The idea of a puzzle based on noughts-and-crosses came to me over ten years ago: I can’t be sure when. My first idea was that the grid would be 13 × 13 (that never changed) divided into 12 ‘games’, the symmetrically placed central lines spelling out a message and/or instruction. But what? I’d known the alternate US name ‘tic-tac-toe’ as it can be spelt; when I checked in Chambers and saw that the ‘e’ could be dropped, I had one of the lines: TIC-TAC-TO GAMES. At this stage I had decided that, to get the Oes and Xes there would have to be an alphanumerical substitution that would yield only 0 and 1, and that the preamble would have to include an instruction such as ‘multiply by ten, and make a further substitution’ to get the Xes from the 10s. But I lacked a 13-letter line to cross the one I had, one which crossed in the middle at T. Driving up to Sydney’s northern beaches — like many who read this, I suspect, I do a lot of thinking about crosswords while driving — I realized that MULTIPLY BY TEN had 13 letters, though its middle letter was not T. The fourth, however, would do the trick, though mildly upsetting the symmetry.

When I set to work in earnest I had my 13 × 13 grid with the fourth rank and the seventh file, and nothing else. How to fill it? I have had four test-solvers over the years, at least two of whom have asked why did I not make the substitution A = 0, B = 1, C = 0, D = 1, etc or similar, which would have placed no constraint on choice of letters; why the more elaborate mapping? I felt from the outset that that would draw solvers’ attention to the fact that the game involved was a binary one, and I wanted to avoid that. So I next considered A= 0, B = 1, C = 2, D = 0, etc, or some variant of it; there was no way of allotting numerical equivalents cyclically in groups of three without at least one vowel’s having the value 2, so eliminating it/them. I felt not to have a full palette of vowels, including Y, at my disposal would be too much of a constraint. After a bit of experimenting I settled on A = 1, B = 2, C = 1, D = 0, E = 1, etc, so eliminating from the fill B, F, J, N, R, V, and Z. I knew that N and R would be missed, but was not too unhappy about the others.

My idea for the original was that solvers would have to highlight the winning play in whichever games had one; the preamble stated ‘highlight the winning play or plays’. As most ‘games’ would not conform to an actual game of tic-tac-to (I rejected the idea of arranging words so that all games were theoretically feasible), for the good reason that each would have nine plays, that meant that I needed to allow for one unfinished game. I decided to place this critical game in the SE corner, leaving two cells empty. So I looked for a word for the bottom rank that had to have the last three letters yield numbers consistent with a tic-tac-to game. I used some program or application — I can’t remember what — to find a word whose middle letter of 13 was N. From the list I eliminated any with the outlawed letters, and found CHALCANTHITES, quite wonderfully containing ‘Chal can’t hit’: a sentence, rare enough, with clueing possibilities. I looked no further.

It must have been at about this stage that I decided that the method of entry would be different. I wanted to clue both description and instruction completely — ie without unches; the former could be partitioned into TIC-TAC, TOGA, and MES (should I use a French word? or did ME allow of a plural?) and the latter to MULTI-, PLY, BY, and TEN; or the last five letters as BYTE and N. I was not happy about MES or clueing either a one- or a two-letter entry. Dimitry — rather, one of his puzzles — gave me an idea.

During Easter 2007, I was holed up on a farm in Victoria with Dimitry’s Listener #3922 (thanks to Stephen Rice for identifying this). I do not now recall how it happened, or even what the solver had to do, but it must have been a carte blanche, because for some time I began entries in the wrong cells, not leaving sufficient space for the entries, and so had to spill over to the next line; eventually the penny dropped and I was able to complete as intended. But I remember thinking that it could be a useful way of constructing a puzzle. So this would be the puzzle in which I could use it: MES could become MESENTERY, say, or just MESA; there was no limit to the possibilities for N. I considered other break-downs: (FRAN)TIC, TACT, OGAM, ES(SE), say; (TU)MULT, I, PLY, BYTE, N( ) would have a one-letter entry again. I decided to stick with one of the original partitionings. The fill proceeded slowly, as I would at times carelessly include a forbidden letter, usually an N or an R. I should have liked the entries to have been consecutive without any intruding unches, or at least to begin so. My first grid had the first five ‘Across’, but only the first two ‘Down’, entries consecutive; in the one eventually published, three and four respectively (but would have been six had OPSIMATHIC been allowed). Given the method of entry, perhaps it is possible to construct such an unchless grid, but the constraint upon choice of letters increased the difficulty, and the task proved beyond me.

The puzzle sent to the first test-solvers, Dysart and Radix, was harder in that the clues were given without identifying in which rank or file each began. In addition, the preamble did not mention that there were two (at that time) empty cells; instead it stated that ‘127 cells are checked’, which enabled the arithmetically minded solver to deduce that two were empty. The vetters thought that it would be fairer to state that outright. The grid at the time had an unfortunate flaw: both I and A of TIC-TAC were, unintentionally, unches; this meant that, as the clue then stood, the entry could possibly be entered as TACTIC, which rendered the puzzle virtually unsolvable. Solvers were originally required to enter 48 (later, 96) bars; Radix’s much more precise suggestion, which I incorporated, was ’32 boundaries of three bars each — 16 across, 16 down — to create a number of regions’. He also prophetically queried the plural CHALCANTHITES, and suggested a change of title, from ‘The Winning Play/s’ to the present one. I’m not sure why many solvers dislike entering bars — in a grid, that is! — but I accept it, and so that requirement was removed.

I cannot be sure, but I think I submitted the puzzle for vetting in mid-2007; it entered the logjam that Roger and Shane inherited. In the meantime, another Listener puzzle which involved noughts-and-crosses appeared, in 2010 I think, and I feared that that would do for mine. When last year I received a vetter’s report, one of the comments was that the clues to each first entry — to CUISSE-MADAME, ‘Sadaam Hussein and mice chewed this nashi? Yes and no: wrong variety (12)’, and to CASTAWAY, ‘He may find company absent (8)’ — were too hard to make a start in what the other neatly called ‘a demi carte blanche’ (or was it ‘a demi-blanche carte’?). ‘Saddam’ is, of course, the accepted spelling, but I found surprisingly many ‘Sadaam’s on the internet, possible misspellings.

The unching was quite deliberately unXimenean: I felt that the difficulty in placing the entries called for greater checking than usual, though there are probably too many fully checked entries. After I had received the first report I noticed that, of the ten cells in a 2 × 5 block in the NE corner, six contained unches; that seemed unfair in a puzzle with unusual entry method, so I set about loosening it (there are now only two). The puzzle subsequently had another two testers; I think only two of the four mentioned the constraint on letter choice — the lipogrammatic aspect of most of the grid. I shall be interested to see whether many solvers comment on that.

I have received a number of complimentary comments on the clue to LOLIGO. My original was ‘Behold, progressing from head, ‘limbs’ eight (eight? two others) here (6)’. I hope the chief vetter, Roger Phillips, will not mind my revealing that his comment was ‘The things from which letters are to be progressively taken must be words: expecting the solver to take “two others” as a unit is unfair’, and tweaked the clue to the far superior published version: brilliant!

Ozzie

An incredibly detailed and fascinating setter’s blog by Mash can be seen at the following address on the Magpie site:

http://www.piemag.com/2011/11/27/the-4th-dimension-from-the-mind-of-mash/

If you enjoyed the puzzle and don’t already subscribe to Magpie, may I suggest that you do so without delay!

Mervyn Peake puzzle? Well, it’s been done, hasn’t it? There was that one back in CROSSWORD magazine 20+ years ago, and then that Inquisitor in 2008. Well, yes, now you come to mention it, both by me.

[Actually, the first one of those had some wonderful anagrams in it, inspired by observing that TITUS GROAN becomes OUTSTARING; this led to the discovery that GORMENGHAST, with the S replaced by U, can become GRANGEMOUTH, and TITUS ALONE (S -> X) transforms to EXULTATION; incidentally, none of this TEA and Sympathy malarkey in those days – it was all done by sweat of brow and brute force anagramming.
“Brute force anagramming? You were lucky – when I were a lad, I had to chisel out the individual letters of my anagrams from the cold granite.”
“A chisel? You had a chisel? Luxury…”]

Sorry, where was I?

Noting the centenary of Mervyn Peake was a reflex action, but finding that it fell on a Saturday upped the ante immediately. I drafted a grid quite quickly, but, while it was a perfectly fine puzzle idea (and we’ll come back to it in due course) it wasn’t that mysterious beast, a Listener puzzle. So I cast around for a further idea, and looked at the opening paragraph of the trilogy. The depiction of the castle there gave me the idea for the skyline effect – I particularly wanted one turret ‘point[ing] blasphemously at heaven’. That meant the puzzle would be carte blanche – pointless giving the outcome at the outset – which in turn led to some interesting solving challenges, and some thought into how to address them. Obviously, with a CB, you expect some hints about the symmetry to emerge (generally from the pattern of across word-lengths appearing as you solve), but here they would be deeply hidden by the fact that the first few clues would not adhere to any symmetry. However, making the main body of the puzzle symmetrical would still allow for some hints as to the pattern to emerge. Against that, I wanted some castellations to ensure you placed the Downs correctly. I may be very much on the side of symmetry in puzzles, but that doesn’t stop me wanting to play with the concept!

I also planned some very long entries – especially vertically – as anchors for the grid. But, as it happens, MERVYN PEAKE and GORMENGHAST are the same length (indeed, the R, N and A are shared, as that 2008 Inquisitor puzzle utilised), so two of my long answers were immediately unclued, and of much less value.

Obviously one of the first things you need to do for an anniversary puzzle is to ‘bag’ the date, and this I did. It turned out to be my last formal contact with Derek Arthur as an editor. His death was a tremendous shock. On a more practical level, the moratorium imposed while the editorship regrouped provided a worrying impediment to the timing. Not least, my earthquake puzzle moved to January this year – rather close to (indeed, slightly under) the six-month deadline for a setter’s repeated appearance if the Peake puzzle was to hit its anniversary. But the editors agreed to make an exception, even though, by this time, they had another CB for the preceding week – in fact, I had to change a corner of the original grid because of a clash of vocabulary between puzzles.

Editor 1 failed to finish, and came back for a hint (the division point between across and down clues) – which editor 2 decided against giving solvers, so I reckoned that this would appear as a tough test. What we all three managed to miss was the absence of a clear statement that the clues were in the right order, which, at the moment of writing (late afternoon on 9 July, NZ time), is causing agonies of indecision in a certain online forum.

Later note: having now seen comments from solvers, it’s also clear that the statement about numbers was less than helpful. I’ve tended to submit carte blanches with clues asterisked ever since I spotted (on a set of proofs) the unfortunate occurrence of a rare two-sentenced clue having its second sentence start on a fresh line – without numbers (or some other distinguishing feature), of course, this looks like a new clue. Numbers do help with notes and so on but (in good old retrospect) might not have been the best solution here. I have had a Listener CB with a preamble clearly stating that the clues were in random order (come to think of it, that was the one with the two-sentenced clue), but this wasn’t the moment for a repeat! [If you want to try this puzzle, click the image on the right; solution in a few days - Ed]

At some point I revisited the title. It was originally ‘Keep in View’ (told you it was a visual stimulus). Looking at that one day, I suddenly realised the obvious anagram and replaced it for submission, rather tentatively, as I already had wind of some of the publicity that would surround the centenary, and thought it might be a bit of a giveaway. Editor 1 loved it, so that was that. I do know one solver twigged the theme before solving a clue, but I gather that didn’t make it much easier.

Let’s just go back to that other puzzle I mentioned. I don’t like to waste ideas, and this was a perfectly good puzzle, so I finished it off, and it went in as an Inquisitor. Once again, I had to ‘bag’ the date and a reminder to that effect was my last exchange with Mike Laws. (So I am forswearing anniversary puzzles for a while at least.) But, for those of you who also do the Inquisitor, Plumrot is me, and the genesis of that pseudonym will accompany the solution notes.

As a happier coincidence, my copies of the centenary reissues of the Gormenghast books (including the newly discovered sequel by Peake’s widow) arrived today, on the date of the centenary itself. (And the puzzle, of course.)

This makes a third CB in sequence. Looking back, I’d note that each idea is strongly visual – quite unusual for me; each also plays slightly differently with symmetry. 50-50 had the concept of the nested Ls, which I didn’t want to show at the outset, (much the same as the visual element of A Keep); Heart relied heavily on the image of sliding two halves of the grid. You may be relieved to know that my next Listener idea has a grid with bars and numbers.

Phi

This puzzle elicited strong emotions, and following a little arm twisting, here are some remarks from the setter’s perspective. In context, recent Sabre puzzles (4115 Invisible Ink II; 4082 Pangrams; 4058 Whirly-Birly, etc.) have received numerous comments such as “What has happened to Sabre’s sting?”, “Disappointingly easy for this setter”, “Not the Sabre of yore”. The current puzzle was in some sense a reaction: I wanted to construct a testing puzzle, yet not an unfair puzzle. It is the easiest thing in the world to construct an impossible crossword, but to achieve a high level of difficulty while maintaining fairness is a very delicate balance, which I constantly struggle to get right.

Some agree that 4140 achieved this, and some would disagree. Knight’s moves provide an easy way of achieving the difficulty. I toyed with the idea of a 5 × 5 central square in a 13 × 13 grid in which cell entries had to be “knighted”, making a less fraught endgame. But as a general tenet, I dislike the 13-grid over the 12-grid if the theme doesn’t mandate the larger size. There should be some thematic words appearing from the jumbles, and as 6 × 6 took precedence in my thoughts, I immediately saw what the title of the puzzle should be, with its potential for a mischievous trap (I prefer “mischievous” to “mean”). Thus KNIG· · · · ·ANT was fixed, and the two knights GAWAIN and MODRED duly noted. Construction began with my determination to use both SOUTHERNWOOD (a smooth anagram), and BLOODIED with the misdirection of being able to use BOODIED=moped, which I hoped would raise a smile. (Jim Evans was my first Listener editor, whose comments were always tremendously helpful. One I recall is his criticism that my clues were “sterile and dry – get solvers to smile!”) Several other pre-planned words were worked in, and after more sheets of paper were thrown away than I care to number, the diagram was complete. Clueing was the usual pleasure, and after more revisions and rewrites than I can recall, the puzzle was submitted to the Listener vetters.

You forget about these things until your inbox registers a communication from listenercrossword.com, and with beating heart you open the e-mail… Rejection! Slowly it sinks in. The puzzle is essentially unsolvable. And of course, the criticisms of the editors are spot on. In this version of the puzzle, the “disaster zones” (vetters’ phrase) of grid bottom and grid right contained unchecked entries (at least, checked only by knight’s move answers); further, the only answers starting within the 6 × 6 square were on the perimeter of the square. With one or two clues unsolved the endgame was truly impossible (the fact that the vetters between them solved the puzzle is mightily impressive in retrospect!).

A correspondence ensued about whether I could save the puzzle, and how to achieve this. The upshot was to try and rework the “disaster zones” so as to make every answer checked in at least two letters; and to add in several more answers that started within the central 6 × 6 square, thereby reducing the multitude of possibilities for the endgame. This turned out not to be so easy: answers were not allowed to start on the diagonal because this would have upset my KNIGHT ERRANT trap; and it was surprisingly difficult to find words of length and quality using knight’s moves when I hadn’t planned for this. To cut a long story short, minor tweaking turned into major surgery, and only about 20% of the original grid survived. I was determined to keep SOUTHERNWOOD and BLOODIED, and also HOSTRY that had fortuitously presented itself during the initial composition. It was important too to include some less common letters: a solver might reasonably assume that two Xs share the same cell, softening the endgame. I tried for Js and Qs but could only manage W,X,Y. MODRED, poor fellow, had to change his direction, but eventually all came together. This time listenercrossword.com brought good news, and June 4 saw the puzzle’s appearance in the Times.

With hindsight, if I had the opportunity to make any changes, I would change the clue to CAPAS at 5D which caused much confusion with its alternative CAPES. I had simply not noticed this. The clue should have used wordplay for CA-PAS or CAP-AS etc, to eliminate the CAPES answer. I wonder about the clue for DINOTHERE, which many solvers said was the last to be solved, in many cases working backwards from a filled grid. Originally I had the clue as “Fossil elephant – tips of decayed ivory not present”, much easier than “Fossil elephant – tips of decayed ivory missing”. But then those jabs about Sabre’s sting kicked in. My apologies to those who found this a frustrating exercise, particularly to those who gave up; my thanks to those who relished the challenge and were generous with their compliments.

Finally, here is the logical argument submitited with the puzzle that demonstrates uniqueness of the placement of knight’s letters, independent of foreknowledge of GAWAIN and MODRED. After solving all clues (!), the grid is as follows:

We number the cells of the central 6 × 6 square by Cartesian coordinates with bottom left cell as (1, 1). We also think of this cell as black, with squares alternating in colour as on a chessboard. This allows parity considerations: in particular, that the W of SOUTHERNWOOD must fall on a white square. The first E of BORDEREAUX, the X of IXTLE, and the first L of HALLIAN are forced; followed by the R of OBTURATION, the A of RACINOS, and the I of NIRLY. We have the following grid:

If the second R of SURREYND falls in cell (5, 5), then the whole of SURREYND is determined, and there is no white square available for the W of SOUTHERNWOOD, with its corresponding sequence -HERNWO. Thus the second R of SURREYND falls in (1, 5). The Y of SURREYND must fall on a black square. If (5, 5), then -YND cannot be found; if (2, 2), then -YN cannot be found; and (2, 6) is unreachable. Thus the Y of SURREYND falls in (4, 2). Of the six possibilities (using parity) for the W of SOUTHERNWOOD, the following five are immediately ruled out: (2, 5), (4, 5), (6, 1), (6, 5) (for which no -HERNWO possible), and (2, 1) (which fully determines SOUTHERNWOOD, leaving no possibility for the H of SLOETHORN). Thus the W falls in (3, 4), which now determines the H of SLOETHORN. The grid is now:

The A of OBTURATION, the first E and O of REWORDED, are now forced, then the T of PERITRICHA, and second L of MANMILLINER. Finally, the N of HALLIAN and the U of BORDEREAUX complete the square.

Sabre

Maybe this one was not the most testing of challenges, but I’d hope that it provided adequate entertainment for the army of discerning Listener solvers. As a veteran of barely ten composed puzzles, it was indeed an honour to find one of my offerings considered suitable for Listener publication – however, I can’t claim complete credit for this, so first and foremost, a note of thanks:

Firstly, for their help and encouragement when I initially attempted setting, a big nod to Homer and Zero – their advice served to point me in the right direction (or away from the wrong one) at that early stage. Thanks also to subsequent testers of my efforts – and specifically in the case of Cross-reference to Chalicea and Radix. Lastly a doff of the topper in the direction of Chris, Dave, George and Shirley for their excellent solvers’ blogs, and to all who kindly take the time and trouble to submit feedback on any puzzles by whatever means – it is much appreciated, often constructively helpful, and of course adds an extra dimension to the whole effort.

Now for the puzzle itself – I was going to say that ‘the idea came to me …’ but this was more a case of ‘an idea came to me …’ since the finished article differed somewhat from the original seed. My primary intention was to construct a puzzle in which the words were more important than any kind of endgame (for which I’m afraid some simple highlighting had to suffice). I was toying with options for generating two answers from one clue when it occurred to me that I might instead use the clue to generate two words but then merge those back into a single answer. It was then a short step to decide that two half-words, run together in opposite directions, might generate some mileage, and may prove to be a bit different for solvers.

My first step was to start to compile a list of words which could work, and I soon recognised how many possibilities there might be. I also had grand ideas at that point to do something ‘clever’ with the discarded half words, like anagrammatise them and let those results appear maybe as extra words in other clues – for example, DEMIREPS and SUBLATED would combine to give SUBLIMED, but their discarded halves might additionally yield PREDATES. Sadly good intentions did not prevail, and that effort proved short-lived … my list of possible words shrunk significantly, and I was also stuck for any smart idea as to the purpose of those spare words (save in generation of an extra-word message, which threatened further layers of complexity and ingenuity which was, for me, in short supply at the time). I consequently opted just to lose the discards.

Just before constructing the grid, the Dolittle animal occurred to me as a means of indicating or reflecting the required treatment of the clues – that immediately obliged me to opt for 12×12 instead of my preferred 13×13, so I could neatly accommodate ‘Pushmi-Pullyu’. It also gave me a suitably vague working title of ‘In my life’ (as in Rex Harrison’s song from the Dolittle film musical).

I decided to focus a little more on the half-and-half elements … how about a number of ‘normal’ clues, then half of the remainder generating a message, and half being composite clues. I then hit on an idea of triple element clues for those composites, giving me a further opportunity for halving – seven would contain definitions to the two component words with wordplay only to the composite grid entry. The other seven would have a definition only of the grid entry with two sets of wordplay to the two component words. Indeed, after some effort, that’s how the first version ended up, was tested, tidied up and submitted.

The one casualty at that stage was the title – I had realised that I needed to use Hugh Lofting in the message, to point to his original creature (which was a gazelle-unicorn cross). The first film version had used the same animal, a llama, at both ends – so as a literary prototype, that would have just translated to palindromes in the puzzle. Hence, ‘In my life’ was out, and ‘Cross-reference’ took its place.

First vetting then caused me a major rethink. Roger suggested some significant improvements – bar off four original words to improve some under-unching, and switch all the triplet clues to pure full doubles, with full wordplay and definition to both component words (and no indication of the grid entry). He felt that the original version left some of the grid entries guessable with no strict obligation on the solver to solve all components of those clues. It made a lot of sense, consequently I got even more practice at (re)writing double clues.

After that major surgery, the puzzle passed to dear Derek, who had a bombshell in store for me. He only actually changed one clue in the puzzle, but did mention that he thought he had seen the theme before, and quite recently. I thought immediately that I was holed below the waterline, but an ensuing exchange of correspondence established that the other puzzle had merely served to leave the Lofting name fairly fresh in his mind, and that he thought there was still plenty to differentiate my offering. I subsequently managed to track the other puzzle down – it was in the Crossword Club mag (to which I have since begun to subscribe!). Fortunately, that puzzle, whilst referencing Pushmi-Pullyu, employed entire words which formed another word when written backwards, so my use of the two half-words was indeed a different treatment.

End of story – a few months’ wait, and there was my (first?) Listener – on almost the 125th anniversary of Hugh Lofting’s birth.

Nudd.
Remembering Derek Arthur and Mike Laws, two fine editors and setters. The crossword world is much the poorer without them.