed a good way to go. So, a ten letter title was required, with no repeating letter. Some internet research produced some possible candidates: methodical, aneuploidy, co-equality, educations, chloramine, scathingly. However that meant pinning certain letters to certain numbers and as some of the numbers appear more often (9 the most, then 3, then 0, 2 and 8 equally), the vowels often ended up attached to less useful numbers. So initial attempts at grid fills foundered, but with a bit of persistence I did manage to make “scathingly” work – 23 across clues, 22 down clues and all 100 cells filled.
I thought about it for a while and decided I didn’t like it – there were a lot of 3 letter answers, average answer length was just 3.8, and number 7 had a group of four letters associated with it and I would have preferred all of them to have either two or three. I had at least got every letter of the alphabet in, but some only once, and some of the clashes appeared only once, so it felt like I was asking a lot of solvers. Actually I was asking even more: because I didn’t want clue numbers cluttering up the grid which would end up containing a lot of numbers, I had left the grid as a carte blanche so solvers would have to work out where entries went too. It was at this point too that I realised I had the pi digits ending 899 rather than 900. I was able to fix that with a couple of grid entry changes, but the whole thing was feeling like it was a bit of a mess! The only upside was that the 899/900 ending addressed a concern I had that a solver who spotted 14159 … in row two might be able to just write out the numbers of pi to submit the solution without having solved the majority of clues (Or so I thought, until I saw Jaguar’s LWO blog!)
So another try – what about a two word phrase, if one word isn’t working? Flying shot, stamp hinge, crying wolf, crazy quilt – all tried, none producing anything better. Then “stormy wind” – that produced a grid with 41 entries, average entry length 4.0 and all 100 cells filled. It felt reasonable enough that I wrote the clues for the puzzle, including a hidden message to encode the letter groups. “Mess” was one of the entries, so I made its clue the only normal clue, with a suggestion in the preamble that it provided a hint (as a synonym of “pi”).
Despite getting to this stage, I wasn’t happy on reflection. There wasn’t enough checking of clashing letters, some letters still appeared only once and solvers still had a carte blanche to contend with. “Mess” was beginning to seem only too apt. And I wasn’t happy with the title – it was just too arbitrary and wasn’t even a proper phrase. So I gave up again.
But I came back to it again after a while. Maybe the mess synonym for pi had percolated a little, because I started considering whether synonyms for pi could form a title. Maybe “crazy fonts”? Saintly fog? Godly snafu? Prig symbol? Holy figure? Moral type (with the missing tenth letter group being zero)? “Godly mix-up” seemed to be the most promising …
But I couldn’t get a grid to work with it. With across and down entries, I couldn’t even find a way to fill all 100 cells. Time to cheat then – who says entries can’t go left as well as right, up as well as down? I tried some of that, and it was getting more promising. However, now I’d given myself more leeway, I felt more of an onus to produce a grid that gives solvers a fair chance – so I decided every letter of the alphabet had to appear, and had to appear at least twice. Also every cross-check within the letter groups had to appear at least twice, so not being able to solve just one clue shouldn’t leave the solver stumped. With those constraints, even four directions proved not to be enough, especially in the corners of the grid. So now I decided to allow myself diagonal entries too. With all that in place, I was able to produce a grid with average word length up to 4.5. I was reasonably happy, having met the requirements I set myself and with the title giving some PDM potential.
So time to write some clues and work out how to tell solvers what needs to happen. Using superfluous words to create a message had been used by another Listener puzzle around this time, so I shamelessly stole that idea but decided to do it in only half the clues to make it a little trickier. Using the resulting grid as a key to produce the final stage seemed an obvious device and one that hopefully wouldn’t give the game away to solvers until they reached this stage. That did mean the superfluous words had to give two messages, the first to use the title as a key to encode the letter groups as digits zero to nine and then secondly to use the grid numbers to find the right letter in each clue, which produced the third message to recode seventeen cells in top and bottom rows. I still didn’t want clue numbers appearing in the grid but I didn’t think I could justify a carte blanche where solvers had to deduce not only the starting point of entries but also which of eight directions they went in, especially given the number of clashes – so I provided co-ordinates around the grid, with clues giving the starting point and a direction. A few weeks of working on clues to work within those constraints and it was done, just two and a half years after starting work on the original idea. I pressed the send button on the e-mail and waited.
A month later, I got an e-mail but it was to request some revisions to System Analysts, which had then been submitted over two years previously (and overtaken by Lawbreaker which went in a year later and had got through to publication relatively quickly). There was about a month of discussion over six of the clues on that one (quite a high proportion out of only sixteen!) and then that was good to go.
Forward about fifteen months, and the first feedback arrived. You may have already spotted from the above that changes were needed. Some of the superfluous words weren’t clear and neither were the two resulting messages. And the co-ordinate system for locating grid entries was deemed unnecessary. As ever, the feedback was extremely helpful , especially in suggesting that the first step (using the title as key) could be given in the preamble. I took on board the issues with superfluous words and ditched that device in favour of the more usual extra letter in wordplay. This just meant rethinking the message and then rewriting every clue (and I thought six out of sixteen was a high proportion earlier!) which took a few weeks as it coincided with a busy work period. A few more revisions and – as ever – a number of improvements from both editors resulted in the published puzzle. I’m very grateful to both editors for their feedback and help.
It just remains to thank all those who have commented on this site and others and of course via John Green (whose marvellous compilation of the feedback and report on errors makes setting a Listener feel even more of a privilege). I’m glad that most solvers enjoyed the challenge but I’m also grateful for the various steers on how it could have been better.