I’ve greatly enjoyed solving some of the Christmas Listeners over the years, so have thought for a while about a suitable theme for a Christmas puzzle. On holiday last summer I had printed out from the Times site some randomly selected Listeners that predated when I started solving the puzzle seriously in 2005 or so. One of these was Double Carte Blanche by Phi (Listener 3760 from February 2004). I won’t spoil the theme for anyone who wishes to solve this puzzle, but it involved two grids which had to be placed on top of one another. I was impressed with this, and it was whilst thinking about possible Christmas themes that these two things coincided and I thought of setting a puzzle that turned into an Advent Calendar, where ‘doors’ would be opened in one grid to reveal letters in the second grid, and thus a seasonal message.

The first, and most obvious, hurdle was the fact that the grids could not be too large, as the space in The Times is constrained. I sat down to try to come up with a single grid with a suitable message that I was confident would fit into a reasonable sized grid. I figured that any message presented horizontally would have to be on alternate lines so as to avoid the grid simply falling apart once doors had been cut out. With a message that would probably have to include the words ‘GREETINGS’ or ‘CHRISTMAS’, the minimum width of the grid would have to be nine letters. After much thinking and playing around, I settled on the message SEASONS GREETINGS TO ALL SOLVERS AND SETTERS, knowing that I had a bit of play in there so that I could switch the words SOLVERS and SETTERS around if I ran into trouble. This also had the plus point that it was almost symmetrical about the central column. I toyed with having ‘TO ALL’ displayed without a space in between, but it seemed tidier not to.

Once I had the message determined, and had thought that this fitted nicely into a 9 by 11 grid, the next step was to work out how to generate numbers in the first grid. There would have to be 24 of them, and the obvious way to do this was by having clashes and taking the numerical difference of the letters. At this point, a few things became apparent. The first was that this would make life pretty tricky for the solver (that was a lot of clashes to fit into a 99 cell grid – almost a quarter of the cells), and secondly, it was going to be impossible to construct a grid whereby each of the clash cells would reveal a letter contained in the ‘SEASONS GREETINGS…..’ message. The message had 38 cells, and I would need more than this to enable each clash cell to reveal a thematic letter. However, I really wanted the clashes to be in symmetrically placed cells, so I could help the solver a little bit with locating these.

At this point I paused for a week or so, and tried to decide whether I should go for a larger grid after all, even though this would take it out of the realms of The Listener, and by necessity to The Magpie, where larger grids would probably be acceptable. In the end, I decided to press on with the original plan, reasoning that the revealing of the thematic letters would, assuming that only a single grid would be submitted, be a kind of check that solvers had also completed the ‘underneath’ grid.

By this time I was ready to start grid construction, and that was when the problems really started. Some setters, such as Elap, have programming knowledge that means they can write programmes to generate complex grids. Radix uses the Mathematica program to do this sort of thing. I, however, am not up to that, so sat down, quite literally, with pencil and squared paper, to come up with grid A. I thought it wouldn’t be too difficult to come up with a grid that contained 24 clashes of the required type.

Boy, was I wrong.

It took three months, on and off, of working on this, to come up with a suitable grid. Many times I reached the point of having 16 or 17 clashes sorted out, and then simply ran out of words to generate the rest. I almost gave up, but battled on, and after certainly more than 100 hours work, I came up with what I thought would turn out to be the final grid A. I breathed a sigh of relief, went to bed very happy – and then woke up next morning, came to check what I had done, and found that X-A was not 24, as I had assumed towards the end of the fill, but was in fact 25. At this point I did give up for a few days, and decided that I just wasn’t up to the job. In the end, my wife asked me why I was so grumpy. I told her, and she, understandingly, said that she was sure I could sort it out. I’m glad I listened, as the mistake actually turned out to be relatively easy to sort out. A couple of hours later, grid A was complete.

Grid B, by comparison, was a straightforward fill, with the only thing to consider being the letters contributing towards the thematic message. Before starting, I decided on my instruction to solvers, which would be:

PLACE A OVER B AND OPEN DOORS

Hence I needed 23 entries in the grid. This was duly achieved:

Readers at this point may realise that the above is not actually what finally appeared in The Times!

I was conscious that grid B would be a lot easier for solvers to complete than would grid A, so I decided to use a clueing gimmick that would hopefully not allow the instruction to appear to be ascertained too easily by solvers. This was a simple variation on a theme, whereby superfluous letters would need to be arranged according to the alphabetical order of the actual answers. I would have preferred to use misprinted definitions, my current favourite gimmick, but these clues often end up being quite long due to the limitations of the misprints, and I knew that clues had to be concise. It was time to start writing clues.

This was reasonably challenging, as a) I seemed to have some difficult words (TREKSCHUITS!), b) I wanted to make clues in grid A reasonably straightforward, and c) with 54 clues to write, they all had to be fairly concise so as to not go over the limit on the number of words that can be used in a puzzle. This kept me busy for a few weeks (although I did later wonder why I worried, when a Pieman puzzle was published that had 80 clues or thereabouts).

The clues written, my test solver turned things round fairly quickly, and it was time to e-mail the puzzle to the Listener editors.

In due course I heard back from the first vetter, and the news was good… kind of. He was happy with the puzzle to go forward as it was, but had concerns over the unching in grid B, where EGURGITATE and ASSISTANTS had 5 unches, too many for 10 letter entries. The vetter also had a few concerns over the interpretation of the thematic message and thus the denouement.

After some discussion, two new entries were added, GYP and RUT:

This, however, meant that the thematic message would have to change to be 25 letters long. We agreed on the (hopefully) clearer message of CUT OUT DOORS AND PASTE A OVER B (as it would be impossible to cut out doors after pasting A over B). This was better, but unfortunately meant that I had to rewrite almost all of the clues for this grid. I toyed briefly with asking the first vetter to just stick with the original version of the puzzle (after all, he had been happy to do this), but I figured that the effort involved would result in a much better puzzle, so started writing clues again. This took a couple of weeks, but finally the puzzle was complete.

Of all the puzzles that I’ve set to date, this one probably took the most effort, so it’s good to see that it has been mostly well received by solvers. Apologies to those who found cutting out the doors to be a rather fiddly process, and I definitely owe the statistician a beer or two for the extra problems this must have caused him when checking submissions.